/* MM-align: complex structure alignment by TM-score superposition.
 * Please report issues to yangzhanglab@umich.edu
 *
 * References to cite:
 * Srayanta Mukherjee, Yang Zhang. Nucleic Acids Research 2009; 37:e83 
 *
 * DISCLAIMER:
 *  Permission to use, copy, modify, and distribute this program for any
 *  purpose, with or without fee, is hereby granted, provided that the
 *  notices on the head, the reference information, and this copyright
 *  notice appear in all copies or substantial portions of the Software.
 *  It is provided "as is" without express or implied warranty.
 *
 * =========================
 * How to install the program
 * =========================
 * The following command compiles the program in your Linux computer:
 *
 *     g++ -static -O3 -ffast-math -lm -o MMalign MMalign.cpp
 *
 * The '-static' flag should be removed on Mac OS, which does not support
 * building static executables.
 *
 * MMalign.cpp does not natively compile on Windows OS due to its lack of
 * POSIX support. Nonetheless, MMalign.cpp is fully tested and known to
 * work on Windows Subsystem for Linux (WSL2) for Windows 10 onwards.
 *
 * ===================
 * How to use MM-align
 * ===================
 * You can run the program without argument to obtain the document.
 * Briefly, you can compare two structures by:
 *
 *     ./MMalign structure1.pdb structure2.pdb
 *
 * ==============
 * update history
 * ==============
 * 2019/04/08: A C/C++ code of MM-align was constructed by Chengxin Zhang
 * 2019/09/09: Fix bug in output display where unalign chains were missing.
 * 2019/10/06: Fix bug in homo-oligomer alignment.
 * 2019/10/07: Combine all codes into a single cpp file. A code block from
 *             pstream.h (by Jonathan Wakely) is included for compressed
 *             file reading. This code block and its original license notice
 *             are marked in this file.
 * 2019/10/13: Refine chain assignment for dimer alignment.
 * 2019/10/16: Add new option to read multi-model structure file
 * 2019/10/21: Show model index in output for -ter 0.
 *             Add -mirror option for aligning mirrored structures.
 * 2021/08/16: Rewrite dynamic programming for non-cross-chain alignment.
 */

#include <stdio.h>
#include <stdlib.h>
#include <math.h>
#include <time.h>
#include <string.h>
#include <malloc.h>
#include <float.h>
#include <sstream>
#include <iostream>
#include <iomanip>
#include <fstream>
#include <vector>
#include <iterator>
#include <algorithm>
#include <string>
#include <iomanip>
#include <map>

using namespace std;

void print_version()
{
    cout << 
"\n"
" **********************************************************************\n"
" * MM-align (Version 20210816): complex structure alignment           *\n"
" * References: S Mukherjee, Y Zhang. Nucl Acids Res 37(11):e83 (2009) *\n"
" * Please email comments and suggestions to yangzhanglab@umich.edu    *\n"
" **********************************************************************"
    << endl;
}

void print_extra_help()
{
    cout <<
"Additional options:\n"
"    -fast    Fast but slightly inaccurate alignment\n"
"\n"
"    -dir1    Use a list of PDB chains listed by 'chain1_list' under\n"
"             'chain1_folder' as all chains for the first complex.\n"
"             Note that the slash is necessary.\n"
"             $ MMalign -dir1 chain1_folder/ chain1_list complex2\n"
"\n"
"    -dir2    Use a list of PDB chains listed by'chain2_list'\n"
"             under 'chain2_folder' as all chains for the second complex.\n"
"             $ MMalign complex1 -dir2 chain2_folder/ chain2_list\n"
"\n"
"    -suffix  (Only when -dir1 and/or -dir2 are set, default is empty)\n"
"             add file name suffix to files listed by chain1_list or chain2_list\n"
"\n"
"    -atom    4-character atom name used to represent a residue.\n"
"             Default is \" C3'\" for RNA/DNA and \" CA \" for proteins\n"
"             (note the spaces before and after CA).\n"
"\n"
"    -mol     Types of molecules to align\n""Molecule type: RNA or protein\n"
"             auto   : (default) align both proteins and nucleic acids\n"
"             protein: only align proteins\n"
"             RNA    : only align nucleic acids (RNA and DNA)\n"
"\n"
"    -split   Whether to split PDB file into multiple chains\n"
"             2: (default) treat each chain as a seperate chain (-ter should be <=1)\n"
"             1: treat each MODEL as a separate chain (-ter should be 0)\n"
"                and joins all chains in a MODEL into a single chain.\n"
"\n"
"    -outfmt  Output format\n"
"             0: (default) full output\n"
"             1: fasta format compact output\n"
"             2: tabular format very compact output\n"
"            -1: full output, but without version or citation information\n"
"\n"
"    -TMcut   -1: (default) do not consider TMcut\n"
"             Values in [0.5,1): Do not proceed with TM-align for this\n"
"                 structure pair if TM-score is unlikely to reach TMcut.\n"
"                 TMcut is normalized is set by -a option:\n"
"                 -2: normalized by longer structure length\n"
"                 -1: normalized by shorter structure length\n"
"                  0: (default, same as F) normalized by second structure\n"
"                  1: same as T, normalized by average structure length\n"
"\n"
"    -mirror  Whether to align the mirror image of input structure\n"
"             0: (default) do not align mirrored structure\n"
"             1: align mirror of chain1 to origin chain2\n"
"\n"
"    -het     Whether to align residues marked as 'HETATM' in addition to 'ATOM  '\n"
"             0: (default) only align 'ATOM  ' residues\n"
"             1: align both 'ATOM  ' and 'HETATM' residues\n"
"\n"
"    -infmt1  Input format for complex1\n"
"    -infmt2  Input format for complex2\n"
"            -1: (default) automatically detect PDB or PDBx/mmCIF format\n"
"             0: PDB format\n"
"             1: SPICKER format\n"
"             2: xyz format\n"
"             3: PDBx/mmCIF format\n"
    <<endl;
}

void print_help(bool h_opt=false)
{
    print_version();
    cout <<
"\n"
"Usage: MMalign complex1.pdb complex2.pdb [Options]\n"
"\n"
"Options:\n"
"    -a    TM-score normalized by the average length of two structures\n"
"          T or F, (default F)\n"
"\n"
"    -m    Output MM-align rotation matrix\n"
"\n"
"    -d    TM-score scaled by an assigned d0, e.g. 5 Angstroms\n"
"\n"
"    -o    Output the superposition of complex1.pdb to MM_sup.pdb\n"
"          $ MMalign complex1.pdb complex2.pdb -o MM_sup.pdb\n"
"          To view superposed full-atom structures:\n"
"          $ pymol MM_sup.pdb complex2.pdb\n"
"\n"
"    -full Whether to show full alignment result, including alignment of\n"
"          individual chains. T or F, (default F)\n"
"\n"
"    -ter  Whether to read all MODELs in a multi-model structure file\n"
"          1: (default) only read the first model, recommended for alignment\n"
"             of asymetric units.\n"
"          0: read all MODEL, recomended for alignment of biological\n"
"             assemblies, i.e., biological units (biounits).\n"
"\n"
"    -v    Print the version of MM-align\n"
"\n"
"    -h    Print the full help message\n"
"\n"
"    (Options -a, -d, -m, -o won't change the final structure alignment)\n\n"
"Example usages:\n"
"    MMalign complex1.pdb complex2.pdb\n"
"    MMalign complex1.pdb complex2.pdb -d 5.0\n"
"    MMalign complex1.pdb complex2.pdb -a T -o complex1.sup\n"
"    MMalign complex1.pdb complex2.pdb -m matrix.txt\n"
    <<endl;

    if (h_opt) print_extra_help();

    exit(EXIT_SUCCESS);
}


/* The following code block (2255 lines) is from pstream.h.
 * It is included for reading gzip and bz2 compressed files.
 * The original license of pstream.h is attached below:
 *
 * PStreams - POSIX Process I/O for C++
 * Copyright (C) 2001 - 2017 Jonathan Wakely
 * Distributed under the Boost Software License, Version 1.0.
 * (See accompanying file LICENSE_1_0.txt or copy at
 * http://www.boost.org/LICENSE_1_0.txt)
 *  
 * @file pstream.h
 * @brief Declares all PStreams classes.
 * @author Jonathan Wakely
 *
 * Defines classes redi::ipstream, redi::opstream, redi::pstream
 * and redi::rpstream.
 */

#ifndef REDI_PSTREAM_H_SEEN
#define REDI_PSTREAM_H_SEEN

#include <ios>
#include <streambuf>
#include <istream>
#include <ostream>
#include <string>
#include <vector>
#include <algorithm>    // for min()
#include <cerrno>       // for errno
#include <cstddef>      // for size_t, NULL
#include <cstdlib>      // for exit()
#include <sys/types.h>  // for pid_t
#include <sys/wait.h>   // for waitpid()
#include <sys/ioctl.h>  // for ioctl() and FIONREAD
#if defined(__sun)
# include <sys/filio.h> // for FIONREAD on Solaris 2.5
#endif
#include <unistd.h>     // for pipe() fork() exec() and filedes functions
#include <signal.h>     // for kill()
#include <fcntl.h>      // for fcntl()
#if REDI_EVISCERATE_PSTREAMS
# include <stdio.h>     // for FILE, fdopen()
#endif


/// The library version.
#define PSTREAMS_VERSION 0x0101   // 1.0.1

/**
 *  @namespace redi
 *  @brief  All PStreams classes are declared in namespace redi.
 *
 *  Like the standard iostreams, PStreams is a set of class templates,
 *  taking a character type and traits type. As with the standard streams
 *  they are most likely to be used with @c char and the default
 *  traits type, so typedefs for this most common case are provided.
 *
 *  The @c pstream_common class template is not intended to be used directly,
 *  it is used internally to provide the common functionality for the
 *  other stream classes.
 */
namespace redi
{
  /// Common base class providing constants and typenames.
  struct pstreams
  {
    /// Type used to specify how to connect to the process.
    typedef std::ios_base::openmode           pmode;

    /// Type used to hold the arguments for a command.
    typedef std::vector<std::string>          argv_type;

    /// Type used for file descriptors.
    typedef int                               fd_type;

    static const pmode pstdin  = std::ios_base::out; ///< Write to stdin
    static const pmode pstdout = std::ios_base::in;  ///< Read from stdout
    static const pmode pstderr = std::ios_base::app; ///< Read from stderr

    /// Create a new process group for the child process.
    static const pmode newpg   = std::ios_base::trunc;

  protected:
    enum { bufsz = 32 };  ///< Size of pstreambuf buffers.
    enum { pbsz  = 2 };   ///< Number of putback characters kept.
  };

  /// Class template for stream buffer.
  template <typename CharT, typename Traits = std::char_traits<CharT> >
    class basic_pstreambuf
    : public std::basic_streambuf<CharT, Traits>
    , public pstreams
    {
    public:
      // Type definitions for dependent types
      typedef CharT                             char_type;
      typedef Traits                            traits_type;
      typedef typename traits_type::int_type    int_type;
      typedef typename traits_type::off_type    off_type;
      typedef typename traits_type::pos_type    pos_type;
      /** @deprecated use pstreams::fd_type instead. */
      typedef fd_type                           fd_t;

      /// Default constructor.
      basic_pstreambuf();

      /// Constructor that initialises the buffer with @a cmd.
      basic_pstreambuf(const std::string& cmd, pmode mode);

      /// Constructor that initialises the buffer with @a file and @a argv.
      basic_pstreambuf( const std::string& file,
                        const argv_type& argv,
                        pmode mode );

      /// Destructor.
      ~basic_pstreambuf();

      /// Initialise the stream buffer with @a cmd.
      basic_pstreambuf*
      open(const std::string& cmd, pmode mode);

      /// Initialise the stream buffer with @a file and @a argv.
      basic_pstreambuf*
      open(const std::string& file, const argv_type& argv, pmode mode);

      /// Close the stream buffer and wait for the process to exit.
      basic_pstreambuf*
      close();

      /// Send a signal to the process.
      basic_pstreambuf*
      kill(int signal = SIGTERM);

      /// Send a signal to the process' process group.
      basic_pstreambuf*
      killpg(int signal = SIGTERM);

      /// Close the pipe connected to the process' stdin.
      void
      peof();

      /// Change active input source.
      bool
      read_err(bool readerr = true);

      /// Report whether the stream buffer has been initialised.
      bool
      is_open() const;

      /// Report whether the process has exited.
      bool
      exited();

#if REDI_EVISCERATE_PSTREAMS
      /// Obtain FILE pointers for each of the process' standard streams.
      std::size_t
      fopen(FILE*& in, FILE*& out, FILE*& err);
#endif

      /// Return the exit status of the process.
      int
      status() const;

      /// Return the error number (errno) for the most recent failed operation.
      int
      error() const;

    protected:
      /// Transfer characters to the pipe when character buffer overflows.
      int_type
      overflow(int_type c);

      /// Transfer characters from the pipe when the character buffer is empty.
      int_type
      underflow();

      /// Make a character available to be returned by the next extraction.
      int_type
      pbackfail(int_type c = traits_type::eof());

      /// Write any buffered characters to the stream.
      int
      sync();

      /// Insert multiple characters into the pipe.
      std::streamsize
      xsputn(const char_type* s, std::streamsize n);

      /// Insert a sequence of characters into the pipe.
      std::streamsize
      write(const char_type* s, std::streamsize n);

      /// Extract a sequence of characters from the pipe.
      std::streamsize
      read(char_type* s, std::streamsize n);

      /// Report how many characters can be read from active input without blocking.
      std::streamsize
      showmanyc();

    protected:
      /// Enumerated type to indicate whether stdout or stderr is to be read.
      enum buf_read_src { rsrc_out = 0, rsrc_err = 1 };

      /// Initialise pipes and fork process.
      pid_t
      fork(pmode mode);

      /// Wait for the child process to exit.
      int
      wait(bool nohang = false);

      /// Return the file descriptor for the output pipe.
      fd_type&
      wpipe();

      /// Return the file descriptor for the active input pipe.
      fd_type&
      rpipe();

      /// Return the file descriptor for the specified input pipe.
      fd_type&
      rpipe(buf_read_src which);

      void
      create_buffers(pmode mode);

      void
      destroy_buffers(pmode mode);

      /// Writes buffered characters to the process' stdin pipe.
      bool
      empty_buffer();

      bool
      fill_buffer(bool non_blocking = false);

      /// Return the active input buffer.
      char_type*
      rbuffer();

      buf_read_src
      switch_read_buffer(buf_read_src);

    private:
      basic_pstreambuf(const basic_pstreambuf&);
      basic_pstreambuf& operator=(const basic_pstreambuf&);

      void
      init_rbuffers();

      pid_t         ppid_;        // pid of process
      fd_type       wpipe_;       // pipe used to write to process' stdin
      fd_type       rpipe_[2];    // two pipes to read from, stdout and stderr
      char_type*    wbuffer_;
      char_type*    rbuffer_[2];
      char_type*    rbufstate_[3];
      /// Index into rpipe_[] to indicate active source for read operations.
      buf_read_src  rsrc_;
      int           status_;      // hold exit status of child process
      int           error_;       // hold errno if fork() or exec() fails
    };

  /// Class template for common base class.
  template <typename CharT, typename Traits = std::char_traits<CharT> >
    class pstream_common
    : virtual public std::basic_ios<CharT, Traits>
    , virtual public pstreams
    {
    protected:
      typedef basic_pstreambuf<CharT, Traits>       streambuf_type;

      typedef pstreams::pmode                       pmode;
      typedef pstreams::argv_type                   argv_type;

      /// Default constructor.
      pstream_common();

      /// Constructor that initialises the stream by starting a process.
      pstream_common(const std::string& cmd, pmode mode);

      /// Constructor that initialises the stream by starting a process.
      pstream_common(const std::string& file, const argv_type& argv, pmode mode);

      /// Pure virtual destructor.
      virtual
      ~pstream_common() = 0;

      /// Start a process.
      void
      do_open(const std::string& cmd, pmode mode);

      /// Start a process.
      void
      do_open(const std::string& file, const argv_type& argv, pmode mode);

    public:
      /// Close the pipe.
      void
      close();

      /// Report whether the stream's buffer has been initialised.
      bool
      is_open() const;

      /// Return the command used to initialise the stream.
      const std::string&
      command() const;

      /// Return a pointer to the stream buffer.
      streambuf_type*
      rdbuf() const;

#if REDI_EVISCERATE_PSTREAMS
      /// Obtain FILE pointers for each of the process' standard streams.
      std::size_t
      fopen(FILE*& in, FILE*& out, FILE*& err);
#endif

    protected:
      std::string       command_; ///< The command used to start the process.
      streambuf_type    buf_;     ///< The stream buffer.
    };


  /**
   * @class basic_ipstream
   * @brief Class template for Input PStreams.
   *
   * Reading from an ipstream reads the command's standard output and/or
   * standard error (depending on how the ipstream is opened)
   * and the command's standard input is the same as that of the process
   * that created the object, unless altered by the command itself.
   */

  template <typename CharT, typename Traits = std::char_traits<CharT> >
    class basic_ipstream
    : public std::basic_istream<CharT, Traits>
    , public pstream_common<CharT, Traits>
    , virtual public pstreams
    {
      typedef std::basic_istream<CharT, Traits>     istream_type;
      typedef pstream_common<CharT, Traits>         pbase_type;

      using pbase_type::buf_;  // declare name in this scope

      // Ensure a basic_ipstream will read from at least one pipe
      pmode readable(pmode mode)
      {
        if (!(mode & (pstdout|pstderr)))
          mode |= pstdout;
        return mode;
      }

    public:
      /// Type used to specify how to connect to the process.
      typedef typename pbase_type::pmode            pmode;

      /// Type used to hold the arguments for a command.
      typedef typename pbase_type::argv_type        argv_type;

      /// Default constructor, creates an uninitialised stream.
      basic_ipstream()
      : istream_type(NULL), pbase_type()
      { }

      /**
       * @brief Constructor that initialises the stream by starting a process.
       *
       * Initialises the stream buffer by calling do_open() with the supplied
       * arguments.
       *
       * @param cmd   a string containing a shell command.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, pmode)
       */
      explicit
      basic_ipstream(const std::string& cmd, pmode mode = pstdout)
      : istream_type(NULL), pbase_type(cmd, readable(mode))
      { }

      /**
       * @brief Constructor that initialises the stream by starting a process.
       *
       * Initialises the stream buffer by calling do_open() with the supplied
       * arguments.
       *
       * @param file  a string containing the pathname of a program to execute.
       * @param argv  a vector of argument strings passed to the new program.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, const argv_type&, pmode)
       */
      basic_ipstream( const std::string& file,
                      const argv_type& argv,
                      pmode mode = pstdout )
      : istream_type(NULL), pbase_type(file, argv, readable(mode))
      { }

      /**
       * @brief Constructor that initialises the stream by starting a process.
       *
       * Initialises the stream buffer by calling
       * @c do_open(argv[0],argv,mode|pstdout)
       *
       * @param argv  a vector of argument strings passed to the new program.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, const argv_type&, pmode)
       */
      explicit
      basic_ipstream(const argv_type& argv, pmode mode = pstdout)
      : istream_type(NULL), pbase_type(argv.at(0), argv, readable(mode))
      { }

#if __cplusplus >= 201103L
      template<typename T>
        explicit
        basic_ipstream(std::initializer_list<T> args, pmode mode = pstdout)
        : basic_ipstream(argv_type(args.begin(), args.end()), mode)
        { }
#endif

      /**
       * @brief Destructor.
       *
       * Closes the stream and waits for the child to exit.
       */
      ~basic_ipstream()
      { }

      /**
       * @brief Start a process.
       *
       * Calls do_open( @a cmd , @a mode|pstdout ).
       *
       * @param cmd   a string containing a shell command.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, pmode)
       */
      void
      open(const std::string& cmd, pmode mode = pstdout)
      {
        this->do_open(cmd, readable(mode));
      }

      /**
       * @brief Start a process.
       *
       * Calls do_open( @a file , @a argv , @a mode|pstdout ).
       *
       * @param file  a string containing the pathname of a program to execute.
       * @param argv  a vector of argument strings passed to the new program.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, const argv_type&, pmode)
       */
      void
      open( const std::string& file,
            const argv_type& argv,
            pmode mode = pstdout )
      {
        this->do_open(file, argv, readable(mode));
      }

      /**
       * @brief Set streambuf to read from process' @c stdout.
       * @return  @c *this
       */
      basic_ipstream&
      out()
      {
        this->buf_.read_err(false);
        return *this;
      }

      /**
       * @brief Set streambuf to read from process' @c stderr.
       * @return  @c *this
       */
      basic_ipstream&
      err()
      {
        this->buf_.read_err(true);
        return *this;
      }
    };


  /**
   * @class basic_opstream
   * @brief Class template for Output PStreams.
   *
   * Writing to an open opstream writes to the standard input of the command;
   * the command's standard output is the same as that of the process that
   * created the pstream object, unless altered by the command itself.
   */

  template <typename CharT, typename Traits = std::char_traits<CharT> >
    class basic_opstream
    : public std::basic_ostream<CharT, Traits>
    , public pstream_common<CharT, Traits>
    , virtual public pstreams
    {
      typedef std::basic_ostream<CharT, Traits>     ostream_type;
      typedef pstream_common<CharT, Traits>         pbase_type;

      using pbase_type::buf_;  // declare name in this scope

    public:
      /// Type used to specify how to connect to the process.
      typedef typename pbase_type::pmode            pmode;

      /// Type used to hold the arguments for a command.
      typedef typename pbase_type::argv_type        argv_type;

      /// Default constructor, creates an uninitialised stream.
      basic_opstream()
      : ostream_type(NULL), pbase_type()
      { }

      /**
       * @brief Constructor that initialises the stream by starting a process.
       *
       * Initialises the stream buffer by calling do_open() with the supplied
       * arguments.
       *
       * @param cmd   a string containing a shell command.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, pmode)
       */
      explicit
      basic_opstream(const std::string& cmd, pmode mode = pstdin)
      : ostream_type(NULL), pbase_type(cmd, mode|pstdin)
      { }

      /**
       * @brief Constructor that initialises the stream by starting a process.
       *
       * Initialises the stream buffer by calling do_open() with the supplied
       * arguments.
       *
       * @param file  a string containing the pathname of a program to execute.
       * @param argv  a vector of argument strings passed to the new program.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, const argv_type&, pmode)
       */
      basic_opstream( const std::string& file,
                      const argv_type& argv,
                      pmode mode = pstdin )
      : ostream_type(NULL), pbase_type(file, argv, mode|pstdin)
      { }

      /**
       * @brief Constructor that initialises the stream by starting a process.
       *
       * Initialises the stream buffer by calling
       * @c do_open(argv[0],argv,mode|pstdin)
       *
       * @param argv  a vector of argument strings passed to the new program.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, const argv_type&, pmode)
       */
      explicit
      basic_opstream(const argv_type& argv, pmode mode = pstdin)
      : ostream_type(NULL), pbase_type(argv.at(0), argv, mode|pstdin)
      { }

#if __cplusplus >= 201103L
      /**
       * @brief Constructor that initialises the stream by starting a process.
       *
       * @param args  a list of argument strings passed to the new program.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, const argv_type&, pmode)
       */
      template<typename T>
        explicit
        basic_opstream(std::initializer_list<T> args, pmode mode = pstdin)
        : basic_opstream(argv_type(args.begin(), args.end()), mode)
        { }
#endif

      /**
       * @brief Destructor
       *
       * Closes the stream and waits for the child to exit.
       */
      ~basic_opstream() { }

      /**
       * @brief Start a process.
       *
       * Calls do_open( @a cmd , @a mode|pstdin ).
       *
       * @param cmd   a string containing a shell command.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, pmode)
       */
      void
      open(const std::string& cmd, pmode mode = pstdin)
      {
        this->do_open(cmd, mode|pstdin);
      }

      /**
       * @brief Start a process.
       *
       * Calls do_open( @a file , @a argv , @a mode|pstdin ).
       *
       * @param file  a string containing the pathname of a program to execute.
       * @param argv  a vector of argument strings passed to the new program.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, const argv_type&, pmode)
       */
      void
      open( const std::string& file,
            const argv_type& argv,
            pmode mode = pstdin)
      {
        this->do_open(file, argv, mode|pstdin);
      }
    };


  /**
   * @class basic_pstream
   * @brief Class template for Bidirectional PStreams.
   *
   * Writing to a pstream opened with @c pmode @c pstdin writes to the
   * standard input of the command.
   * Reading from a pstream opened with @c pmode @c pstdout and/or @c pstderr
   * reads the command's standard output and/or standard error.
   * Any of the process' @c stdin, @c stdout or @c stderr that is not
   * connected to the pstream (as specified by the @c pmode)
   * will be the same as the process that created the pstream object,
   * unless altered by the command itself.
   */
  template <typename CharT, typename Traits = std::char_traits<CharT> >
    class basic_pstream
    : public std::basic_iostream<CharT, Traits>
    , public pstream_common<CharT, Traits>
    , virtual public pstreams
    {
      typedef std::basic_iostream<CharT, Traits>    iostream_type;
      typedef pstream_common<CharT, Traits>         pbase_type;

      using pbase_type::buf_;  // declare name in this scope

    public:
      /// Type used to specify how to connect to the process.
      typedef typename pbase_type::pmode            pmode;

      /// Type used to hold the arguments for a command.
      typedef typename pbase_type::argv_type        argv_type;

      /// Default constructor, creates an uninitialised stream.
      basic_pstream()
      : iostream_type(NULL), pbase_type()
      { }

      /**
       * @brief Constructor that initialises the stream by starting a process.
       *
       * Initialises the stream buffer by calling do_open() with the supplied
       * arguments.
       *
       * @param cmd   a string containing a shell command.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, pmode)
       */
      explicit
      basic_pstream(const std::string& cmd, pmode mode = pstdout|pstdin)
      : iostream_type(NULL), pbase_type(cmd, mode)
      { }

      /**
       * @brief Constructor that initialises the stream by starting a process.
       *
       * Initialises the stream buffer by calling do_open() with the supplied
       * arguments.
       *
       * @param file  a string containing the pathname of a program to execute.
       * @param argv  a vector of argument strings passed to the new program.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, const argv_type&, pmode)
       */
      basic_pstream( const std::string& file,
                     const argv_type& argv,
                     pmode mode = pstdout|pstdin )
      : iostream_type(NULL), pbase_type(file, argv, mode)
      { }

      /**
       * @brief Constructor that initialises the stream by starting a process.
       *
       * Initialises the stream buffer by calling
       * @c do_open(argv[0],argv,mode)
       *
       * @param argv  a vector of argument strings passed to the new program.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, const argv_type&, pmode)
       */
      explicit
      basic_pstream(const argv_type& argv, pmode mode = pstdout|pstdin)
      : iostream_type(NULL), pbase_type(argv.at(0), argv, mode)
      { }

#if __cplusplus >= 201103L
      /**
       * @brief Constructor that initialises the stream by starting a process.
       *
       * @param l     a list of argument strings passed to the new program.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, const argv_type&, pmode)
       */
      template<typename T>
        explicit
        basic_pstream(std::initializer_list<T> l, pmode mode = pstdout|pstdin)
        : basic_pstream(argv_type(l.begin(), l.end()), mode)
        { }
#endif

      /**
       * @brief Destructor
       *
       * Closes the stream and waits for the child to exit.
       */
      ~basic_pstream() { }

      /**
       * @brief Start a process.
       *
       * Calls do_open( @a cnd , @a mode ).
       *
       * @param cmd   a string containing a shell command.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, pmode)
       */
      void
      open(const std::string& cmd, pmode mode = pstdout|pstdin)
      {
        this->do_open(cmd, mode);
      }

      /**
       * @brief Start a process.
       *
       * Calls do_open( @a file , @a argv , @a mode ).
       *
       * @param file  a string containing the pathname of a program to execute.
       * @param argv  a vector of argument strings passed to the new program.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, const argv_type&, pmode)
       */
      void
      open( const std::string& file,
            const argv_type& argv,
            pmode mode = pstdout|pstdin )
      {
        this->do_open(file, argv, mode);
      }

      /**
       * @brief Set streambuf to read from process' @c stdout.
       * @return  @c *this
       */
      basic_pstream&
      out()
      {
        this->buf_.read_err(false);
        return *this;
      }

      /**
       * @brief Set streambuf to read from process' @c stderr.
       * @return  @c *this
       */
      basic_pstream&
      err()
      {
        this->buf_.read_err(true);
        return *this;
      }
    };


  /**
   * @class basic_rpstream
   * @brief Class template for Restricted PStreams.
   *
   * Writing to an rpstream opened with @c pmode @c pstdin writes to the
   * standard input of the command.
   * It is not possible to read directly from an rpstream object, to use
   * an rpstream as in istream you must call either basic_rpstream::out()
   * or basic_rpstream::err(). This is to prevent accidental reads from
   * the wrong input source. If the rpstream was not opened with @c pmode
   * @c pstderr then the class cannot read the process' @c stderr, and
   * basic_rpstream::err() will return an istream that reads from the
   * process' @c stdout, and vice versa.
   * Reading from an rpstream opened with @c pmode @c pstdout and/or
   * @c pstderr reads the command's standard output and/or standard error.
   * Any of the process' @c stdin, @c stdout or @c stderr that is not
   * connected to the pstream (as specified by the @c pmode)
   * will be the same as the process that created the pstream object,
   * unless altered by the command itself.
   */

  template <typename CharT, typename Traits = std::char_traits<CharT> >
    class basic_rpstream
    : public std::basic_ostream<CharT, Traits>
    , private std::basic_istream<CharT, Traits>
    , private pstream_common<CharT, Traits>
    , virtual public pstreams
    {
      typedef std::basic_ostream<CharT, Traits>     ostream_type;
      typedef std::basic_istream<CharT, Traits>     istream_type;
      typedef pstream_common<CharT, Traits>         pbase_type;

      using pbase_type::buf_;  // declare name in this scope

    public:
      /// Type used to specify how to connect to the process.
      typedef typename pbase_type::pmode            pmode;

      /// Type used to hold the arguments for a command.
      typedef typename pbase_type::argv_type        argv_type;

      /// Default constructor, creates an uninitialised stream.
      basic_rpstream()
      : ostream_type(NULL), istream_type(NULL), pbase_type()
      { }

      /**
       * @brief  Constructor that initialises the stream by starting a process.
       *
       * Initialises the stream buffer by calling do_open() with the supplied
       * arguments.
       *
       * @param cmd   a string containing a shell command.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, pmode)
       */
      explicit
      basic_rpstream(const std::string& cmd, pmode mode = pstdout|pstdin)
      : ostream_type(NULL) , istream_type(NULL) , pbase_type(cmd, mode)
      { }

      /**
       * @brief  Constructor that initialises the stream by starting a process.
       *
       * Initialises the stream buffer by calling do_open() with the supplied
       * arguments.
       *
       * @param file a string containing the pathname of a program to execute.
       * @param argv a vector of argument strings passed to the new program.
       * @param mode the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, const argv_type&, pmode)
       */
      basic_rpstream( const std::string& file,
                      const argv_type& argv,
                      pmode mode = pstdout|pstdin )
      : ostream_type(NULL), istream_type(NULL), pbase_type(file, argv, mode)
      { }

      /**
       * @brief Constructor that initialises the stream by starting a process.
       *
       * Initialises the stream buffer by calling
       * @c do_open(argv[0],argv,mode)
       *
       * @param argv  a vector of argument strings passed to the new program.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, const argv_type&, pmode)
       */
      explicit
      basic_rpstream(const argv_type& argv, pmode mode = pstdout|pstdin)
      : ostream_type(NULL), istream_type(NULL),
        pbase_type(argv.at(0), argv, mode)
      { }

#if __cplusplus >= 201103L
      /**
       * @brief Constructor that initialises the stream by starting a process.
       *
       * @param l     a list of argument strings passed to the new program.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, const argv_type&, pmode)
       */
      template<typename T>
        explicit
        basic_rpstream(std::initializer_list<T> l, pmode mode = pstdout|pstdin)
        : basic_rpstream(argv_type(l.begin(), l.end()), mode)
        { }
#endif

      /// Destructor
      ~basic_rpstream() { }

      /**
       * @brief  Start a process.
       *
       * Calls do_open( @a cmd , @a mode ).
       *
       * @param cmd   a string containing a shell command.
       * @param mode  the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, pmode)
       */
      void
      open(const std::string& cmd, pmode mode = pstdout|pstdin)
      {
        this->do_open(cmd, mode);
      }

      /**
       * @brief  Start a process.
       *
       * Calls do_open( @a file , @a argv , @a mode ).
       *
       * @param file a string containing the pathname of a program to execute.
       * @param argv a vector of argument strings passed to the new program.
       * @param mode the I/O mode to use when opening the pipe.
       * @see   do_open(const std::string&, const argv_type&, pmode)
       */
      void
      open( const std::string& file,
            const argv_type& argv,
            pmode mode = pstdout|pstdin )
      {
        this->do_open(file, argv, mode);
      }

      /**
       * @brief  Obtain a reference to the istream that reads
       *         the process' @c stdout.
       * @return @c *this
       */
      istream_type&
      out()
      {
        this->buf_.read_err(false);
        return *this;
      }

      /**
       * @brief  Obtain a reference to the istream that reads
       *         the process' @c stderr.
       * @return @c *this
       */
      istream_type&
      err()
      {
        this->buf_.read_err(true);
        return *this;
      }
    };


  /// Type definition for common template specialisation.
  typedef basic_pstreambuf<char> pstreambuf;
  /// Type definition for common template specialisation.
  typedef basic_ipstream<char> ipstream;
  /// Type definition for common template specialisation.
  typedef basic_opstream<char> opstream;
  /// Type definition for common template specialisation.
  typedef basic_pstream<char> pstream;
  /// Type definition for common template specialisation.
  typedef basic_rpstream<char> rpstream;


  /**
   * When inserted into an output pstream the manipulator calls
   * basic_pstreambuf<C,T>::peof() to close the output pipe,
   * causing the child process to receive the end-of-file indicator
   * on subsequent reads from its @c stdin stream.
   *
   * @brief   Manipulator to close the pipe connected to the process' stdin.
   * @param   s  An output PStream class.
   * @return  The stream object the manipulator was invoked on.
   * @see     basic_pstreambuf<C,T>::peof()
   * @relates basic_opstream basic_pstream basic_rpstream
   */
  template <typename C, typename T>
    inline std::basic_ostream<C,T>&
    peof(std::basic_ostream<C,T>& s)
    {
      typedef basic_pstreambuf<C,T> pstreambuf_type;
      if (pstreambuf_type* p = dynamic_cast<pstreambuf_type*>(s.rdbuf()))
        p->peof();
      return s;
    }


  /*
   * member definitions for pstreambuf
   */


  /**
   * @class basic_pstreambuf
   * Provides underlying streambuf functionality for the PStreams classes.
   */

  /** Creates an uninitialised stream buffer. */
  template <typename C, typename T>
    inline
    basic_pstreambuf<C,T>::basic_pstreambuf()
    : ppid_(-1)   // initialise to -1 to indicate no process run yet.
    , wpipe_(-1)
    , wbuffer_(NULL)
    , rsrc_(rsrc_out)
    , status_(-1)
    , error_(0)
    {
      init_rbuffers();
    }

  /**
   * Initialises the stream buffer by calling open() with the supplied
   * arguments.
   *
   * @param cmd   a string containing a shell command.
   * @param mode  the I/O mode to use when opening the pipe.
   * @see   open()
   */
  template <typename C, typename T>
    inline
    basic_pstreambuf<C,T>::basic_pstreambuf(const std::string& cmd, pmode mode)
    : ppid_(-1)   // initialise to -1 to indicate no process run yet.
    , wpipe_(-1)
    , wbuffer_(NULL)
    , rsrc_(rsrc_out)
    , status_(-1)
    , error_(0)
    {
      init_rbuffers();
      open(cmd, mode);
    }

  /**
   * Initialises the stream buffer by calling open() with the supplied
   * arguments.
   *
   * @param file  a string containing the name of a program to execute.
   * @param argv  a vector of argument strings passsed to the new program.
   * @param mode  the I/O mode to use when opening the pipe.
   * @see   open()
   */
  template <typename C, typename T>
    inline
    basic_pstreambuf<C,T>::basic_pstreambuf( const std::string& file,
                                             const argv_type& argv,
                                             pmode mode )
    : ppid_(-1)   // initialise to -1 to indicate no process run yet.
    , wpipe_(-1)
    , wbuffer_(NULL)
    , rsrc_(rsrc_out)
    , status_(-1)
    , error_(0)
    {
      init_rbuffers();
      open(file, argv, mode);
    }

  /**
   * Closes the stream by calling close().
   * @see close()
   */
  template <typename C, typename T>
    inline
    basic_pstreambuf<C,T>::~basic_pstreambuf()
    {
      close();
    }

  /**
   * Starts a new process by passing @a command to the shell (/bin/sh)
   * and opens pipes to the process with the specified @a mode.
   *
   * If @a mode contains @c pstdout the initial read source will be
   * the child process' stdout, otherwise if @a mode  contains @c pstderr
   * the initial read source will be the child's stderr.
   *
   * Will duplicate the actions of  the  shell  in searching for an
   * executable file if the specified file name does not contain a slash (/)
   * character.
   *
   * @warning
   * There is no way to tell whether the shell command succeeded, this
   * function will always succeed unless resource limits (such as
   * memory usage, or number of processes or open files) are exceeded.
   * This means is_open() will return true even if @a command cannot
   * be executed.
   * Use pstreambuf::open(const std::string&, const argv_type&, pmode)
   * if you need to know whether the command failed to execute.
   *
   * @param   command  a string containing a shell command.
   * @param   mode     a bitwise OR of one or more of @c out, @c in, @c err.
   * @return  NULL if the shell could not be started or the
   *          pipes could not be opened, @c this otherwise.
   * @see     <b>execl</b>(3)
   */
  template <typename C, typename T>
    basic_pstreambuf<C,T>*
    basic_pstreambuf<C,T>::open(const std::string& command, pmode mode)
    {
      const char * shell_path = "/bin/sh";
#if 0
      const std::string argv[] = { "sh", "-c", command };
      return this->open(shell_path, argv_type(argv, argv+3), mode);
#else
      basic_pstreambuf<C,T>* ret = NULL;

      if (!is_open())
      {
        switch(fork(mode))
        {
        case 0 :
          // this is the new process, exec command
          ::execl(shell_path, "sh", "-c", command.c_str(), (char*)NULL);

          // can only reach this point if exec() failed

          // parent can get exit code from waitpid()
          ::_exit(errno);
          // using std::exit() would make static dtors run twice

        case -1 :
          // couldn't fork, error already handled in pstreambuf::fork()
          break;

        default :
          // this is the parent process
          // activate buffers
          create_buffers(mode);
          ret = this;
        }
      }
      return ret;
#endif
    }

  /**
   * @brief  Helper function to close a file descriptor.
   *
   * Inspects @a fd and calls <b>close</b>(3) if it has a non-negative value.
   *
   * @param   fd  a file descriptor.
   * @relates basic_pstreambuf
   */
  inline void
  close_fd(pstreams::fd_type& fd)
  {
    if (fd >= 0 && ::close(fd) == 0)
      fd = -1;
  }

  /**
   * @brief  Helper function to close an array of file descriptors.
   *
   * Calls @c close_fd() on each member of the array.
   * The length of the array is determined automatically by
   * template argument deduction to avoid errors.
   *
   * @param   fds  an array of file descriptors.
   * @relates basic_pstreambuf
   */
  template <int N>
    inline void
    close_fd_array(pstreams::fd_type (&fds)[N])
    {
      for (std::size_t i = 0; i < N; ++i)
        close_fd(fds[i]);
    }

  /**
   * Starts a new process by executing @a file with the arguments in
   * @a argv and opens pipes to the process with the specified @a mode.
   *
   * By convention @c argv[0] should be the file name of the file being
   * executed.
   *
   * If @a mode contains @c pstdout the initial read source will be
   * the child process' stdout, otherwise if @a mode  contains @c pstderr
   * the initial read source will be the child's stderr.
   *
   * Will duplicate the actions of  the  shell  in searching for an
   * executable file if the specified file name does not contain a slash (/)
   * character.
   *
   * Iff @a file is successfully executed then is_open() will return true.
   * Otherwise, pstreambuf::error() can be used to obtain the value of
   * @c errno that was set by <b>execvp</b>(3) in the child process.
   *
   * The exit status of the new process will be returned by
   * pstreambuf::status() after pstreambuf::exited() returns true.
   *
   * @param   file  a string containing the pathname of a program to execute.
   * @param   argv  a vector of argument strings passed to the new program.
   * @param   mode  a bitwise OR of one or more of @c out, @c in and @c err.
   * @return  NULL if a pipe could not be opened or if the program could
   *          not be executed, @c this otherwise.
   * @see     <b>execvp</b>(3)
   */
  template <typename C, typename T>
    basic_pstreambuf<C,T>*
    basic_pstreambuf<C,T>::open( const std::string& file,
                                 const argv_type& argv,
                                 pmode mode )
    {
      basic_pstreambuf<C,T>* ret = NULL;

      if (!is_open())
      {
        // constants for read/write ends of pipe
        enum { RD, WR };

        // open another pipe and set close-on-exec
        fd_type ck_exec[] = { -1, -1 };
        if (-1 == ::pipe(ck_exec)
            || -1 == ::fcntl(ck_exec[RD], F_SETFD, FD_CLOEXEC)
            || -1 == ::fcntl(ck_exec[WR], F_SETFD, FD_CLOEXEC))
        {
          error_ = errno;
          close_fd_array(ck_exec);
        }
        else
        {
          switch(fork(mode))
          {
          case 0 :
            // this is the new process, exec command
            {
              char** arg_v = new char*[argv.size()+1];
              for (std::size_t i = 0; i < argv.size(); ++i)
              {
                const std::string& src = argv[i];
                char*& dest = arg_v[i];
                dest = new char[src.size()+1];
                dest[ src.copy(dest, src.size()) ] = '\0';
              }
              arg_v[argv.size()] = NULL;

              ::execvp(file.c_str(), arg_v);

              // can only reach this point if exec() failed

              // parent can get error code from ck_exec pipe
              error_ = errno;

              while (::write(ck_exec[WR], &error_, sizeof(error_)) == -1
                  && errno == EINTR)
              { }

              ::close(ck_exec[WR]);
              ::close(ck_exec[RD]);

              ::_exit(error_);
              // using std::exit() would make static dtors run twice
            }

          case -1 :
            // couldn't fork, error already handled in pstreambuf::fork()
            close_fd_array(ck_exec);
            break;

          default :
            // this is the parent process

            // check child called exec() successfully
            ::close(ck_exec[WR]);
            switch (::read(ck_exec[RD], &error_, sizeof(error_)))
            {
            case 0:
              // activate buffers
              create_buffers(mode);
              ret = this;
              break;
            case -1:
              error_ = errno;
              break;
            default:
              // error_ contains error code from child
              // call wait() to clean up and set ppid_ to 0
              this->wait();
              break;
            }
            ::close(ck_exec[RD]);
          }
        }
      }
      return ret;
    }

  /**
   * Creates pipes as specified by @a mode and calls @c fork() to create
   * a new process. If the fork is successful the parent process stores
   * the child's PID and the opened pipes and the child process replaces
   * its standard streams with the opened pipes.
   *
   * If an error occurs the error code will be set to one of the possible
   * errors for @c pipe() or @c fork().
   * See your system's documentation for these error codes.
   *
   * @param   mode  an OR of pmodes specifying which of the child's
   *                standard streams to connect to.
   * @return  On success the PID of the child is returned in the parent's
   *          context and zero is returned in the child's context.
   *          On error -1 is returned and the error code is set appropriately.
   */
  template <typename C, typename T>
    pid_t
    basic_pstreambuf<C,T>::fork(pmode mode)
    {
      pid_t pid = -1;

      // Three pairs of file descriptors, for pipes connected to the
      // process' stdin, stdout and stderr
      // (stored in a single array so close_fd_array() can close all at once)
      fd_type fd[] = { -1, -1, -1, -1, -1, -1 };
      fd_type* const pin = fd;
      fd_type* const pout = fd+2;
      fd_type* const perr = fd+4;

      // constants for read/write ends of pipe
      enum { RD, WR };

      // N.B.
      // For the pstreambuf pin is an output stream and
      // pout and perr are input streams.

      if (!error_ && mode&pstdin && ::pipe(pin))
        error_ = errno;

      if (!error_ && mode&pstdout && ::pipe(pout))
        error_ = errno;

      if (!error_ && mode&pstderr && ::pipe(perr))
        error_ = errno;

      if (!error_)
      {
        pid = ::fork();
        switch (pid)
        {
          case 0 :
          {
            // this is the new process

            // for each open pipe close one end and redirect the
            // respective standard stream to the other end

            if (*pin >= 0)
            {
              ::close(pin[WR]);
              ::dup2(pin[RD], STDIN_FILENO);
              ::close(pin[RD]);
            }
            if (*pout >= 0)
            {
              ::close(pout[RD]);
              ::dup2(pout[WR], STDOUT_FILENO);
              ::close(pout[WR]);
            }
            if (*perr >= 0)
            {
              ::close(perr[RD]);
              ::dup2(perr[WR], STDERR_FILENO);
              ::close(perr[WR]);
            }

#ifdef _POSIX_JOB_CONTROL
            if (mode&newpg)
              ::setpgid(0, 0); // Change to a new process group
#endif

            break;
          }
          case -1 :
          {
            // couldn't fork for some reason
            error_ = errno;
            // close any open pipes
            close_fd_array(fd);
            break;
          }
          default :
          {
            // this is the parent process, store process' pid
            ppid_ = pid;

            // store one end of open pipes and close other end
            if (*pin >= 0)
            {
              wpipe_ = pin[WR];
              ::close(pin[RD]);
            }
            if (*pout >= 0)
            {
              rpipe_[rsrc_out] = pout[RD];
              ::close(pout[WR]);
            }
            if (*perr >= 0)
            {
              rpipe_[rsrc_err] = perr[RD];
              ::close(perr[WR]);
            }
          }
        }
      }
      else
      {
        // close any pipes we opened before failure
        close_fd_array(fd);
      }
      return pid;
    }

  /**
   * Closes all pipes and calls wait() to wait for the process to finish.
   * If an error occurs the error code will be set to one of the possible
   * errors for @c waitpid().
   * See your system's documentation for these errors.
   *
   * @return  @c this on successful close or @c NULL if there is no
   *          process to close or if an error occurs.
   */
  template <typename C, typename T>
    basic_pstreambuf<C,T>*
    basic_pstreambuf<C,T>::close()
    {
      const bool running = is_open();

      sync(); // this might call wait() and reap the child process

      // rather than trying to work out whether or not we need to clean up
      // just do it anyway, all cleanup functions are safe to call twice.

      destroy_buffers(pstdin|pstdout|pstderr);

      // close pipes before wait() so child gets EOF/SIGPIPE
      close_fd(wpipe_);
      close_fd_array(rpipe_);

      do
      {
        error_ = 0;
      } while (wait() == -1 && error() == EINTR);

      return running ? this : NULL;
    }

  /**
   *  Called on construction to initialise the arrays used for reading.
   */
  template <typename C, typename T>
    inline void
    basic_pstreambuf<C,T>::init_rbuffers()
    {
      rpipe_[rsrc_out] = rpipe_[rsrc_err] = -1;
      rbuffer_[rsrc_out] = rbuffer_[rsrc_err] = NULL;
      rbufstate_[0] = rbufstate_[1] = rbufstate_[2] = NULL;
    }

  template <typename C, typename T>
    void
    basic_pstreambuf<C,T>::create_buffers(pmode mode)
    {
      if (mode & pstdin)
      {
        delete[] wbuffer_;
        wbuffer_ = new char_type[bufsz];
        this->setp(wbuffer_, wbuffer_ + bufsz);
      }
      if (mode & pstdout)
      {
        delete[] rbuffer_[rsrc_out];
        rbuffer_[rsrc_out] = new char_type[bufsz];
        rsrc_ = rsrc_out;
        this->setg(rbuffer_[rsrc_out] + pbsz, rbuffer_[rsrc_out] + pbsz,
            rbuffer_[rsrc_out] + pbsz);
      }
      if (mode & pstderr)
      {
        delete[] rbuffer_[rsrc_err];
        rbuffer_[rsrc_err] = new char_type[bufsz];
        if (!(mode & pstdout))
        {
          rsrc_ = rsrc_err;
          this->setg(rbuffer_[rsrc_err] + pbsz, rbuffer_[rsrc_err] + pbsz,
              rbuffer_[rsrc_err] + pbsz);
        }
      }
    }

  template <typename C, typename T>
    void
    basic_pstreambuf<C,T>::destroy_buffers(pmode mode)
    {
      if (mode & pstdin)
      {
        this->setp(NULL, NULL);
        delete[] wbuffer_;
        wbuffer_ = NULL;
      }
      if (mode & pstdout)
      {
        if (rsrc_ == rsrc_out)
          this->setg(NULL, NULL, NULL);
        delete[] rbuffer_[rsrc_out];
        rbuffer_[rsrc_out] = NULL;
      }
      if (mode & pstderr)
      {
        if (rsrc_ == rsrc_err)
          this->setg(NULL, NULL, NULL);
        delete[] rbuffer_[rsrc_err];
        rbuffer_[rsrc_err] = NULL;
      }
    }

  template <typename C, typename T>
    typename basic_pstreambuf<C,T>::buf_read_src
    basic_pstreambuf<C,T>::switch_read_buffer(buf_read_src src)
    {
      if (rsrc_ != src)
      {
        char_type* tmpbufstate[] = {this->eback(), this->gptr(), this->egptr()};
        this->setg(rbufstate_[0], rbufstate_[1], rbufstate_[2]);
        for (std::size_t i = 0; i < 3; ++i)
          rbufstate_[i] = tmpbufstate[i];
        rsrc_ = src;
      }
      return rsrc_;
    }

  /**
   * Suspends execution and waits for the associated process to exit, or
   * until a signal is delivered whose action is to terminate the current
   * process or to call a signal handling function. If the process has
   * already exited (i.e. it is a "zombie" process) then wait() returns
   * immediately.  Waiting for the child process causes all its system
   * resources to be freed.
   *
   * error() will return EINTR if wait() is interrupted by a signal.
   *
   * @param   nohang  true to return immediately if the process has not exited.
   * @return  1 if the process has exited and wait() has not yet been called.
   *          0 if @a nohang is true and the process has not exited yet.
   *          -1 if no process has been started or if an error occurs,
   *          in which case the error can be found using error().
   */
  template <typename C, typename T>
    int
    basic_pstreambuf<C,T>::wait(bool nohang)
    {
      int child_exited = -1;
      if (is_open())
      {
        int exit_status;
        switch(::waitpid(ppid_, &exit_status, nohang ? WNOHANG : 0))
        {
          case 0 :
            // nohang was true and process has not exited
            child_exited = 0;
            break;
          case -1 :
            error_ = errno;
            break;
          default :
            // process has exited
            ppid_ = 0;
            status_ = exit_status;
            child_exited = 1;
            // Close wpipe, would get SIGPIPE if we used it.
            destroy_buffers(pstdin);
            close_fd(wpipe_);
            // Must free read buffers and pipes on destruction
            // or next call to open()/close()
            break;
        }
      }
      return child_exited;
    }

  /**
   * Sends the specified signal to the process.  A signal can be used to
   * terminate a child process that would not exit otherwise.
   *
   * If an error occurs the error code will be set to one of the possible
   * errors for @c kill().  See your system's documentation for these errors.
   *
   * @param   signal  A signal to send to the child process.
   * @return  @c this or @c NULL if @c kill() fails.
   */
  template <typename C, typename T>
    inline basic_pstreambuf<C,T>*
    basic_pstreambuf<C,T>::kill(int signal)
    {
      basic_pstreambuf<C,T>* ret = NULL;
      if (is_open())
      {
        if (::kill(ppid_, signal))
          error_ = errno;
        else
        {
#if 0
          // TODO call exited() to check for exit and clean up? leave to user?
          if (signal==SIGTERM || signal==SIGKILL)
            this->exited();
#endif
          ret = this;
        }
      }
      return ret;
    }

  /**
   * Sends the specified signal to the process group of the child process.
   * A signal can be used to terminate a child process that would not exit
   * otherwise, or to kill the process and its own children.
   *
   * If an error occurs the error code will be set to one of the possible
   * errors for @c getpgid() or @c kill().  See your system's documentation
   * for these errors. If the child is in the current process group then
   * NULL will be returned and the error code set to EPERM.
   *
   * @param   signal  A signal to send to the child process.
   * @return  @c this on success or @c NULL on failure.
   */
  template <typename C, typename T>
    inline basic_pstreambuf<C,T>*
    basic_pstreambuf<C,T>::killpg(int signal)
    {
      basic_pstreambuf<C,T>* ret = NULL;
#ifdef _POSIX_JOB_CONTROL
      if (is_open())
      {
        pid_t pgid = ::getpgid(ppid_);
        if (pgid == -1)
          error_ = errno;
        else if (pgid == ::getpgrp())
          error_ = EPERM;  // Don't commit suicide
        else if (::killpg(pgid, signal))
          error_ = errno;
        else
          ret = this;
      }
#else
      error_ = ENOTSUP;
#endif
      return ret;
    }

  /**
   *  This function can call pstreambuf::wait() and so may change the
   *  object's state if the child process has already exited.
   *
   *  @return  True if the associated process has exited, false otherwise.
   *  @see     basic_pstreambuf<C,T>::wait()
   */
  template <typename C, typename T>
    inline bool
    basic_pstreambuf<C,T>::exited()
    {
      return ppid_ == 0 || wait(true)==1;
    }


  /**
   *  @return  The exit status of the child process, or -1 if wait()
   *           has not yet been called to wait for the child to exit.
   *  @see     basic_pstreambuf<C,T>::wait()
   */
  template <typename C, typename T>
    inline int
    basic_pstreambuf<C,T>::status() const
    {
      return status_;
    }

  /**
   *  @return  The error code of the most recently failed operation, or zero.
   */
  template <typename C, typename T>
    inline int
    basic_pstreambuf<C,T>::error() const
    {
      return error_;
    }

  /**
   *  Closes the output pipe, causing the child process to receive the
   *  end-of-file indicator on subsequent reads from its @c stdin stream.
   */
  template <typename C, typename T>
    inline void
    basic_pstreambuf<C,T>::peof()
    {
      sync();
      destroy_buffers(pstdin);
      close_fd(wpipe_);
    }

  /**
   * Unlike pstreambuf::exited(), this function will not call wait() and
   * so will not change the object's state.  This means that once a child
   * process is executed successfully this function will continue to
   * return true even after the process exits (until wait() is called.)
   *
   * @return  true if a previous call to open() succeeded and wait() has
   *          not been called and determined that the process has exited,
   *          false otherwise.
   */
  template <typename C, typename T>
    inline bool
    basic_pstreambuf<C,T>::is_open() const
    {
      return ppid_ > 0;
    }

  /**
   * Toggle the stream used for reading. If @a readerr is @c true then the
   * process' @c stderr output will be used for subsequent extractions, if
   * @a readerr is false the the process' stdout will be used.
   * @param   readerr  @c true to read @c stderr, @c false to read @c stdout.
   * @return  @c true if the requested stream is open and will be used for
   *          subsequent extractions, @c false otherwise.
   */
  template <typename C, typename T>
    inline bool
    basic_pstreambuf<C,T>::read_err(bool readerr)
    {
      buf_read_src src = readerr ? rsrc_err : rsrc_out;
      if (rpipe_[src]>=0)
      {
        switch_read_buffer(src);
        return true;
      }
      return false;
    }

  /**
   * Called when the internal character buffer is not present or is full,
   * to transfer the buffer contents to the pipe.
   *
   * @param   c  a character to be written to the pipe.
   * @return  @c traits_type::eof() if an error occurs, otherwise if @a c
   *          is not equal to @c traits_type::eof() it will be buffered and
   *          a value other than @c traits_type::eof() returned to indicate
   *          success.
   */
  template <typename C, typename T>
    typename basic_pstreambuf<C,T>::int_type
    basic_pstreambuf<C,T>::overflow(int_type c)
    {
      if (!empty_buffer())
        return traits_type::eof();
      else if (!traits_type::eq_int_type(c, traits_type::eof()))
        return this->sputc(c);
      else
        return traits_type::not_eof(c);
    }


  template <typename C, typename T>
    int
    basic_pstreambuf<C,T>::sync()
    {
      return !exited() && empty_buffer() ? 0 : -1;
    }

  /**
   * @param   s  character buffer.
   * @param   n  buffer length.
   * @return  the number of characters written.
   */
  template <typename C, typename T>
    std::streamsize
    basic_pstreambuf<C,T>::xsputn(const char_type* s, std::streamsize n)
    {
      std::streamsize done = 0;
      while (done < n)
      {
        if (std::streamsize nbuf = this->epptr() - this->pptr())
        {
          nbuf = std::min(nbuf, n - done);
          traits_type::copy(this->pptr(), s + done, nbuf);
          this->pbump(nbuf);
          done += nbuf;
        }
        else if (!empty_buffer())
          break;
      }
      return done;
    }

  /**
   * @return  true if the buffer was emptied, false otherwise.
   */
  template <typename C, typename T>
    bool
    basic_pstreambuf<C,T>::empty_buffer()
    {
      const std::streamsize count = this->pptr() - this->pbase();
      if (count > 0)
      {
        const std::streamsize written = this->write(this->wbuffer_, count);
        if (written > 0)
        {
          if (const std::streamsize unwritten = count - written)
            traits_type::move(this->pbase(), this->pbase()+written, unwritten);
          this->pbump(-written);
          return true;
        }
      }
      return false;
    }

  /**
   * Called when the internal character buffer is is empty, to re-fill it
   * from the pipe.
   *
   * @return The first available character in the buffer,
   * or @c traits_type::eof() in case of failure.
   */
  template <typename C, typename T>
    typename basic_pstreambuf<C,T>::int_type
    basic_pstreambuf<C,T>::underflow()
    {
      if (this->gptr() < this->egptr() || fill_buffer())
        return traits_type::to_int_type(*this->gptr());
      else
        return traits_type::eof();
    }

  /**
   * Attempts to make @a c available as the next character to be read by
   * @c sgetc().
   *
   * @param   c   a character to make available for extraction.
   * @return  @a c if the character can be made available,
   *          @c traits_type::eof() otherwise.
   */
  template <typename C, typename T>
    typename basic_pstreambuf<C,T>::int_type
    basic_pstreambuf<C,T>::pbackfail(int_type c)
    {
      if (this->gptr() != this->eback())
      {
        this->gbump(-1);
        if (!traits_type::eq_int_type(c, traits_type::eof()))
          *this->gptr() = traits_type::to_char_type(c);
        return traits_type::not_eof(c);
      }
      else
         return traits_type::eof();
    }

  template <typename C, typename T>
    std::streamsize
    basic_pstreambuf<C,T>::showmanyc()
    {
      int avail = 0;
      if (sizeof(char_type) == 1)
        avail = fill_buffer(true) ? this->egptr() - this->gptr() : -1;
#ifdef FIONREAD
      else
      {
        if (::ioctl(rpipe(), FIONREAD, &avail) == -1)
          avail = -1;
        else if (avail)
          avail /= sizeof(char_type);
      }
#endif
      return std::streamsize(avail);
    }

  /**
   * @return  true if the buffer was filled, false otherwise.
   */
  template <typename C, typename T>
    bool
    basic_pstreambuf<C,T>::fill_buffer(bool non_blocking)
    {
      const std::streamsize pb1 = this->gptr() - this->eback();
      const std::streamsize pb2 = pbsz;
      const std::streamsize npb = std::min(pb1, pb2);

      char_type* const rbuf = rbuffer();

      if (npb)
        traits_type::move(rbuf + pbsz - npb, this->gptr() - npb, npb);

      std::streamsize rc = -1;

      if (non_blocking)
      {
        const int flags = ::fcntl(rpipe(), F_GETFL);
        if (flags != -1)
        {
          const bool blocking = !(flags & O_NONBLOCK);
          if (blocking)
            ::fcntl(rpipe(), F_SETFL, flags | O_NONBLOCK);  // set non-blocking

          error_ = 0;
          rc = read(rbuf + pbsz, bufsz - pbsz);

          if (rc == -1 && error_ == EAGAIN)  // nothing available
            rc = 0;
          else if (rc == 0)  // EOF
            rc = -1;

          if (blocking)
            ::fcntl(rpipe(), F_SETFL, flags); // restore
        }
      }
      else
        rc = read(rbuf + pbsz, bufsz - pbsz);

      if (rc > 0 || (rc == 0 && non_blocking))
      {
        this->setg( rbuf + pbsz - npb,
                    rbuf + pbsz,
                    rbuf + pbsz + rc );
        return true;
      }
      else
      {
        this->setg(NULL, NULL, NULL);
        return false;
      }
    }

  /**
   * Writes up to @a n characters to the pipe from the buffer @a s.
   *
   * @param   s  character buffer.
   * @param   n  buffer length.
   * @return  the number of characters written.
   */
  template <typename C, typename T>
    inline std::streamsize
    basic_pstreambuf<C,T>::write(const char_type* s, std::streamsize n)
    {
      std::streamsize nwritten = 0;
      if (wpipe() >= 0)
      {
        nwritten = ::write(wpipe(), s, n * sizeof(char_type));
        if (nwritten == -1)
          error_ = errno;
        else
          nwritten /= sizeof(char_type);
      }
      return nwritten;
    }

  /**
   * Reads up to @a n characters from the pipe to the buffer @a s.
   *
   * @param   s  character buffer.
   * @param   n  buffer length.
   * @return  the number of characters read.
   */
  template <typename C, typename T>
    inline std::streamsize
    basic_pstreambuf<C,T>::read(char_type* s, std::streamsize n)
    {
      std::streamsize nread = 0;
      if (rpipe() >= 0)
      {
        nread = ::read(rpipe(), s, n * sizeof(char_type));
        if (nread == -1)
          error_ = errno;
        else
          nread /= sizeof(char_type);
      }
      return nread;
    }

  /** @return a reference to the output file descriptor */
  template <typename C, typename T>
    inline pstreams::fd_type&
    basic_pstreambuf<C,T>::wpipe()
    {
      return wpipe_;
    }

  /** @return a reference to the active input file descriptor */
  template <typename C, typename T>
    inline pstreams::fd_type&
    basic_pstreambuf<C,T>::rpipe()
    {
      return rpipe_[rsrc_];
    }

  /** @return a reference to the specified input file descriptor */
  template <typename C, typename T>
    inline pstreams::fd_type&
    basic_pstreambuf<C,T>::rpipe(buf_read_src which)
    {
      return rpipe_[which];
    }

  /** @return a pointer to the start of the active input buffer area. */
  template <typename C, typename T>
    inline typename basic_pstreambuf<C,T>::char_type*
    basic_pstreambuf<C,T>::rbuffer()
    {
      return rbuffer_[rsrc_];
    }


  /*
   * member definitions for pstream_common
   */

  /**
   * @class pstream_common
   * Abstract Base Class providing common functionality for basic_ipstream,
   * basic_opstream and basic_pstream.
   * pstream_common manages the basic_pstreambuf stream buffer that is used
   * by the derived classes to initialise an iostream class.
   */

  /** Creates an uninitialised stream. */
  template <typename C, typename T>
    inline
    pstream_common<C,T>::pstream_common()
    : std::basic_ios<C,T>(NULL)
    , command_()
    , buf_()
    {
      this->std::basic_ios<C,T>::rdbuf(&buf_);
    }

  /**
   * Initialises the stream buffer by calling
   * do_open( @a command , @a mode )
   *
   * @param cmd   a string containing a shell command.
   * @param mode  the I/O mode to use when opening the pipe.
   * @see   do_open(const std::string&, pmode)
   */
  template <typename C, typename T>
    inline
    pstream_common<C,T>::pstream_common(const std::string& cmd, pmode mode)
    : std::basic_ios<C,T>(NULL)
    , command_(cmd)
    , buf_()
    {
      this->std::basic_ios<C,T>::rdbuf(&buf_);
      do_open(cmd, mode);
    }

  /**
   * Initialises the stream buffer by calling
   * do_open( @a file , @a argv , @a mode )
   *
   * @param file  a string containing the pathname of a program to execute.
   * @param argv  a vector of argument strings passed to the new program.
   * @param mode  the I/O mode to use when opening the pipe.
   * @see do_open(const std::string&, const argv_type&, pmode)
   */
  template <typename C, typename T>
    inline
    pstream_common<C,T>::pstream_common( const std::string& file,
                                         const argv_type& argv,
                                         pmode mode )
    : std::basic_ios<C,T>(NULL)
    , command_(file)
    , buf_()
    {
      this->std::basic_ios<C,T>::rdbuf(&buf_);
      do_open(file, argv, mode);
    }

  /**
   * This is a pure virtual function to make @c pstream_common abstract.
   * Because it is the destructor it will be called by derived classes
   * and so must be defined.  It is also protected, to discourage use of
   * the PStreams classes through pointers or references to the base class.
   *
   * @sa If defining a pure virtual seems odd you should read
   * http://www.gotw.ca/gotw/031.htm (and the rest of the site as well!)
   */
  template <typename C, typename T>
    inline
    pstream_common<C,T>::~pstream_common()
    {
    }

  /**
   * Calls rdbuf()->open( @a command , @a mode )
   * and sets @c failbit on error.
   *
   * @param cmd   a string containing a shell command.
   * @param mode  the I/O mode to use when opening the pipe.
   * @see   basic_pstreambuf::open(const std::string&, pmode)
   */
  template <typename C, typename T>
    inline void
    pstream_common<C,T>::do_open(const std::string& cmd, pmode mode)
    {
      if (!buf_.open((command_=cmd), mode))
        this->setstate(std::ios_base::failbit);
    }

  /**
   * Calls rdbuf()->open( @a file, @a  argv, @a mode )
   * and sets @c failbit on error.
   *
   * @param file  a string containing the pathname of a program to execute.
   * @param argv  a vector of argument strings passed to the new program.
   * @param mode  the I/O mode to use when opening the pipe.
   * @see   basic_pstreambuf::open(const std::string&, const argv_type&, pmode)
   */
  template <typename C, typename T>
    inline void
    pstream_common<C,T>::do_open( const std::string& file,
                                  const argv_type& argv,
                                  pmode mode )
    {
      if (!buf_.open((command_=file), argv, mode))
        this->setstate(std::ios_base::failbit);
    }

  /** Calls rdbuf->close() and sets @c failbit on error. */
  template <typename C, typename T>
    inline void
    pstream_common<C,T>::close()
    {
      if (!buf_.close())
        this->setstate(std::ios_base::failbit);
    }

  /**
   * @return  rdbuf()->is_open().
   * @see     basic_pstreambuf::is_open()
   */
  template <typename C, typename T>
    inline bool
    pstream_common<C,T>::is_open() const
    {
      return buf_.is_open();
    }

  /** @return a string containing the command used to initialise the stream. */
  template <typename C, typename T>
    inline const std::string&
    pstream_common<C,T>::command() const
    {
      return command_;
    }

  /** @return a pointer to the private stream buffer member. */
  // TODO  document behaviour if buffer replaced.
  template <typename C, typename T>
    inline typename pstream_common<C,T>::streambuf_type*
    pstream_common<C,T>::rdbuf() const
    {
      return const_cast<streambuf_type*>(&buf_);
    }


#if REDI_EVISCERATE_PSTREAMS
  /**
   * @def REDI_EVISCERATE_PSTREAMS
   * If this macro has a non-zero value then certain internals of the
   * @c basic_pstreambuf template class are exposed. In general this is
   * a Bad Thing, as the internal implementation is largely undocumented
   * and may be subject to change at any time, so this feature is only
   * provided because it might make PStreams useful in situations where
   * it is necessary to do Bad Things.
   */

  /**
   * @warning  This function exposes the internals of the stream buffer and
   *           should be used with caution. It is the caller's responsibility
   *           to flush streams etc. in order to clear any buffered data.
   *           The POSIX.1 function <b>fdopen</b>(3) is used to obtain the
   *           @c FILE pointers from the streambuf's private file descriptor
   *           members so consult your system's documentation for
   *           <b>fdopen</b>(3).
   *
   * @param   in    A FILE* that will refer to the process' stdin.
   * @param   out   A FILE* that will refer to the process' stdout.
   * @param   err   A FILE* that will refer to the process' stderr.
   * @return  An OR of zero or more of @c pstdin, @c pstdout, @c pstderr.
   *
   * For each open stream shared with the child process a @c FILE* is
   * obtained and assigned to the corresponding parameter. For closed
   * streams @c NULL is assigned to the parameter.
   * The return value can be tested to see which parameters should be
   * @c !NULL by masking with the corresponding @c pmode value.
   *
   * @see <b>fdopen</b>(3)
   */
  template <typename C, typename T>
    std::size_t
    basic_pstreambuf<C,T>::fopen(FILE*& in, FILE*& out, FILE*& err)
    {
      in = out = err = NULL;
      std::size_t open_files = 0;
      if (wpipe() > -1)
      {
        if ((in = ::fdopen(wpipe(), "w")))
        {
            open_files |= pstdin;
        }
      }
      if (rpipe(rsrc_out) > -1)
      {
        if ((out = ::fdopen(rpipe(rsrc_out), "r")))
        {
            open_files |= pstdout;
        }
      }
      if (rpipe(rsrc_err) > -1)
      {
        if ((err = ::fdopen(rpipe(rsrc_err), "r")))
        {
            open_files |= pstderr;
        }
      }
      return open_files;
    }

  /**
   *  @warning This function exposes the internals of the stream buffer and
   *  should be used with caution.
   *
   *  @param  in   A FILE* that will refer to the process' stdin.
   *  @param  out  A FILE* that will refer to the process' stdout.
   *  @param  err  A FILE* that will refer to the process' stderr.
   *  @return A bitwise-or of zero or more of @c pstdin, @c pstdout, @c pstderr.
   *  @see    basic_pstreambuf::fopen()
   */
  template <typename C, typename T>
    inline std::size_t
    pstream_common<C,T>::fopen(FILE*& fin, FILE*& fout, FILE*& ferr)
    {
      return buf_.fopen(fin, fout, ferr);
    }

#endif // REDI_EVISCERATE_PSTREAMS


} // namespace redi

/**
 * @mainpage PStreams Reference
 * @htmlinclude mainpage.html
 */

#endif  // REDI_PSTREAM_H_SEEN

/* This is the end of code block taken from pstream.h. 
 * The following codes are for MM-align */

/* File parsing and basic geometry operations */

void PrintErrorAndQuit(const string sErrorString)
{
    cout << sErrorString << endl;
    exit(1);
}

template <typename T> inline T getmin(const T &a, const T &b)
{
    return b<a?b:a;
}

template <class A> void NewArray(A *** array, int Narray1, int Narray2)
{
    *array=new A* [Narray1];
    for(int i=0; i<Narray1; i++) *(*array+i)=new A [Narray2];
}

template <class A> void DeleteArray(A *** array, int Narray)
{
    for(int i=0; i<Narray; i++)
        if(*(*array+i)) delete [] *(*array+i);
    if(Narray) delete [] (*array);
    (*array)=NULL;
}

string AAmap(char A)
{
    if (A=='A') return "ALA";
    if (A=='B') return "ASX";
    if (A=='C') return "CYS";
    if (A=='D') return "ASP";
    if (A=='E') return "GLU";
    if (A=='F') return "PHE";
    if (A=='G') return "GLY";
    if (A=='H') return "HIS";
    if (A=='I') return "ILE";
    if (A=='K') return "LYS";
    if (A=='L') return "LEU";
    if (A=='M') return "MET";
    if (A=='N') return "ASN";
    if (A=='O') return "PYL";
    if (A=='P') return "PRO";
    if (A=='Q') return "GLN";
    if (A=='R') return "ARG";
    if (A=='S') return "SER";
    if (A=='T') return "THR";
    if (A=='U') return "SEC";
    if (A=='V') return "VAL";
    if (A=='W') return "TRP";    
    if (A=='Y') return "TYR";
    if (A=='Z') return "GLX";
    if ('a'<=A && A<='z') return "  "+string(1,char(toupper(A)));
    return "UNK";
}

char AAmap(const string &AA)
{
    if (AA.compare("ALA")==0 || AA.compare("DAL")==0) return 'A';
    if (AA.compare("ASX")==0) return 'B';
    if (AA.compare("CYS")==0 || AA.compare("DCY")==0) return 'C';
    if (AA.compare("ASP")==0 || AA.compare("DAS")==0) return 'D';
    if (AA.compare("GLU")==0 || AA.compare("DGL")==0) return 'E';
    if (AA.compare("PHE")==0 || AA.compare("DPN")==0) return 'F';
    if (AA.compare("GLY")==0) return 'G';
    if (AA.compare("HIS")==0 || AA.compare("DHI")==0) return 'H';
    if (AA.compare("ILE")==0 || AA.compare("DIL")==0) return 'I';
    if (AA.compare("LYS")==0 || AA.compare("DLY")==0) return 'K';
    if (AA.compare("LEU")==0 || AA.compare("DLE")==0) return 'L';
    if (AA.compare("MET")==0 || AA.compare("MED")==0 ||
        AA.compare("MSE")==0) return 'M';
    if (AA.compare("ASN")==0 || AA.compare("DSG")==0) return 'N';
    if (AA.compare("PYL")==0) return 'O';
    if (AA.compare("PRO")==0 || AA.compare("DPR")==0) return 'P';
    if (AA.compare("GLN")==0 || AA.compare("DGN")==0) return 'Q';
    if (AA.compare("ARG")==0 || AA.compare("DAR")==0) return 'R';
    if (AA.compare("SER")==0 || AA.compare("DSN")==0) return 'S';
    if (AA.compare("THR")==0 || AA.compare("DTH")==0) return 'T';
    if (AA.compare("SEC")==0) return 'U';
    if (AA.compare("VAL")==0 || AA.compare("DVA")==0) return 'V';
    if (AA.compare("TRP")==0 || AA.compare("DTR")==0) return 'W';    
    if (AA.compare("TYR")==0 || AA.compare("DTY")==0) return 'Y';
    if (AA.compare("GLX")==0) return 'Z';

    if (AA.compare(0,2," D")==0) return tolower(AA[2]);
    if (AA.compare(0,2,"  ")==0) return tolower(AA[2]);
    return 'X';
}

/* split a long string into vectors by whitespace 
 * line          - input string
 * line_vec      - output vector 
 * delimiter     - delimiter */
void split(const string &line, vector<string> &line_vec,
    const char delimiter=' ')
{
    bool within_word = false;
    for (size_t pos=0;pos<line.size();pos++)
    {
        if (line[pos]==delimiter)
        {
            within_word = false;
            continue;
        }
        if (!within_word)
        {
            within_word = true;
            line_vec.push_back("");
        }
        line_vec.back()+=line[pos];
    }
}

size_t get_PDB_lines(const string filename,
    vector<vector<string> >&PDB_lines, vector<string> &chainID_list,
    vector<int> &mol_vec, const int ter_opt, const int infmt_opt,
    const string atom_opt, const int split_opt, const int het_opt)
{
    size_t i=0; // resi i.e. atom index
    string line;
    char chainID=0;
    string resi="";
    bool select_atom=false;
    size_t model_idx=0;
    vector<string> tmp_str_vec;
    
    int compress_type=0; // uncompressed file
    ifstream fin;
    redi::ipstream fin_gz; // if file is compressed
    if (filename.size()>=3 && 
        filename.substr(filename.size()-3,3)==".gz")
    {
        fin_gz.open("zcat '"+filename+"'");
        compress_type=1;
    }
    else if (filename.size()>=4 && 
        filename.substr(filename.size()-4,4)==".bz2")
    {
        fin_gz.open("bzcat '"+filename+"'");
        compress_type=2;
    }
    else fin.open(filename.c_str());

    if (infmt_opt==0||infmt_opt==-1) // PDB format
    {
        while (compress_type?fin_gz.good():fin.good())
        {
            if (compress_type) getline(fin_gz, line);
            else               getline(fin, line);
            if (infmt_opt==-1 && line.compare(0,5,"loop_")==0) // PDBx/mmCIF
                return get_PDB_lines(filename,PDB_lines,chainID_list,
                    mol_vec, ter_opt, 3, atom_opt, split_opt,het_opt);
            if (i > 0)
            {
                if      (ter_opt>=1 && line.compare(0,3,"END")==0) break;
                else if (ter_opt>=3 && line.compare(0,3,"TER")==0) break;
            }
            if (split_opt && line.compare(0,3,"END")==0) chainID=0;
            if (line.size()>=54 && (line[16]==' ' || line[16]=='A') && (
                (line.compare(0, 6, "ATOM  ")==0) || 
                (line.compare(0, 6, "HETATM")==0 && het_opt==1) ||
                (line.compare(0, 6, "HETATM")==0 && het_opt==2 && 
                 line.compare(17,3, "MSE")==0)))
            {
                if (atom_opt=="auto")
                {
                    if (line[17]==' ' && (line[18]=='D'||line[18]==' '))
                         select_atom=(line.compare(12,4," C3'")==0);
                    else select_atom=(line.compare(12,4," CA ")==0);
                }
                else     select_atom=(line.compare(12,4,atom_opt)==0);
                if (select_atom)
                {
                    if (!chainID)
                    {
                        chainID=line[21];
                        model_idx++;
                        stringstream i8_stream;
                        i=0;
                        if (split_opt==2) // split by chain
                        {
                            if (chainID==' ')
                            {
                                if (ter_opt>=1) i8_stream << ":_";
                                else i8_stream<<':'<<model_idx<<",_";
                            }
                            else
                            {
                                if (ter_opt>=1) i8_stream << ':' << chainID;
                                else i8_stream<<':'<<model_idx<<','<<chainID;
                            }
                            chainID_list.push_back(i8_stream.str());
                        }
                        else if (split_opt==1) // split by model
                        {
                            i8_stream << ':' << model_idx;
                            chainID_list.push_back(i8_stream.str());
                        }
                        PDB_lines.push_back(tmp_str_vec);
                        mol_vec.push_back(0);
                    }
                    else if (ter_opt>=2 && chainID!=line[21]) break;
                    if (split_opt==2 && chainID!=line[21])
                    {
                        chainID=line[21];
                        i=0;
                        stringstream i8_stream;
                        if (chainID==' ')
                        {
                            if (ter_opt>=1) i8_stream << ":_";
                            else i8_stream<<':'<<model_idx<<",_";
                        }
                        else
                        {
                            if (ter_opt>=1) i8_stream << ':' << chainID;
                            else i8_stream<<':'<<model_idx<<','<<chainID;
                        }
                        chainID_list.push_back(i8_stream.str());
                        PDB_lines.push_back(tmp_str_vec);
                        mol_vec.push_back(0);
                    }

                    if (resi==line.substr(22,5))
                        cerr<<"Warning! Duplicated residue "<<resi<<endl;
                    resi=line.substr(22,5); // including insertion code

                    PDB_lines.back().push_back(line);
                    if (line[17]==' ' && (line[18]=='D'||line[18]==' ')) mol_vec.back()++;
                    else mol_vec.back()--;
                    i++;
                }
            }
        }
    }
    else if (infmt_opt==1) // SPICKER format
    {
        size_t L=0;
        float x,y,z;
        stringstream i8_stream;
        while (compress_type?fin_gz.good():fin.good())
        {
            if (compress_type) fin_gz>>L>>x>>y>>z;
            else               fin   >>L>>x>>y>>z;
            if (compress_type) getline(fin_gz, line);
            else               getline(fin, line);
            if (!(compress_type?fin_gz.good():fin.good())) break;
            model_idx++;
            stringstream i8_stream;
            i8_stream << ':' << model_idx;
            chainID_list.push_back(i8_stream.str());
            PDB_lines.push_back(tmp_str_vec);
            mol_vec.push_back(0);
            for (i=0;i<L;i++)
            {
                if (compress_type) fin_gz>>x>>y>>z;
                else               fin   >>x>>y>>z;
                i8_stream<<"ATOM   "<<setw(4)<<i+1<<"  CA  UNK  "<<setw(4)
                    <<i+1<<"    "<<setiosflags(ios::fixed)<<setprecision(3)
                    <<setw(8)<<x<<setw(8)<<y<<setw(8)<<z;
                line=i8_stream.str();
                i8_stream.str(string());
                PDB_lines.back().push_back(line);
            }
            if (compress_type) getline(fin_gz, line);
            else               getline(fin, line);
        }
    }
    else if (infmt_opt==2) // xyz format
    {
        size_t L=0;
        stringstream i8_stream;
        while (compress_type?fin_gz.good():fin.good())
        {
            if (compress_type) getline(fin_gz, line);
            else               getline(fin, line);
            L=atoi(line.c_str());
            if (compress_type) getline(fin_gz, line);
            else               getline(fin, line);
            for (i=0;i<line.size();i++)
                if (line[i]==' '||line[i]=='\t') break;
            if (!(compress_type?fin_gz.good():fin.good())) break;
            chainID_list.push_back(':'+line.substr(0,i));
            PDB_lines.push_back(tmp_str_vec);
            mol_vec.push_back(0);
            for (i=0;i<L;i++)
            {
                if (compress_type) getline(fin_gz, line);
                else               getline(fin, line);
                i8_stream<<"ATOM   "<<setw(4)<<i+1<<"  CA  "
                    <<AAmap(line[0])<<"  "<<setw(4)<<i+1<<"    "
                    <<line.substr(2,8)<<line.substr(11,8)<<line.substr(20,8);
                line=i8_stream.str();
                i8_stream.str(string());
                PDB_lines.back().push_back(line);
                if (line[0]>='a' && line[0]<='z') mol_vec.back()++; // RNA
                else mol_vec.back()--;
            }
        }
    }
    else if (infmt_opt==3) // PDBx/mmCIF format
    {
        bool loop_ = false; // not reading following content
        map<string,int> _atom_site;
        int atom_site_pos;
        vector<string> line_vec;
        string alt_id=".";  // alternative location indicator
        string asym_id="."; // this is similar to chainID, except that
                            // chainID is char while asym_id is a string
                            // with possibly multiple char
        string prev_asym_id="";
        string AA="";       // residue name
        string atom="";
        string prev_resi="";
        string model_index=""; // the same as model_idx but type is string
        stringstream i8_stream;
        while (compress_type?fin_gz.good():fin.good())
        {
            if (compress_type) getline(fin_gz, line);
            else               getline(fin, line);
            if (line.size()==0) continue;
            if (loop_) loop_ = line.compare(0,2,"# ");
            if (!loop_)
            {
                if (line.compare(0,5,"loop_")) continue;
                while(1)
                {
                    if (compress_type)
                    {
                        if (fin_gz.good()) getline(fin_gz, line);
                        else PrintErrorAndQuit("ERROR! Unexpected end of "+filename);
                    }
                    else
                    {
                        if (fin.good()) getline(fin, line);
                        else PrintErrorAndQuit("ERROR! Unexpected end of "+filename);
                    }
                    if (line.size()) break;
                }
                if (line.compare(0,11,"_atom_site.")) continue;

                loop_=true;
                _atom_site.clear();
                atom_site_pos=0;
                _atom_site[line.substr(11,line.size()-12)]=atom_site_pos;

                while(1)
                {
                    if (compress_type) getline(fin_gz, line);
                    else               getline(fin, line);
                    if (line.size()==0) continue;
                    if (line.compare(0,11,"_atom_site.")) break;
                    _atom_site[line.substr(11,line.size()-12)]=++atom_site_pos;
                }


                if (_atom_site.count("group_PDB")*
                    _atom_site.count("label_atom_id")*
                    _atom_site.count("label_comp_id")*
                   (_atom_site.count("auth_asym_id")+
                    _atom_site.count("label_asym_id"))*
                   (_atom_site.count("auth_seq_id")+
                    _atom_site.count("label_seq_id"))*
                    _atom_site.count("Cartn_x")*
                    _atom_site.count("Cartn_y")*
                    _atom_site.count("Cartn_z")==0)
                {
                    loop_ = false;
                    cerr<<"Warning! Missing one of the following _atom_site data items: group_PDB, label_atom_id, label_comp_id, auth_asym_id/label_asym_id, auth_seq_id/label_seq_id, Cartn_x, Cartn_y, Cartn_z"<<endl;
                    continue;
                }
            }

            line_vec.clear();
            split(line,line_vec);
            if ((line_vec[_atom_site["group_PDB"]]!="ATOM" &&
                 line_vec[_atom_site["group_PDB"]]!="HETATM") ||
                (line_vec[_atom_site["group_PDB"]]=="HETATM" &&
                 (het_opt==0 || 
                 (het_opt==2 && line_vec[_atom_site["label_comp_id"]]!="MSE")))
                ) continue;
            
            alt_id=".";
            if (_atom_site.count("label_alt_id")) // in 39.4 % of entries
                alt_id=line_vec[_atom_site["label_alt_id"]];
            if (alt_id!="." && alt_id!="A") continue;

            atom=line_vec[_atom_site["label_atom_id"]];
            if (atom[0]=='"') atom=atom.substr(1);
            if (atom.size() && atom[atom.size()-1]=='"')
                atom=atom.substr(0,atom.size()-1);
            if (atom.size()==0) continue;
            if      (atom.size()==1) atom=" "+atom+"  ";
            else if (atom.size()==2) atom=" "+atom+" "; // wrong for sidechain H
            else if (atom.size()==3) atom=" "+atom;
            else if (atom.size()>=5) continue;

            AA=line_vec[_atom_site["label_comp_id"]]; // residue name
            if      (AA.size()==1) AA="  "+AA;
            else if (AA.size()==2) AA=" " +AA;
            else if (AA.size()>=4) continue;

            if (atom_opt=="auto")
            {
                if (AA[0]==' ' && (AA[1]=='D'||AA[1]==' ')) // DNA || RNA
                     select_atom=(atom==" C3'");
                else select_atom=(atom==" CA ");
            }
            else     select_atom=(atom==atom_opt);

            if (!select_atom) continue;

            if (_atom_site.count("auth_asym_id"))
                 asym_id=line_vec[_atom_site["auth_asym_id"]];
            else asym_id=line_vec[_atom_site["label_asym_id"]];
            if (asym_id==".") asym_id=" ";
            
            if (_atom_site.count("pdbx_PDB_model_num") && 
                model_index!=line_vec[_atom_site["pdbx_PDB_model_num"]])
            {
                model_index=line_vec[_atom_site["pdbx_PDB_model_num"]];
                if (PDB_lines.size() && ter_opt>=1) break;
                if (PDB_lines.size()==0 || split_opt>=1)
                {
                    PDB_lines.push_back(tmp_str_vec);
                    mol_vec.push_back(0);
                    prev_asym_id=asym_id;

                    if (split_opt==1 && ter_opt==0) chainID_list.push_back(
                        ':'+model_index);
                    else if (split_opt==2 && ter_opt==0)
                        chainID_list.push_back(':'+model_index+','+asym_id);
                    else //if (split_opt==2 && ter_opt==1)
                        chainID_list.push_back(':'+asym_id);
                    //else
                        //chainID_list.push_back("");
                }
            }

            if (prev_asym_id!=asym_id)
            {
                if (prev_asym_id!="" && ter_opt>=2) break;
                if (split_opt>=2)
                {
                    PDB_lines.push_back(tmp_str_vec);
                    mol_vec.push_back(0);

                    if (split_opt==1 && ter_opt==0) chainID_list.push_back(
                        ':'+model_index);
                    else if (split_opt==2 && ter_opt==0)
                        chainID_list.push_back(':'+model_index+','+asym_id);
                    else //if (split_opt==2 && ter_opt==1)
                        chainID_list.push_back(':'+asym_id);
                    //else
                        //chainID_list.push_back("");
                }
            }
            if (prev_asym_id!=asym_id) prev_asym_id=asym_id;

            if (AA[0]==' ' && (AA[1]=='D'||AA[1]==' ')) mol_vec.back()++;
            else mol_vec.back()--;

            if (_atom_site.count("auth_seq_id"))
                 resi=line_vec[_atom_site["auth_seq_id"]];
            else resi=line_vec[_atom_site["label_seq_id"]];
            if (_atom_site.count("pdbx_PDB_ins_code") && 
                line_vec[_atom_site["pdbx_PDB_ins_code"]]!="?")
                resi+=line_vec[_atom_site["pdbx_PDB_ins_code"]][0];
            else resi+=" ";

            if (prev_resi==resi)
                cerr<<"Warning! Duplicated residue "<<resi<<endl;
            prev_resi=resi;

            i++;
            i8_stream<<"ATOM  "
                <<setw(5)<<i<<" "<<atom<<" "<<AA<<" "<<asym_id[0]
                <<setw(5)<<resi.substr(0,5)<<"   "
                <<setw(8)<<line_vec[_atom_site["Cartn_x"]].substr(0,8)
                <<setw(8)<<line_vec[_atom_site["Cartn_y"]].substr(0,8)
                <<setw(8)<<line_vec[_atom_site["Cartn_z"]].substr(0,8);
            PDB_lines.back().push_back(i8_stream.str());
            i8_stream.str(string());
        }
        _atom_site.clear();
        line_vec.clear();
        alt_id.clear();
        asym_id.clear();
        AA.clear();
    }

    if (compress_type) fin_gz.close();
    else               fin.close();
    line.clear();
    if (!split_opt) chainID_list.push_back("");
    return PDB_lines.size();
}

/* read fasta file from filename. sequence is stored into FASTA_lines
 * while sequence name is stored into chainID_list.
 * if ter_opt >=1, only read the first sequence.
 * if ter_opt ==0, read all sequences.
 * if split_opt >=1 and ter_opt ==0, each sequence is a separate entry.
 * if split_opt ==0 and ter_opt ==0, all sequences are combined into one */
size_t get_FASTA_lines(const string filename,
    vector<vector<string> >&FASTA_lines, vector<string> &chainID_list,
    vector<int> &mol_vec, const int ter_opt=3, const int split_opt=0)
{
    string line;
    vector<string> tmp_str_vec;
    size_t l;
    
    int compress_type=0; // uncompressed file
    ifstream fin;
    redi::ipstream fin_gz; // if file is compressed
    if (filename.size()>=3 && 
        filename.substr(filename.size()-3,3)==".gz")
    {
        fin_gz.open("zcat '"+filename+"'");
        compress_type=1;
    }
    else if (filename.size()>=4 && 
        filename.substr(filename.size()-4,4)==".bz2")
    {
        fin_gz.open("bzcat '"+filename+"'");
        compress_type=2;
    }
    else fin.open(filename.c_str());

    while (compress_type?fin_gz.good():fin.good())
    {
        if (compress_type) getline(fin_gz, line);
        else               getline(fin, line);
        if (line.size()==0 || line[0]=='#') continue;

        if (line[0]=='>')
        {
            if (FASTA_lines.size())
            {
                if (ter_opt) break;
                if (split_opt==0) continue;
            }
            FASTA_lines.push_back(tmp_str_vec);
            FASTA_lines.back().push_back("");
            mol_vec.push_back(0);
            if (ter_opt==0 && split_opt)
            {
                line[0]=':';
                chainID_list.push_back(line);
            }
            else chainID_list.push_back("");
        }
        else
        {
            FASTA_lines.back()[0]+=line;
            for (l=0;l<line.size();l++) mol_vec.back()+=
                ('a'<=line[l] && line[l]<='z')-('A'<=line[l] && line[l]<='Z');
        }
    }

    line.clear();
    if (compress_type) fin_gz.close();
    else               fin.close();
    return FASTA_lines.size();
}


/* extract pairwise sequence alignment from residue index vectors,
 * assuming that "sequence" contains two empty strings.
 * return length of alignment, including gap. */
int extract_aln_from_resi(vector<string> &sequence, char *seqx, char *seqy,
    const vector<string> resi_vec1, const vector<string> resi_vec2,
    const int byresi_opt)
{
    sequence.clear();
    sequence.push_back("");
    sequence.push_back("");

    int i1=0; // positions in resi_vec1
    int i2=0; // positions in resi_vec2
    int xlen=resi_vec1.size();
    int ylen=resi_vec2.size();
    map<string,string> chainID_map1;
    map<string,string> chainID_map2;
    if (byresi_opt==3)
    {
        vector<string> chainID_vec;
        string chainID;
        stringstream ss;
        int i;
        for (i=0;i<xlen;i++)
        {
            chainID=resi_vec1[i].substr(5);
            if (!chainID_vec.size()|| chainID_vec.back()!=chainID)
            {
                chainID_vec.push_back(chainID);
                ss<<chainID_vec.size();
                chainID_map1[chainID]=ss.str();
                ss.str("");
            }
        }
        chainID_vec.clear();
        for (i=0;i<ylen;i++)
        {
            chainID=resi_vec2[i].substr(5);
            if (!chainID_vec.size()|| chainID_vec.back()!=chainID)
            {
                chainID_vec.push_back(chainID);
                ss<<chainID_vec.size();
                chainID_map2[chainID]=ss.str();
                ss.str("");
            }
        }
        vector<string>().swap(chainID_vec);
    }
    string chainID1="";
    string chainID2="";
    string chainID1_prev="";
    string chainID2_prev="";
    while(i1<xlen && i2<ylen)
    {
        if (byresi_opt==2)
        {
            chainID1=resi_vec1[i1].substr(5);
            chainID2=resi_vec2[i2].substr(5);
        }
        else if (byresi_opt==3)
        {
            chainID1=chainID_map1[resi_vec1[i1].substr(5)];
            chainID2=chainID_map2[resi_vec2[i2].substr(5)];
        }

        if (chainID1==chainID2)
        {
            if (atoi(resi_vec1[i1].substr(0,4).c_str())<
                atoi(resi_vec2[i2].substr(0,4).c_str()))
            {
                sequence[0]+=seqx[i1++];
                sequence[1]+='-';
            }
            else if (atoi(resi_vec1[i1].substr(0,4).c_str())>
                     atoi(resi_vec2[i2].substr(0,4).c_str()))
            {
                sequence[0]+='-';
                sequence[1]+=seqy[i2++];
            }
            else
            {
                sequence[0]+=seqx[i1++];
                sequence[1]+=seqy[i2++];
            }
            chainID1_prev=chainID1;
            chainID2_prev=chainID2;
        }
        else
        {
            if (chainID1_prev==chainID1 && chainID2_prev!=chainID2)
            {
                sequence[0]+=seqx[i1++];
                sequence[1]+='-';
                chainID1_prev=chainID1;
            }
            else if (chainID1_prev!=chainID1 && chainID2_prev==chainID2)
            {
                sequence[0]+='-';
                sequence[1]+=seqy[i2++];
                chainID2_prev=chainID2;
            }
            else
            {
                sequence[0]+=seqx[i1++];
                sequence[1]+=seqy[i2++];
                chainID1_prev=chainID1;
                chainID2_prev=chainID2;
            }
        }
        
    }
    map<string,string>().swap(chainID_map1);
    map<string,string>().swap(chainID_map2);
    chainID1.clear();
    chainID2.clear();
    chainID1_prev.clear();
    chainID2_prev.clear();
    return sequence[0].size();
}

int read_PDB(const vector<string> &PDB_lines, double **a, char *seq,
    vector<string> &resi_vec, const int read_resi)
{
    size_t i;
    for (i=0;i<PDB_lines.size();i++)
    {
        a[i][0] = atof(PDB_lines[i].substr(30, 8).c_str());
        a[i][1] = atof(PDB_lines[i].substr(38, 8).c_str());
        a[i][2] = atof(PDB_lines[i].substr(46, 8).c_str());
        seq[i]  = AAmap(PDB_lines[i].substr(17, 3));

        if (read_resi>=2) resi_vec.push_back(PDB_lines[i].substr(22,5)+
                                             PDB_lines[i][21]);
        if (read_resi==1) resi_vec.push_back(PDB_lines[i].substr(22,5));
    }
    seq[i]='\0'; 
    return i;
}

double dist(double x[3], double y[3])
{
    double d1=x[0]-y[0];
    double d2=x[1]-y[1];
    double d3=x[2]-y[2];
 
    return (d1*d1 + d2*d2 + d3*d3);
}

double dot(double *a, double *b)
{
    return (a[0] * b[0] + a[1] * b[1] + a[2] * b[2]);
}

void transform(double t[3], double u[3][3], double *x, double *x1)
{
    x1[0]=t[0]+dot(&u[0][0], x);
    x1[1]=t[1]+dot(&u[1][0], x);
    x1[2]=t[2]+dot(&u[2][0], x);
}

void do_rotation(double **x, double **x1, int len, double t[3], double u[3][3])
{
    for(int i=0; i<len; i++)
    {
        transform(t, u, &x[i][0], &x1[i][0]);
    }    
}

/* strip white space at the begining or end of string */
string Trim(const string &inputString)
{
    string result = inputString;
    int idxBegin = inputString.find_first_not_of(" \n\r\t");
    int idxEnd = inputString.find_last_not_of(" \n\r\t");
    if (idxBegin >= 0 && idxEnd >= 0)
        result = inputString.substr(idxBegin, idxEnd + 1 - idxBegin);
    return result;
}

/* read user specified pairwise alignment from 'fname_lign' to 'sequence'.
 * This function should only be called by main function, as it will
 * terminate a program if wrong alignment is given */
void read_user_alignment(vector<string>&sequence, const string &fname_lign,
    const int i_opt)
{
    if (fname_lign == "")
        PrintErrorAndQuit("Please provide a file name for option -i!");
    // open alignment file
    int n_p = 0;// number of structures in alignment file
    string line;
    
    ifstream fileIn(fname_lign.c_str());
    if (fileIn.is_open())
    {
        while (fileIn.good())
        {
            getline(fileIn, line);
            if (line.compare(0, 1, ">") == 0)// Flag for a new structure
            {
                if (n_p >= 2) break;
                sequence.push_back("");
                n_p++;
            }
            else if (n_p > 0 && line!="") sequence.back()+=line;
        }
        fileIn.close();
    }
    else PrintErrorAndQuit("ERROR! Alignment file does not exist.");
    
    if (n_p < 2)
        PrintErrorAndQuit("ERROR: Fasta format is wrong, two proteins should be included.");
    if (sequence[0].size() != sequence[1].size())
        PrintErrorAndQuit("ERROR! FASTA file is wrong. The length in alignment should be equal for the two aligned proteins.");
    if (i_opt==3)
    {
        int aligned_resNum=0;
        for (size_t i=0;i<sequence[0].size();i++)
            aligned_resNum+=(sequence[0][i]!='-' && sequence[1][i]!='-');
        if (aligned_resNum<3)
            PrintErrorAndQuit("ERROR! Superposition is undefined for <3 aligned residues.");
    }
    line.clear();
    return;
}

/* read list of entries from 'name' to 'chain_list'.
 * dir_opt is the folder name (prefix).
 * suffix_opt is the file name extension (suffix_opt).
 * This function should only be called by main function, as it will
 * terminate a program if wrong alignment is given */
void file2chainlist(vector<string>&chain_list, const string &name,
    const string &dir_opt, const string &suffix_opt)
{
    ifstream fp(name.c_str());
    if (! fp.is_open())
        PrintErrorAndQuit(("Can not open file: "+name+'\n').c_str());
    string line;
    while (fp.good())
    {
        getline(fp, line);
        if (! line.size()) continue;
        chain_list.push_back(dir_opt+Trim(line)+suffix_opt);
    }
    fp.close();
    line.clear();
}

/* These functions implement d0 normalization. The d0 for final TM-score
 * output is implemented by parameter_set4final. For both RNA alignment
 * and protein alignment, using d0 set by parameter_set4search yields
 * slightly better results during initial alignment-superposition iteration.
 */

void parameter_set4search(const int xlen, const int ylen,
    double &D0_MIN, double &Lnorm,
    double &score_d8, double &d0, double &d0_search, double &dcu0)
{
    //parameter initialization for searching: D0_MIN, Lnorm, d0, d0_search, score_d8
    D0_MIN=0.5; 
    dcu0=4.25;                       //update 3.85-->4.25
 
    Lnorm=getmin(xlen, ylen);        //normalize TMscore by this in searching
    if (Lnorm<=19)                    //update 15-->19
        d0=0.168;                   //update 0.5-->0.168
    else d0=(1.24*pow((Lnorm*1.0-15), 1.0/3)-1.8);
    D0_MIN=d0+0.8;              //this should be moved to above
    d0=D0_MIN;                  //update: best for search    

    d0_search=d0;
    if (d0_search>8)   d0_search=8;
    if (d0_search<4.5) d0_search=4.5;

    score_d8=1.5*pow(Lnorm*1.0, 0.3)+3.5; //remove pairs with dis>d8 during search & final
}

void parameter_set4final_C3prime(const double len, double &D0_MIN,
    double &Lnorm, double &d0, double &d0_search)
{
    D0_MIN=0.3; 
 
    Lnorm=len;            //normalize TMscore by this in searching
    if(Lnorm<=11) d0=0.3;
    else if(Lnorm>11&&Lnorm<=15) d0=0.4;
    else if(Lnorm>15&&Lnorm<=19) d0=0.5;
    else if(Lnorm>19&&Lnorm<=23) d0=0.6;
    else if(Lnorm>23&&Lnorm<30)  d0=0.7;
    else d0=(0.6*pow((Lnorm*1.0-0.5), 1.0/2)-2.5);

    d0_search=d0;
    if (d0_search>8)   d0_search=8;
    if (d0_search<4.5) d0_search=4.5;
}

void parameter_set4final(const double len, double &D0_MIN, double &Lnorm,
    double &d0, double &d0_search, const int mol_type)
{
    if (mol_type>0) // RNA
    {
        parameter_set4final_C3prime(len, D0_MIN, Lnorm,
            d0, d0_search);
        return;
    }
    D0_MIN=0.5; 
 
    Lnorm=len;            //normalize TMscore by this in searching
    if (Lnorm<=21) d0=0.5;          
    else d0=(1.24*pow((Lnorm*1.0-15), 1.0/3)-1.8);
    if (d0<D0_MIN) d0=D0_MIN;   
    d0_search=d0;
    if (d0_search>8)   d0_search=8;
    if (d0_search<4.5) d0_search=4.5;
}

void parameter_set4scale(const int len, const double d_s, double &Lnorm,
    double &d0, double &d0_search)
{
    d0=d_s;          
    Lnorm=len;            //normalize TMscore by this in searching
    d0_search=d0;
    if (d0_search>8)   d0_search=8;
    if (d0_search<4.5) d0_search=4.5;  
}

/**************************************************************************
Implemetation of Kabsch algoritm for finding the best rotation matrix
---------------------------------------------------------------------------
x    - x(i,m) are coordinates of atom m in set x            (input)
y    - y(i,m) are coordinates of atom m in set y            (input)
n    - n is number of atom pairs                            (input)
mode  - 0:calculate rms only                                (input)
1:calculate u,t only                                (takes medium)
2:calculate rms,u,t                                 (takes longer)
rms   - sum of w*(ux+t-y)**2 over all atom pairs            (output)
u    - u(i,j) is   rotation  matrix for best superposition  (output)
t    - t(i)   is translation vector for best superposition  (output)
**************************************************************************/
bool Kabsch(double **x, double **y, int n, int mode, double *rms,
    double t[3], double u[3][3])
{
    int i, j, m, m1, l, k;
    double e0, rms1, d, h, g;
    double cth, sth, sqrth, p, det, sigma;
    double xc[3], yc[3];
    double a[3][3], b[3][3], r[3][3], e[3], rr[6], ss[6];
    double sqrt3 = 1.73205080756888, tol = 0.01;
    int ip[] = { 0, 1, 3, 1, 2, 4, 3, 4, 5 };
    int ip2312[] = { 1, 2, 0, 1 };

    int a_failed = 0, b_failed = 0;
    double epsilon = 0.00000001;

    //initialization
    *rms = 0;
    rms1 = 0;
    e0 = 0;
    double c1[3], c2[3];
    double s1[3], s2[3];
    double sx[3], sy[3], sz[3];
    for (i = 0; i < 3; i++)
    {
        s1[i] = 0.0;
        s2[i] = 0.0;

        sx[i] = 0.0;
        sy[i] = 0.0;
        sz[i] = 0.0;
    }

    for (i = 0; i<3; i++)
    {
        xc[i] = 0.0;
        yc[i] = 0.0;
        t[i] = 0.0;
        for (j = 0; j<3; j++)
        {
            u[i][j] = 0.0;
            r[i][j] = 0.0;
            a[i][j] = 0.0;
            if (i == j)
            {
                u[i][j] = 1.0;
                a[i][j] = 1.0;
            }
        }
    }

    if (n<1) return false;

    //compute centers for vector sets x, y
    for (i = 0; i<n; i++)
    {
        for (j = 0; j < 3; j++)
        {
            c1[j] = x[i][j];
            c2[j] = y[i][j];

            s1[j] += c1[j];
            s2[j] += c2[j];
        }

        for (j = 0; j < 3; j++)
        {
            sx[j] += c1[0] * c2[j];
            sy[j] += c1[1] * c2[j];
            sz[j] += c1[2] * c2[j];
        }
    }
    for (i = 0; i < 3; i++)
    {
        xc[i] = s1[i] / n;
        yc[i] = s2[i] / n;
    }
    if (mode == 2 || mode == 0)
        for (int mm = 0; mm < n; mm++)
            for (int nn = 0; nn < 3; nn++)
                e0 += (x[mm][nn] - xc[nn]) * (x[mm][nn] - xc[nn]) + 
                      (y[mm][nn] - yc[nn]) * (y[mm][nn] - yc[nn]);
    for (j = 0; j < 3; j++)
    {
        r[j][0] = sx[j] - s1[0] * s2[j] / n;
        r[j][1] = sy[j] - s1[1] * s2[j] / n;
        r[j][2] = sz[j] - s1[2] * s2[j] / n;
    }

    //compute determinant of matrix r
    det = r[0][0] * (r[1][1] * r[2][2] - r[1][2] * r[2][1])\
        - r[0][1] * (r[1][0] * r[2][2] - r[1][2] * r[2][0])\
        + r[0][2] * (r[1][0] * r[2][1] - r[1][1] * r[2][0]);
    sigma = det;

    //compute tras(r)*r
    m = 0;
    for (j = 0; j<3; j++)
    {
        for (i = 0; i <= j; i++)
        {
            rr[m] = r[0][i] * r[0][j] + r[1][i] * r[1][j] + r[2][i] * r[2][j];
            m++;
        }
    }

    double spur = (rr[0] + rr[2] + rr[5]) / 3.0;
    double cof = (((((rr[2] * rr[5] - rr[4] * rr[4]) + rr[0] * rr[5])\
        - rr[3] * rr[3]) + rr[0] * rr[2]) - rr[1] * rr[1]) / 3.0;
    det = det*det;

    for (i = 0; i<3; i++) e[i] = spur;

    if (spur>0)
    {
        d = spur*spur;
        h = d - cof;
        g = (spur*cof - det) / 2.0 - spur*h;

        if (h>0)
        {
            sqrth = sqrt(h);
            d = h*h*h - g*g;
            if (d<0.0) d = 0.0;
            d = atan2(sqrt(d), -g) / 3.0;
            cth = sqrth * cos(d);
            sth = sqrth*sqrt3*sin(d);
            e[0] = (spur + cth) + cth;
            e[1] = (spur - cth) + sth;
            e[2] = (spur - cth) - sth;

            if (mode != 0)
            {//compute a                
                for (l = 0; l<3; l = l + 2)
                {
                    d = e[l];
                    ss[0] = (d - rr[2]) * (d - rr[5]) - rr[4] * rr[4];
                    ss[1] = (d - rr[5]) * rr[1] + rr[3] * rr[4];
                    ss[2] = (d - rr[0]) * (d - rr[5]) - rr[3] * rr[3];
                    ss[3] = (d - rr[2]) * rr[3] + rr[1] * rr[4];
                    ss[4] = (d - rr[0]) * rr[4] + rr[1] * rr[3];
                    ss[5] = (d - rr[0]) * (d - rr[2]) - rr[1] * rr[1];

                    if (fabs(ss[0]) <= epsilon) ss[0] = 0.0;
                    if (fabs(ss[1]) <= epsilon) ss[1] = 0.0;
                    if (fabs(ss[2]) <= epsilon) ss[2] = 0.0;
                    if (fabs(ss[3]) <= epsilon) ss[3] = 0.0;
                    if (fabs(ss[4]) <= epsilon) ss[4] = 0.0;
                    if (fabs(ss[5]) <= epsilon) ss[5] = 0.0;

                    if (fabs(ss[0]) >= fabs(ss[2]))
                    {
                        j = 0;
                        if (fabs(ss[0]) < fabs(ss[5])) j = 2;
                    }
                    else if (fabs(ss[2]) >= fabs(ss[5])) j = 1;
                    else j = 2;

                    d = 0.0;
                    j = 3 * j;
                    for (i = 0; i<3; i++)
                    {
                        k = ip[i + j];
                        a[i][l] = ss[k];
                        d = d + ss[k] * ss[k];
                    }


                    //if( d > 0.0 ) d = 1.0 / sqrt(d);
                    if (d > epsilon) d = 1.0 / sqrt(d);
                    else d = 0.0;
                    for (i = 0; i<3; i++) a[i][l] = a[i][l] * d;
                }//for l

                d = a[0][0] * a[0][2] + a[1][0] * a[1][2] + a[2][0] * a[2][2];
                if ((e[0] - e[1]) >(e[1] - e[2]))
                {
                    m1 = 2;
                    m = 0;
                }
                else
                {
                    m1 = 0;
                    m = 2;
                }
                p = 0;
                for (i = 0; i<3; i++)
                {
                    a[i][m1] = a[i][m1] - d*a[i][m];
                    p = p + a[i][m1] * a[i][m1];
                }
                if (p <= tol)
                {
                    p = 1.0;
                    for (i = 0; i<3; i++)
                    {
                        if (p < fabs(a[i][m])) continue;
                        p = fabs(a[i][m]);
                        j = i;
                    }
                    k = ip2312[j];
                    l = ip2312[j + 1];
                    p = sqrt(a[k][m] * a[k][m] + a[l][m] * a[l][m]);
                    if (p > tol)
                    {
                        a[j][m1] = 0.0;
                        a[k][m1] = -a[l][m] / p;
                        a[l][m1] = a[k][m] / p;
                    }
                    else a_failed = 1;
                }//if p<=tol
                else
                {
                    p = 1.0 / sqrt(p);
                    for (i = 0; i<3; i++) a[i][m1] = a[i][m1] * p;
                }//else p<=tol  
                if (a_failed != 1)
                {
                    a[0][1] = a[1][2] * a[2][0] - a[1][0] * a[2][2];
                    a[1][1] = a[2][2] * a[0][0] - a[2][0] * a[0][2];
                    a[2][1] = a[0][2] * a[1][0] - a[0][0] * a[1][2];
                }
            }//if(mode!=0)       
        }//h>0

        //compute b anyway
        if (mode != 0 && a_failed != 1)//a is computed correctly
        {
            //compute b
            for (l = 0; l<2; l++)
            {
                d = 0.0;
                for (i = 0; i<3; i++)
                {
                    b[i][l] = r[i][0] * a[0][l] + 
                              r[i][1] * a[1][l] + r[i][2] * a[2][l];
                    d = d + b[i][l] * b[i][l];
                }
                //if( d > 0 ) d = 1.0 / sqrt(d);
                if (d > epsilon) d = 1.0 / sqrt(d);
                else d = 0.0;
                for (i = 0; i<3; i++) b[i][l] = b[i][l] * d;
            }
            d = b[0][0] * b[0][1] + b[1][0] * b[1][1] + b[2][0] * b[2][1];
            p = 0.0;

            for (i = 0; i<3; i++)
            {
                b[i][1] = b[i][1] - d*b[i][0];
                p += b[i][1] * b[i][1];
            }

            if (p <= tol)
            {
                p = 1.0;
                for (i = 0; i<3; i++)
                {
                    if (p<fabs(b[i][0])) continue;
                    p = fabs(b[i][0]);
                    j = i;
                }
                k = ip2312[j];
                l = ip2312[j + 1];
                p = sqrt(b[k][0] * b[k][0] + b[l][0] * b[l][0]);
                if (p > tol)
                {
                    b[j][1] = 0.0;
                    b[k][1] = -b[l][0] / p;
                    b[l][1] = b[k][0] / p;
                }
                else b_failed = 1;
            }//if( p <= tol )
            else
            {
                p = 1.0 / sqrt(p);
                for (i = 0; i<3; i++) b[i][1] = b[i][1] * p;
            }
            if (b_failed != 1)
            {
                b[0][2] = b[1][0] * b[2][1] - b[1][1] * b[2][0];
                b[1][2] = b[2][0] * b[0][1] - b[2][1] * b[0][0];
                b[2][2] = b[0][0] * b[1][1] - b[0][1] * b[1][0];
                //compute u
                for (i = 0; i<3; i++)
                    for (j = 0; j<3; j++)
                        u[i][j] = b[i][0] * a[j][0] + 
                                  b[i][1] * a[j][1] + b[i][2] * a[j][2];
            }

            //compute t
            for (i = 0; i<3; i++)
                t[i] = ((yc[i] - u[i][0] * xc[0]) - u[i][1] * xc[1]) - 
                                                    u[i][2] * xc[2];
        }//if(mode!=0 && a_failed!=1)
    }//spur>0
    else //just compute t and errors
    {
        //compute t
        for (i = 0; i<3; i++)
            t[i] = ((yc[i] - u[i][0] * xc[0]) - u[i][1] * xc[1]) - 
                                                u[i][2] * xc[2];
    }//else spur>0 

    //compute rms
    for (i = 0; i<3; i++)
    {
        if (e[i] < 0) e[i] = 0;
        e[i] = sqrt(e[i]);
    }
    d = e[2];
    if (sigma < 0.0) d = -d;
    d = (d + e[1]) + e[0];

    if (mode == 2 || mode == 0)
    {
        rms1 = (e0 - d) - d;
        if (rms1 < 0.0) rms1 = 0.0;
    }

    *rms = rms1;
    return true;
}

/* Partial implementation of Needleman-Wunsch (NW) dynamic programming for
 * global alignment. The three NWDP_TM functions below are not complete
 * implementation of NW algorithm because gap jumping in the standard Gotoh
 * algorithm is not considered. Since the gap opening and gap extension is
 * the same, this is not a problem. This code was exploited in TM-align
 * because it is about 1.5 times faster than a complete NW implementation.
 * Nevertheless, if gap opening != gap extension shall be implemented in
 * the future, the Gotoh algorithm must be implemented. In rare scenarios,
 * it is also possible to have asymmetric alignment (i.e. 
 * TMalign A.pdb B.pdb and TMalign B.pdb A.pdb have different TM_A and TM_B
 * values) caused by the NWPD_TM implement.
 */

/* Input: score[1:len1, 1:len2], and gap_open
 * Output: j2i[1:len2] \in {1:len1} U {-1}
 * path[0:len1, 0:len2]=1,2,3, from diagonal, horizontal, vertical */
void NWDP_TM(double **score, bool **path, double **val,
    int len1, int len2, double gap_open, int j2i[])
{

    int i, j;
    double h, v, d;

    //initialization
    for(i=0; i<=len1; i++)
    {
        val[i][0]=0;
        //val[i][0]=i*gap_open;
        path[i][0]=false; //not from diagonal
    }

    for(j=0; j<=len2; j++)
    {
        val[0][j]=0;
        //val[0][j]=j*gap_open;
        path[0][j]=false; //not from diagonal
        j2i[j]=-1;    //all are not aligned, only use j2i[1:len2]
    }      


    //decide matrix and path
    for(i=1; i<=len1; i++)
    {
        for(j=1; j<=len2; j++)
        {
            d=val[i-1][j-1]+score[i][j]; //diagonal

            //symbol insertion in horizontal (= a gap in vertical)
            h=val[i-1][j];
            if(path[i-1][j]) h += gap_open; //aligned in last position

            //symbol insertion in vertical
            v=val[i][j-1];
            if(path[i][j-1]) v += gap_open; //aligned in last position


            if(d>=h && d>=v)
            {
                path[i][j]=true; //from diagonal
                val[i][j]=d;
            }
            else 
            {
                path[i][j]=false; //from horizontal
                if(v>=h) val[i][j]=v;
                else val[i][j]=h;
            }
        } //for i
    } //for j

    //trace back to extract the alignment
    i=len1;
    j=len2;
    while(i>0 && j>0)
    {
        if(path[i][j]) //from diagonal
        {
            j2i[j-1]=i-1;
            i--;
            j--;
        }
        else 
        {
            h=val[i-1][j];
            if(path[i-1][j]) h +=gap_open;

            v=val[i][j-1];
            if(path[i][j-1]) v +=gap_open;

            if(v>=h) j--;
            else i--;
        }
    }
}

/* Input: vectors x, y, rotation matrix t, u, scale factor d02, and gap_open
 * Output: j2i[1:len2] \in {1:len1} U {-1}
 * path[0:len1, 0:len2]=1,2,3, from diagonal, horizontal, vertical */
void NWDP_TM(bool **path, double **val, double **x, double **y,
    int len1, int len2, double t[3], double u[3][3],
    double d02, double gap_open, int j2i[])
{
    int i, j;
    double h, v, d;

    //initialization. use old val[i][0] and val[0][j] initialization
    //to minimize difference from TMalign fortran version
    for(i=0; i<=len1; i++)
    {
        val[i][0]=0;
        //val[i][0]=i*gap_open;
        path[i][0]=false; //not from diagonal
    }

    for(j=0; j<=len2; j++)
    {
        val[0][j]=0;
        //val[0][j]=j*gap_open;
        path[0][j]=false; //not from diagonal
        j2i[j]=-1;    //all are not aligned, only use j2i[1:len2]
    }      
    double xx[3], dij;


    //decide matrix and path
    for(i=1; i<=len1; i++)
    {
        transform(t, u, &x[i-1][0], xx);
        for(j=1; j<=len2; j++)
        {
            dij=dist(xx, &y[j-1][0]);    
            d=val[i-1][j-1] +  1.0/(1+dij/d02);

            //symbol insertion in horizontal (= a gap in vertical)
            h=val[i-1][j];
            if(path[i-1][j]) h += gap_open; //aligned in last position

            //symbol insertion in vertical
            v=val[i][j-1];
            if(path[i][j-1]) v += gap_open; //aligned in last position


            if(d>=h && d>=v)
            {
                path[i][j]=true; //from diagonal
                val[i][j]=d;
            }
            else 
            {
                path[i][j]=false; //from horizontal
                if(v>=h) val[i][j]=v;
                else val[i][j]=h;
            }
        } //for i
    } //for j

    //trace back to extract the alignment
    i=len1;
    j=len2;
    while(i>0 && j>0)
    {
        if(path[i][j]) //from diagonal
        {
            j2i[j-1]=i-1;
            i--;
            j--;
        }
        else 
        {
            h=val[i-1][j];
            if(path[i-1][j]) h +=gap_open;

            v=val[i][j-1];
            if(path[i][j-1]) v +=gap_open;

            if(v>=h) j--;
            else i--;
        }
    }
}

/* This is the same as the previous NWDP_TM, except for the lack of rotation
 * Input: vectors x, y, scale factor d02, and gap_open
 * Output: j2i[1:len2] \in {1:len1} U {-1}
 * path[0:len1, 0:len2]=1,2,3, from diagonal, horizontal, vertical */
void NWDP_SE(bool **path, double **val, double **x, double **y,
    int len1, int len2, double d02, double gap_open, int j2i[])
{
    int i, j;
    double h, v, d;

    for(i=0; i<=len1; i++)
    {
        val[i][0]=0;
        path[i][0]=false; //not from diagonal
    }

    for(j=0; j<=len2; j++)
    {
        val[0][j]=0;
        path[0][j]=false; //not from diagonal
        j2i[j]=-1;    //all are not aligned, only use j2i[1:len2]
    }      
    double dij;

    //decide matrix and path
    for(i=1; i<=len1; i++)
    {
        for(j=1; j<=len2; j++)
        {
            dij=dist(&x[i-1][0], &y[j-1][0]);    
            d=val[i-1][j-1] +  1.0/(1+dij/d02);

            //symbol insertion in horizontal (= a gap in vertical)
            h=val[i-1][j];
            if(path[i-1][j]) h += gap_open; //aligned in last position

            //symbol insertion in vertical
            v=val[i][j-1];
            if(path[i][j-1]) v += gap_open; //aligned in last position


            if(d>=h && d>=v)
            {
                path[i][j]=true; //from diagonal
                val[i][j]=d;
            }
            else 
            {
                path[i][j]=false; //from horizontal
                if(v>=h) val[i][j]=v;
                else val[i][j]=h;
            }
        } //for i
    } //for j

    //trace back to extract the alignment
    i=len1;
    j=len2;
    while(i>0 && j>0)
    {
        if(path[i][j]) //from diagonal
        {
            j2i[j-1]=i-1;
            i--;
            j--;
        }
        else 
        {
            h=val[i-1][j];
            if(path[i-1][j]) h +=gap_open;

            v=val[i][j-1];
            if(path[i][j-1]) v +=gap_open;

            if(v>=h) j--;
            else i--;
        }
    }
}

/* +ss
 * Input: secondary structure secx, secy, and gap_open
 * Output: j2i[1:len2] \in {1:len1} U {-1}
 * path[0:len1, 0:len2]=1,2,3, from diagonal, horizontal, vertical */
void NWDP_TM(bool **path, double **val, const char *secx, const char *secy,
    const int len1, const int len2, const double gap_open, int j2i[])
{

    int i, j;
    double h, v, d;

    //initialization
    for(i=0; i<=len1; i++)
    {
        val[i][0]=0;
        //val[i][0]=i*gap_open;
        path[i][0]=false; //not from diagonal
    }

    for(j=0; j<=len2; j++)
    {
        val[0][j]=0;
        //val[0][j]=j*gap_open;
        path[0][j]=false; //not from diagonal
        j2i[j]=-1;    //all are not aligned, only use j2i[1:len2]
    }      

    //decide matrix and path
    for(i=1; i<=len1; i++)
    {
        for(j=1; j<=len2; j++)
        {
            d=val[i-1][j-1] + 1.0*(secx[i-1]==secy[j-1]);

            //symbol insertion in horizontal (= a gap in vertical)
            h=val[i-1][j];
            if(path[i-1][j]) h += gap_open; //aligned in last position

            //symbol insertion in vertical
            v=val[i][j-1];
            if(path[i][j-1]) v += gap_open; //aligned in last position

            if(d>=h && d>=v)
            {
                path[i][j]=true; //from diagonal
                val[i][j]=d;
            }
            else 
            {
                path[i][j]=false; //from horizontal
                if(v>=h) val[i][j]=v;
                else val[i][j]=h;
            }
        } //for i
    } //for j

    //trace back to extract the alignment
    i=len1;
    j=len2;
    while(i>0 && j>0)
    {
        if(path[i][j]) //from diagonal
        {
            j2i[j-1]=i-1;
            i--;
            j--;
        }
        else 
        {
            h=val[i-1][j];
            if(path[i-1][j]) h +=gap_open;

            v=val[i][j-1];
            if(path[i][j-1]) v +=gap_open;

            if(v>=h) j--;
            else i--;
        }
    }
}

/* Functions for the core TMalign algorithm, including the entry function
 * TMalign_main */

//     1, collect those residues with dis<d;
//     2, calculate TMscore
int score_fun8( double **xa, double **ya, int n_ali, double d, int i_ali[],
    double *score1, int score_sum_method, const double Lnorm, 
    const double score_d8, const double d0)
{
    double score_sum=0, di;
    double d_tmp=d*d;
    double d02=d0*d0;
    double score_d8_cut = score_d8*score_d8;
    
    int i, n_cut, inc=0;

    while(1)
    {
        n_cut=0;
        score_sum=0;
        for(i=0; i<n_ali; i++)
        {
            di = dist(xa[i], ya[i]);
            if(di<d_tmp)
            {
                i_ali[n_cut]=i;
                n_cut++;
            }
            if(score_sum_method==8)
            {                
                if(di<=score_d8_cut) score_sum += 1/(1+di/d02);
            }
            else score_sum += 1/(1+di/d02);
        }
        //there are not enough feasible pairs, relieve the threshold         
        if(n_cut<3 && n_ali>3)
        {
            inc++;
            double dinc=(d+inc*0.5);
            d_tmp = dinc * dinc;
        }
        else break;
    }  

    *score1=score_sum/Lnorm;
    return n_cut;
}

int score_fun8_standard(double **xa, double **ya, int n_ali, double d,
    int i_ali[], double *score1, int score_sum_method,
    double score_d8, double d0)
{
    double score_sum = 0, di;
    double d_tmp = d*d;
    double d02 = d0*d0;
    double score_d8_cut = score_d8*score_d8;

    int i, n_cut, inc = 0;
    while (1)
    {
        n_cut = 0;
        score_sum = 0;
        for (i = 0; i<n_ali; i++)
        {
            di = dist(xa[i], ya[i]);
            if (di<d_tmp)
            {
                i_ali[n_cut] = i;
                n_cut++;
            }
            if (score_sum_method == 8)
            {
                if (di <= score_d8_cut) score_sum += 1 / (1 + di / d02);
            }
            else
            {
                score_sum += 1 / (1 + di / d02);
            }
        }
        //there are not enough feasible pairs, relieve the threshold         
        if (n_cut<3 && n_ali>3)
        {
            inc++;
            double dinc = (d + inc*0.5);
            d_tmp = dinc * dinc;
        }
        else break;
    }

    *score1 = score_sum / n_ali;
    return n_cut;
}

double TMscore8_search(double **r1, double **r2, double **xtm, double **ytm,
    double **xt, int Lali, double t0[3], double u0[3][3], int simplify_step,
    int score_sum_method, double *Rcomm, double local_d0_search, double Lnorm,
    double score_d8, double d0)
{
    int i, m;
    double score_max, score, rmsd;    
    const int kmax=Lali;    
    int k_ali[kmax], ka, k;
    double t[3];
    double u[3][3];
    double d;
    

    //iterative parameters
    int n_it=20;            //maximum number of iterations
    int n_init_max=6; //maximum number of different fragment length 
    int L_ini[n_init_max];  //fragment lengths, Lali, Lali/2, Lali/4 ... 4   
    int L_ini_min=4;
    if(Lali<L_ini_min) L_ini_min=Lali;   

    int n_init=0, i_init;      
    for(i=0; i<n_init_max-1; i++)
    {
        n_init++;
        L_ini[i]=(int) (Lali/pow(2.0, (double) i));
        if(L_ini[i]<=L_ini_min)
        {
            L_ini[i]=L_ini_min;
            break;
        }
    }
    if(i==n_init_max-1)
    {
        n_init++;
        L_ini[i]=L_ini_min;
    }
    
    score_max=-1;
    //find the maximum score starting from local structures superposition
    int i_ali[kmax], n_cut;
    int L_frag; //fragment length
    int iL_max; //maximum starting position for the fragment
    
    for(i_init=0; i_init<n_init; i_init++)
    {
        L_frag=L_ini[i_init];
        iL_max=Lali-L_frag;
      
        i=0;   
        while(1)
        {
            //extract the fragment starting from position i 
            ka=0;
            for(k=0; k<L_frag; k++)
            {
                int kk=k+i;
                r1[k][0]=xtm[kk][0];  
                r1[k][1]=xtm[kk][1]; 
                r1[k][2]=xtm[kk][2];   
                
                r2[k][0]=ytm[kk][0];  
                r2[k][1]=ytm[kk][1]; 
                r2[k][2]=ytm[kk][2];
                
                k_ali[ka]=kk;
                ka++;
            }
            
            //extract rotation matrix based on the fragment
            Kabsch(r1, r2, L_frag, 1, &rmsd, t, u);
            if (simplify_step != 1)
                *Rcomm = 0;
            do_rotation(xtm, xt, Lali, t, u);
            
            //get subsegment of this fragment
            d = local_d0_search - 1;
            n_cut=score_fun8(xt, ytm, Lali, d, i_ali, &score, 
                score_sum_method, Lnorm, score_d8, d0);
            if(score>score_max)
            {
                score_max=score;
                
                //save the rotation matrix
                for(k=0; k<3; k++)
                {
                    t0[k]=t[k];
                    u0[k][0]=u[k][0];
                    u0[k][1]=u[k][1];
                    u0[k][2]=u[k][2];
                }
            }
            
            //try to extend the alignment iteratively            
            d = local_d0_search + 1;
            for(int it=0; it<n_it; it++)            
            {
                ka=0;
                for(k=0; k<n_cut; k++)
                {
                    m=i_ali[k];
                    r1[k][0]=xtm[m][0];  
                    r1[k][1]=xtm[m][1]; 
                    r1[k][2]=xtm[m][2];
                    
                    r2[k][0]=ytm[m][0];  
                    r2[k][1]=ytm[m][1]; 
                    r2[k][2]=ytm[m][2];
                    
                    k_ali[ka]=m;
                    ka++;
                } 
                //extract rotation matrix based on the fragment                
                Kabsch(r1, r2, n_cut, 1, &rmsd, t, u);
                do_rotation(xtm, xt, Lali, t, u);
                n_cut=score_fun8(xt, ytm, Lali, d, i_ali, &score, 
                    score_sum_method, Lnorm, score_d8, d0);
                if(score>score_max)
                {
                    score_max=score;

                    //save the rotation matrix
                    for(k=0; k<3; k++)
                    {
                        t0[k]=t[k];
                        u0[k][0]=u[k][0];
                        u0[k][1]=u[k][1];
                        u0[k][2]=u[k][2];
                    }                     
                }
                
                //check if it converges            
                if(n_cut==ka)
                {                
                    for(k=0; k<n_cut; k++)
                    {
                        if(i_ali[k]!=k_ali[k]) break;
                    }
                    if(k==n_cut) break;
                }                                                               
            } //for iteration            

            if(i<iL_max)
            {
                i=i+simplify_step; //shift the fragment        
                if(i>iL_max) i=iL_max;  //do this to use the last missed fragment
            }
            else if(i>=iL_max) break;
        }//while(1)
        //end of one fragment
    }//for(i_init
    return score_max;
}


double TMscore8_search_standard( double **r1, double **r2,
    double **xtm, double **ytm, double **xt, int Lali,
    double t0[3], double u0[3][3], int simplify_step, int score_sum_method,
    double *Rcomm, double local_d0_search, double score_d8, double d0)
{
    int i, m;
    double score_max, score, rmsd;
    const int kmax = Lali;
    int k_ali[kmax], ka, k;
    double t[3];
    double u[3][3];
    double d;

    //iterative parameters
    int n_it = 20;            //maximum number of iterations
    int n_init_max = 6; //maximum number of different fragment length 
    int L_ini[n_init_max];  //fragment lengths, Lali, Lali/2, Lali/4 ... 4   
    int L_ini_min = 4;
    if (Lali<L_ini_min) L_ini_min = Lali;

    int n_init = 0, i_init;
    for (i = 0; i<n_init_max - 1; i++)
    {
        n_init++;
        L_ini[i] = (int)(Lali / pow(2.0, (double)i));
        if (L_ini[i] <= L_ini_min)
        {
            L_ini[i] = L_ini_min;
            break;
        }
    }
    if (i == n_init_max - 1)
    {
        n_init++;
        L_ini[i] = L_ini_min;
    }

    score_max = -1;
    //find the maximum score starting from local structures superposition
    int i_ali[kmax], n_cut;
    int L_frag; //fragment length
    int iL_max; //maximum starting position for the fragment

    for (i_init = 0; i_init<n_init; i_init++)
    {
        L_frag = L_ini[i_init];
        iL_max = Lali - L_frag;

        i = 0;
        while (1)
        {
            //extract the fragment starting from position i 
            ka = 0;
            for (k = 0; k<L_frag; k++)
            {
                int kk = k + i;
                r1[k][0] = xtm[kk][0];
                r1[k][1] = xtm[kk][1];
                r1[k][2] = xtm[kk][2];

                r2[k][0] = ytm[kk][0];
                r2[k][1] = ytm[kk][1];
                r2[k][2] = ytm[kk][2];

                k_ali[ka] = kk;
                ka++;
            }
            //extract rotation matrix based on the fragment
            Kabsch(r1, r2, L_frag, 1, &rmsd, t, u);
            if (simplify_step != 1)
                *Rcomm = 0;
            do_rotation(xtm, xt, Lali, t, u);

            //get subsegment of this fragment
            d = local_d0_search - 1;
            n_cut = score_fun8_standard(xt, ytm, Lali, d, i_ali, &score,
                score_sum_method, score_d8, d0);

            if (score>score_max)
            {
                score_max = score;

                //save the rotation matrix
                for (k = 0; k<3; k++)
                {
                    t0[k] = t[k];
                    u0[k][0] = u[k][0];
                    u0[k][1] = u[k][1];
                    u0[k][2] = u[k][2];
                }
            }

            //try to extend the alignment iteratively            
            d = local_d0_search + 1;
            for (int it = 0; it<n_it; it++)
            {
                ka = 0;
                for (k = 0; k<n_cut; k++)
                {
                    m = i_ali[k];
                    r1[k][0] = xtm[m][0];
                    r1[k][1] = xtm[m][1];
                    r1[k][2] = xtm[m][2];

                    r2[k][0] = ytm[m][0];
                    r2[k][1] = ytm[m][1];
                    r2[k][2] = ytm[m][2];

                    k_ali[ka] = m;
                    ka++;
                }
                //extract rotation matrix based on the fragment                
                Kabsch(r1, r2, n_cut, 1, &rmsd, t, u);
                do_rotation(xtm, xt, Lali, t, u);
                n_cut = score_fun8_standard(xt, ytm, Lali, d, i_ali, &score,
                    score_sum_method, score_d8, d0);
                if (score>score_max)
                {
                    score_max = score;

                    //save the rotation matrix
                    for (k = 0; k<3; k++)
                    {
                        t0[k] = t[k];
                        u0[k][0] = u[k][0];
                        u0[k][1] = u[k][1];
                        u0[k][2] = u[k][2];
                    }
                }

                //check if it converges            
                if (n_cut == ka)
                {
                    for (k = 0; k<n_cut; k++)
                    {
                        if (i_ali[k] != k_ali[k]) break;
                    }
                    if (k == n_cut) break;
                }
            } //for iteration            

            if (i<iL_max)
            {
                i = i + simplify_step; //shift the fragment        
                if (i>iL_max) i = iL_max;  //do this to use the last missed fragment
            }
            else if (i >= iL_max) break;
        }//while(1)
        //end of one fragment
    }//for(i_init
    return score_max;
}

//Comprehensive TMscore search engine
// input:   two vector sets: x, y
//          an alignment invmap0[] between x and y
//          simplify_step: 1 or 40 or other integers
//          score_sum_method: 0 for score over all pairs
//                            8 for socre over the pairs with dist<score_d8
// output:  the best rotaion matrix t, u that results in highest TMscore
double detailed_search(double **r1, double **r2, double **xtm, double **ytm,
    double **xt, double **x, double **y, int xlen, int ylen, 
    int invmap0[], double t[3], double u[3][3], int simplify_step,
    int score_sum_method, double local_d0_search, double Lnorm,
    double score_d8, double d0)
{
    //x is model, y is template, try to superpose onto y
    int i, j, k;     
    double tmscore;
    double rmsd;

    k=0;
    for(i=0; i<ylen; i++) 
    {
        j=invmap0[i];
        if(j>=0) //aligned
        {
            xtm[k][0]=x[j][0];
            xtm[k][1]=x[j][1];
            xtm[k][2]=x[j][2];
                
            ytm[k][0]=y[i][0];
            ytm[k][1]=y[i][1];
            ytm[k][2]=y[i][2];
            k++;
        }
    }

    //detailed search 40-->1
    tmscore = TMscore8_search(r1, r2, xtm, ytm, xt, k, t, u, simplify_step,
        score_sum_method, &rmsd, local_d0_search, Lnorm, score_d8, d0);
    return tmscore;
}

double detailed_search_standard( double **r1, double **r2,
    double **xtm, double **ytm, double **xt, double **x, double **y,
    int xlen, int ylen, int invmap0[], double t[3], double u[3][3],
    int simplify_step, int score_sum_method, double local_d0_search,
    const bool& bNormalize, double Lnorm, double score_d8, double d0)
{
    //x is model, y is template, try to superpose onto y
    int i, j, k;     
    double tmscore;
    double rmsd;

    k=0;
    for(i=0; i<ylen; i++) 
    {
        j=invmap0[i];
        if(j>=0) //aligned
        {
            xtm[k][0]=x[j][0];
            xtm[k][1]=x[j][1];
            xtm[k][2]=x[j][2];
                
            ytm[k][0]=y[i][0];
            ytm[k][1]=y[i][1];
            ytm[k][2]=y[i][2];
            k++;
        }
    }

    //detailed search 40-->1
    tmscore = TMscore8_search_standard( r1, r2, xtm, ytm, xt, k, t, u,
        simplify_step, score_sum_method, &rmsd, local_d0_search, score_d8, d0);
    if (bNormalize)// "-i", to use standard_TMscore, then bNormalize=true, else bNormalize=false; 
        tmscore = tmscore * k / Lnorm;

    return tmscore;
}

//compute the score quickly in three iterations
double get_score_fast( double **r1, double **r2, double **xtm, double **ytm,
    double **x, double **y, int xlen, int ylen, int invmap[],
    double d0, double d0_search, double t[3], double u[3][3])
{
    double rms, tmscore, tmscore1, tmscore2;
    int i, j, k;

    k=0;
    for(j=0; j<ylen; j++)
    {
        i=invmap[j];
        if(i>=0)
        {
            r1[k][0]=x[i][0];
            r1[k][1]=x[i][1];
            r1[k][2]=x[i][2];

            r2[k][0]=y[j][0];
            r2[k][1]=y[j][1];
            r2[k][2]=y[j][2];
            
            xtm[k][0]=x[i][0];
            xtm[k][1]=x[i][1];
            xtm[k][2]=x[i][2];
            
            ytm[k][0]=y[j][0];
            ytm[k][1]=y[j][1];
            ytm[k][2]=y[j][2];                  
            
            k++;
        }
        else if(i!=-1) PrintErrorAndQuit("Wrong map!\n");
    }
    Kabsch(r1, r2, k, 1, &rms, t, u);
    
    //evaluate score   
    double di;
    const int len=k;
    double dis[len];    
    double d00=d0_search;
    double d002=d00*d00;
    double d02=d0*d0;
    
    int n_ali=k;
    double xrot[3];
    tmscore=0;
    for(k=0; k<n_ali; k++)
    {
        transform(t, u, &xtm[k][0], xrot);        
        di=dist(xrot, &ytm[k][0]);
        dis[k]=di;
        tmscore += 1/(1+di/d02);
    }
    
   
   
   //second iteration 
    double d002t=d002;
    while(1)
    {
        j=0;
        for(k=0; k<n_ali; k++)
        {            
            if(dis[k]<=d002t)
            {
                r1[j][0]=xtm[k][0];
                r1[j][1]=xtm[k][1];
                r1[j][2]=xtm[k][2];
                
                r2[j][0]=ytm[k][0];
                r2[j][1]=ytm[k][1];
                r2[j][2]=ytm[k][2];
                
                j++;
            }
        }
        //there are not enough feasible pairs, relieve the threshold 
        if(j<3 && n_ali>3) d002t += 0.5;
        else break;
    }
    
    if(n_ali!=j)
    {
        Kabsch(r1, r2, j, 1, &rms, t, u);
        tmscore1=0;
        for(k=0; k<n_ali; k++)
        {
            transform(t, u, &xtm[k][0], xrot);        
            di=dist(xrot, &ytm[k][0]);
            dis[k]=di;
            tmscore1 += 1/(1+di/d02);
        }
        
        //third iteration
        d002t=d002+1;
       
        while(1)
        {
            j=0;
            for(k=0; k<n_ali; k++)
            {            
                if(dis[k]<=d002t)
                {
                    r1[j][0]=xtm[k][0];
                    r1[j][1]=xtm[k][1];
                    r1[j][2]=xtm[k][2];
                    
                    r2[j][0]=ytm[k][0];
                    r2[j][1]=ytm[k][1];
                    r2[j][2]=ytm[k][2];
                                        
                    j++;
                }
            }
            //there are not enough feasible pairs, relieve the threshold 
            if(j<3 && n_ali>3) d002t += 0.5;
            else break;
        }

        //evaluate the score
        Kabsch(r1, r2, j, 1, &rms, t, u);
        tmscore2=0;
        for(k=0; k<n_ali; k++)
        {
            transform(t, u, &xtm[k][0], xrot);
            di=dist(xrot, &ytm[k][0]);
            tmscore2 += 1/(1+di/d02);
        }    
    }
    else
    {
        tmscore1=tmscore;
        tmscore2=tmscore;
    }
    
    if(tmscore1>=tmscore) tmscore=tmscore1;
    if(tmscore2>=tmscore) tmscore=tmscore2;
    return tmscore; // no need to normalize this score because it will not be used for latter scoring
}


//perform gapless threading to find the best initial alignment
//input: x, y, xlen, ylen
//output: y2x0 stores the best alignment: e.g., 
//y2x0[j]=i means:
//the jth element in y is aligned to the ith element in x if i>=0 
//the jth element in y is aligned to a gap in x if i==-1
double get_initial(double **r1, double **r2, double **xtm, double **ytm,
    double **x, double **y, int xlen, int ylen, int *y2x,
    double d0, double d0_search, const bool fast_opt,
    double t[3], double u[3][3])
{
    int min_len=getmin(xlen, ylen);
    if(min_len<3) PrintErrorAndQuit("Sequence is too short <3!\n");
    
    int min_ali= min_len/2;              //minimum size of considered fragment 
    if(min_ali<=5)  min_ali=5;    
    int n1, n2;
    n1 = -ylen+min_ali; 
    n2 = xlen-min_ali;

    int i, j, k, k_best;
    double tmscore, tmscore_max=-1;

    k_best=n1;
    for(k=n1; k<=n2; k+=(fast_opt)?5:1)
    {
        //get the map
        for(j=0; j<ylen; j++)
        {
            i=j+k;
            if(i>=0 && i<xlen) y2x[j]=i;
            else y2x[j]=-1;
        }
        
        //evaluate the map quickly in three iterations
        //this is not real tmscore, it is used to evaluate the goodness of the initial alignment
        tmscore=get_score_fast(r1, r2, xtm, ytm,
            x, y, xlen, ylen, y2x, d0,d0_search, t, u);
        if(tmscore>=tmscore_max)
        {
            tmscore_max=tmscore;
            k_best=k;
        }
    }
    
    //extract the best map
    k=k_best;
    for(j=0; j<ylen; j++)
    {
        i=j+k;
        if(i>=0 && i<xlen) y2x[j]=i;
        else y2x[j]=-1;
    }    

    return tmscore_max;
}

void smooth(int *sec, int len)
{
    int i, j;
    //smooth single  --x-- => -----
    for (i=2; i<len-2; i++)
    {
        if(sec[i]==2 || sec[i]==4)
        {
            j=sec[i];
            if (sec[i-2]!=j && sec[i-1]!=j && sec[i+1]!=j && sec[i+2]!=j)
                sec[i]=1;
        }
    }

    //   smooth double 
    //   --xx-- => ------
    for (i=0; i<len-5; i++)
    {
        //helix
        if (sec[i]!=2   && sec[i+1]!=2 && sec[i+2]==2 && sec[i+3]==2 &&
            sec[i+4]!=2 && sec[i+5]!= 2)
        {
            sec[i+2]=1;
            sec[i+3]=1;
        }

        //beta
        if (sec[i]!=4   && sec[i+1]!=4 && sec[i+2]==4 && sec[i+3]==4 &&
            sec[i+4]!=4 && sec[i+5]!= 4)
        {
            sec[i+2]=1;
            sec[i+3]=1;
        }
    }

    //smooth connect
    for (i=0; i<len-2; i++)
    {        
        if (sec[i]==2 && sec[i+1]!=2 && sec[i+2]==2) sec[i+1]=2;
        else if(sec[i]==4 && sec[i+1]!=4 && sec[i+2]==4) sec[i+1]=4;
    }

}

char sec_str(double dis13, double dis14, double dis15,
            double dis24, double dis25, double dis35)
{
    char s='C';
    
    double delta=2.1;
    if (fabs(dis15-6.37)<delta && fabs(dis14-5.18)<delta && 
        fabs(dis25-5.18)<delta && fabs(dis13-5.45)<delta &&
        fabs(dis24-5.45)<delta && fabs(dis35-5.45)<delta)
    {
        s='H'; //helix                        
        return s;
    }

    delta=1.42;
    if (fabs(dis15-13  )<delta && fabs(dis14-10.4)<delta &&
        fabs(dis25-10.4)<delta && fabs(dis13-6.1 )<delta &&
        fabs(dis24-6.1 )<delta && fabs(dis35-6.1 )<delta)
    {
        s='E'; //strand
        return s;
    }

    if (dis15 < 8) s='T'; //turn
    return s;
}


/* secondary structure assignment for protein:
 * 1->coil, 2->helix, 3->turn, 4->strand */
void make_sec(double **x, int len, char *sec)
{
    int j1, j2, j3, j4, j5;
    double d13, d14, d15, d24, d25, d35;
    for(int i=0; i<len; i++)
    {     
        sec[i]='C';
        j1=i-2;
        j2=i-1;
        j3=i;
        j4=i+1;
        j5=i+2;        
        
        if(j1>=0 && j5<len)
        {
            d13=sqrt(dist(x[j1], x[j3]));
            d14=sqrt(dist(x[j1], x[j4]));
            d15=sqrt(dist(x[j1], x[j5]));
            d24=sqrt(dist(x[j2], x[j4]));
            d25=sqrt(dist(x[j2], x[j5]));
            d35=sqrt(dist(x[j3], x[j5]));
            sec[i]=sec_str(d13, d14, d15, d24, d25, d35);            
        }    
    } 
    sec[len]=0;
}

/* a c d b: a paired to b, c paired to d */
bool overlap(const int a1,const int b1,const int c1,const int d1,
             const int a2,const int b2,const int c2,const int d2)
{
    return (a2>=a1&&a2<=c1)||(c2>=a1&&c2<=c1)||
           (d2>=a1&&d2<=c1)||(b2>=a1&&b2<=c1)||
           (a2>=d1&&a2<=b1)||(c2>=d1&&c2<=b1)||
           (d2>=d1&&d2<=b1)||(b2>=d1&&b2<=b1);
}

/* find base pairing stacks in RNA*/
void sec_str(int len,char *seq, const vector<vector<bool> >&bp, 
    int a, int b,int &c, int &d)
{
    int i;
    
    for (i=0;i<len;i++)
    {
        if (a+i<len-3 && b-i>0)
        {
            if (a+i<b-i && bp[a+i][b-i]) continue;
            break;
        }
    }
    c=a+i-1;d=b-i+1;
}

/* secondary structure assignment for RNA:
 * 1->unpair, 2->paired with upstream, 3->paired with downstream */
void make_sec(char *seq, double **x, int len, char *sec,const string atom_opt)
{
    int ii,jj,i,j;

    float lb=12.5; // lower bound for " C3'"
    float ub=15.0; // upper bound for " C3'"
    if     (atom_opt==" C4'") {lb=14.0;ub=16.0;}
    else if(atom_opt==" C5'") {lb=16.0;ub=18.0;}
    else if(atom_opt==" O3'") {lb=13.5;ub=16.5;}
    else if(atom_opt==" O5'") {lb=15.5;ub=18.5;}
    else if(atom_opt==" P  ") {lb=16.5;ub=21.0;}

    float dis;
    vector<bool> bp_tmp(len,false);
    vector<vector<bool> > bp(len,bp_tmp);
    bp_tmp.clear();
    for (i=0; i<len; i++)
    {
        sec[i]='.';
        for (j=i+1; j<len; j++)
        {
            if (((seq[i]=='u'||seq[i]=='t')&&(seq[j]=='a'             ))||
                ((seq[i]=='a'             )&&(seq[j]=='u'||seq[j]=='t'))||
                ((seq[i]=='g'             )&&(seq[j]=='c'||seq[j]=='u'))||
                ((seq[i]=='c'||seq[i]=='u')&&(seq[j]=='g'             )))
            {
                dis=sqrt(dist(x[i], x[j]));
                bp[j][i]=bp[i][j]=(dis>lb && dis<ub);
            }
        }
    }
    
    // From 5' to 3': A0 C0 D0 B0: A0 paired to B0, C0 paired to D0
    vector<int> A0,B0,C0,D0;
    for (i=0; i<len-2; i++)
    {
        for (j=i+3; j<len; j++)
        {
            if (!bp[i][j]) continue;
            if (i>0 && j+1<len && bp[i-1][j+1]) continue;
            if (!bp[i+1][j-1]) continue;
            sec_str(len,seq, bp, i,j,ii,jj);
            A0.push_back(i);
            B0.push_back(j);
            C0.push_back(ii);
            D0.push_back(jj);
        }
    }
    
    //int sign;
    for (i=0;i<A0.size();i++)
    {
        /*
        sign=0;
        if(C0[i]-A0[i]<=1)
        {
            for(j=0;j<A0.size();j++)
            {
                if(i==j) continue;

                if((A0[j]>=A0[i]&&A0[j]<=C0[i])||
                   (C0[j]>=A0[i]&&C0[j]<=C0[i])||
                   (D0[j]>=A0[i]&&D0[j]<=C0[i])||
                   (B0[j]>=A0[i]&&B0[j]<=C0[i])||
                   (A0[j]>=D0[i]&&A0[j]<=B0[i])||
                   (C0[j]>=D0[i]&&C0[j]<=B0[i])||
                   (D0[j]>=D0[i]&&D0[j]<=B0[i])||
                   (B0[j]>=D0[i]&&B0[j]<=B0[i]))
                {
                    sign=-1;
                    break;
                }
            }
        }
        if(sign!=0) continue;
        */

        for (j=0;;j++)
        {
            if(A0[i]+j>C0[i]) break;
            sec[A0[i]+j]='<';
            sec[D0[i]+j]='>';
        }
    }
    sec[len]=0;

    /* clean up */
    A0.clear();
    B0.clear();
    C0.clear();
    D0.clear();
    bp.clear();
}

//get initial alignment from secondary structure alignment
//input: x, y, xlen, ylen
//output: y2x stores the best alignment: e.g., 
//y2x[j]=i means:
//the jth element in y is aligned to the ith element in x if i>=0 
//the jth element in y is aligned to a gap in x if i==-1
void get_initial_ss(bool **path, double **val,
    const char *secx, const char *secy, int xlen, int ylen, int *y2x)
{
    double gap_open=-1.0;
    NWDP_TM(path, val, secx, secy, xlen, ylen, gap_open, y2x);
}


// get_initial5 in TMalign fortran, get_initial_local in TMalign c by yangji
//get initial alignment of local structure superposition
//input: x, y, xlen, ylen
//output: y2x stores the best alignment: e.g., 
//y2x[j]=i means:
//the jth element in y is aligned to the ith element in x if i>=0 
//the jth element in y is aligned to a gap in x if i==-1
bool get_initial5( double **r1, double **r2, double **xtm, double **ytm,
    bool **path, double **val,
    double **x, double **y, int xlen, int ylen, int *y2x,
    double d0, double d0_search, const bool fast_opt, const double D0_MIN)
{
    double GL, rmsd;
    double t[3];
    double u[3][3];

    double d01 = d0 + 1.5;
    if (d01 < D0_MIN) d01 = D0_MIN;
    double d02 = d01*d01;

    double GLmax = 0;
    int aL = getmin(xlen, ylen);
    int *invmap = new int[ylen + 1];

    // jump on sequence1-------------->
    int n_jump1 = 0;
    if (xlen > 250)
        n_jump1 = 45;
    else if (xlen > 200)
        n_jump1 = 35;
    else if (xlen > 150)
        n_jump1 = 25;
    else
        n_jump1 = 15;
    if (n_jump1 > (xlen / 3))
        n_jump1 = xlen / 3;

    // jump on sequence2-------------->
    int n_jump2 = 0;
    if (ylen > 250)
        n_jump2 = 45;
    else if (ylen > 200)
        n_jump2 = 35;
    else if (ylen > 150)
        n_jump2 = 25;
    else
        n_jump2 = 15;
    if (n_jump2 > (ylen / 3))
        n_jump2 = ylen / 3;

    // fragment to superimpose-------------->
    int n_frag[2] = { 20, 100 };
    if (n_frag[0] > (aL / 3))
        n_frag[0] = aL / 3;
    if (n_frag[1] > (aL / 2))
        n_frag[1] = aL / 2;

    // start superimpose search-------------->
    if (fast_opt)
    {
        n_jump1*=5;
        n_jump2*=5;
    }
    bool flag = false;
    for (int i_frag = 0; i_frag < 2; i_frag++)
    {
        int m1 = xlen - n_frag[i_frag] + 1;
        int m2 = ylen - n_frag[i_frag] + 1;

        for (int i = 0; i<m1; i = i + n_jump1) //index starts from 0, different from FORTRAN
        {
            for (int j = 0; j<m2; j = j + n_jump2)
            {
                for (int k = 0; k<n_frag[i_frag]; k++) //fragment in y
                {
                    r1[k][0] = x[k + i][0];
                    r1[k][1] = x[k + i][1];
                    r1[k][2] = x[k + i][2];

                    r2[k][0] = y[k + j][0];
                    r2[k][1] = y[k + j][1];
                    r2[k][2] = y[k + j][2];
                }

                // superpose the two structures and rotate it
                Kabsch(r1, r2, n_frag[i_frag], 1, &rmsd, t, u);

                double gap_open = 0.0;
                NWDP_TM(path, val, x, y, xlen, ylen,
                    t, u, d02, gap_open, invmap);
                GL = get_score_fast(r1, r2, xtm, ytm, x, y, xlen, ylen,
                    invmap, d0, d0_search, t, u);
                if (GL>GLmax)
                {
                    GLmax = GL;
                    for (int ii = 0; ii<ylen; ii++) y2x[ii] = invmap[ii];
                    flag = true;
                }
            }
        }
    }

    delete[] invmap;
    return flag;
}

void score_matrix_rmsd_sec( double **r1, double **r2, double **score,
    const char *secx, const char *secy, double **x, double **y,
    int xlen, int ylen, int *y2x, const double D0_MIN, double d0)
{
    double t[3], u[3][3];
    double rmsd, dij;
    double d01=d0+1.5;
    if(d01 < D0_MIN) d01=D0_MIN;
    double d02=d01*d01;

    double xx[3];
    int i, k=0;
    for(int j=0; j<ylen; j++)
    {
        i=y2x[j];
        if(i>=0)
        {
            r1[k][0]=x[i][0];  
            r1[k][1]=x[i][1]; 
            r1[k][2]=x[i][2];   
            
            r2[k][0]=y[j][0];  
            r2[k][1]=y[j][1]; 
            r2[k][2]=y[j][2];
            
            k++;
        }
    }
    Kabsch(r1, r2, k, 1, &rmsd, t, u);

    
    for(int ii=0; ii<xlen; ii++)
    {        
        transform(t, u, &x[ii][0], xx);
        for(int jj=0; jj<ylen; jj++)
        {
            dij=dist(xx, &y[jj][0]); 
            if (secx[ii]==secy[jj])
                score[ii+1][jj+1] = 1.0/(1+dij/d02) + 0.5;
            else
                score[ii+1][jj+1] = 1.0/(1+dij/d02);
        }
    }
}


//get initial alignment from secondary structure and previous alignments
//input: x, y, xlen, ylen
//output: y2x stores the best alignment: e.g., 
//y2x[j]=i means:
//the jth element in y is aligned to the ith element in x if i>=0 
//the jth element in y is aligned to a gap in x if i==-1
void get_initial_ssplus(double **r1, double **r2, double **score, bool **path,
    double **val, const char *secx, const char *secy, double **x, double **y,
    int xlen, int ylen, int *y2x0, int *y2x, const double D0_MIN, double d0)
{
    //create score matrix for DP
    score_matrix_rmsd_sec(r1, r2, score, secx, secy, x, y, xlen, ylen,
        y2x0, D0_MIN,d0);
    
    double gap_open=-1.0;
    NWDP_TM(score, path, val, xlen, ylen, gap_open, y2x);
}


void find_max_frag(double **x, int len, int *start_max,
    int *end_max, double dcu0, const bool fast_opt)
{
    int r_min, fra_min=4;           //minimum fragment for search
    if (fast_opt) fra_min=8;
    int start;
    int Lfr_max=0;

    r_min= (int) (len*1.0/3.0); //minimum fragment, in case too small protein
    if(r_min > fra_min) r_min=fra_min;
    
    int inc=0;
    double dcu0_cut=dcu0*dcu0;;
    double dcu_cut=dcu0_cut;

    while(Lfr_max < r_min)
    {        
        Lfr_max=0;            
        int j=1;    //number of residues at nf-fragment
        start=0;
        for(int i=1; i<len; i++)
        {
            if(dist(x[i-1], x[i]) < dcu_cut)
            {
                j++;

                if(i==(len-1))
                {
                    if(j > Lfr_max) 
                    {
                        Lfr_max=j;
                        *start_max=start;
                        *end_max=i;                        
                    }
                    j=1;
                }
            }
            else
            {
                if(j>Lfr_max) 
                {
                    Lfr_max=j;
                    *start_max=start;
                    *end_max=i-1;                                        
                }

                j=1;
                start=i;
            }
        }// for i;
        
        if(Lfr_max < r_min)
        {
            inc++;
            double dinc=pow(1.1, (double) inc) * dcu0;
            dcu_cut= dinc*dinc;
        }
    }//while <;    
}

//perform fragment gapless threading to find the best initial alignment
//input: x, y, xlen, ylen
//output: y2x0 stores the best alignment: e.g., 
//y2x0[j]=i means:
//the jth element in y is aligned to the ith element in x if i>=0 
//the jth element in y is aligned to a gap in x if i==-1
double get_initial_fgt(double **r1, double **r2, double **xtm, double **ytm,
    double **x, double **y, int xlen, int ylen, 
    int *y2x, double d0, double d0_search,
    double dcu0, const bool fast_opt, double t[3], double u[3][3])
{
    int fra_min=4;           //minimum fragment for search
    if (fast_opt) fra_min=8;
    int fra_min1=fra_min-1;  //cutoff for shift, save time

    int xstart=0, ystart=0, xend=0, yend=0;

    find_max_frag(x, xlen, &xstart, &xend, dcu0, fast_opt);
    find_max_frag(y, ylen, &ystart, &yend, dcu0, fast_opt);


    int Lx = xend-xstart+1;
    int Ly = yend-ystart+1;
    int *ifr, *y2x_;
    int L_fr=getmin(Lx, Ly);
    ifr= new int[L_fr];
    y2x_= new int[ylen+1];

    //select what piece will be used. The original implement may cause 
    //asymetry, but only when xlen==ylen and Lx==Ly
    //if L1=Lfr1 and L2=Lfr2 (normal proteins), it will be the same as initial1

    if(Lx<Ly || (Lx==Ly && xlen<ylen))
    {        
        for(int i=0; i<L_fr; i++) ifr[i]=xstart+i;
    }
    else if(Lx>Ly || (Lx==Ly && xlen>ylen))
    {        
        for(int i=0; i<L_fr; i++) ifr[i]=ystart+i;
    }
    else // solve asymetric for 1x5gA vs 2q7nA5
    {
        /* In this case, L0==xlen==ylen; L_fr==Lx==Ly */
        int L0=xlen;
        double tmscore, tmscore_max=-1;
        int i, j, k;
        int n1, n2;
        int min_len;
        int min_ali;

        /* part 1, normalized by xlen */
        for(i=0; i<L_fr; i++) ifr[i]=xstart+i;

        if(L_fr==L0)
        {
            n1= (int)(L0*0.1); //my index starts from 0
            n2= (int)(L0*0.89);
            j=0;
            for(i=n1; i<= n2; i++)
            {
                ifr[j]=ifr[i];
                j++;
            }
            L_fr=j;
        }

        int L1=L_fr;
        min_len=getmin(L1, ylen);    
        min_ali= (int) (min_len/2.5); //minimum size of considered fragment 
        if(min_ali<=fra_min1)  min_ali=fra_min1;    
        n1 = -ylen+min_ali; 
        n2 = L1-min_ali;

        for(k=n1; k<=n2; k+=(fast_opt)?3:1)
        {
            //get the map
            for(j=0; j<ylen; j++)
            {
                i=j+k;
                if(i>=0 && i<L1) y2x_[j]=ifr[i];
                else             y2x_[j]=-1;
            }

            //evaluate the map quickly in three iterations
            tmscore=get_score_fast(r1, r2, xtm, ytm, x, y, xlen, ylen, y2x_,
                d0, d0_search, t, u);

            if(tmscore>=tmscore_max)
            {
                tmscore_max=tmscore;
                for(j=0; j<ylen; j++) y2x[j]=y2x_[j];
            }
        }

        /* part 2, normalized by ylen */
        L_fr=Ly;
        for(i=0; i<L_fr; i++) ifr[i]=ystart+i;

        if (L_fr==L0)
        {
            n1= (int)(L0*0.1); //my index starts from 0
            n2= (int)(L0*0.89);

            j=0;
            for(i=n1; i<= n2; i++)
            {
                ifr[j]=ifr[i];
                j++;
            }
            L_fr=j;
        }

        int L2=L_fr;
        min_len=getmin(xlen, L2);    
        min_ali= (int) (min_len/2.5); //minimum size of considered fragment 
        if(min_ali<=fra_min1)  min_ali=fra_min1;    
        n1 = -L2+min_ali; 
        n2 = xlen-min_ali;

        for(k=n1; k<=n2; k++)
        {
            //get the map
            for(j=0; j<ylen; j++) y2x_[j]=-1;

            for(j=0; j<L2; j++)
            {
                i=j+k;
                if(i>=0 && i<xlen) y2x_[ifr[j]]=i;
            }
        
            //evaluate the map quickly in three iterations
            tmscore=get_score_fast(r1, r2, xtm, ytm,
                x, y, xlen, ylen, y2x_, d0,d0_search, t, u);
            if(tmscore>=tmscore_max)
            {
                tmscore_max=tmscore;
                for(j=0; j<ylen; j++) y2x[j]=y2x_[j];
            }
        }

        delete [] ifr;
        delete [] y2x_;
        return tmscore_max;
    }

    
    int L0=getmin(xlen, ylen); //non-redundant to get_initial1
    if(L_fr==L0)
    {
        int n1= (int)(L0*0.1); //my index starts from 0
        int n2= (int)(L0*0.89);

        int j=0;
        for(int i=n1; i<= n2; i++)
        {
            ifr[j]=ifr[i];
            j++;
        }
        L_fr=j;
    }


    //gapless threading for the extracted fragment
    double tmscore, tmscore_max=-1;

    if(Lx<Ly || (Lx==Ly && xlen<=ylen))
    {
        int L1=L_fr;
        int min_len=getmin(L1, ylen);    
        int min_ali= (int) (min_len/2.5);              //minimum size of considered fragment 
        if(min_ali<=fra_min1)  min_ali=fra_min1;    
        int n1, n2;
        n1 = -ylen+min_ali; 
        n2 = L1-min_ali;

        int i, j, k;
        for(k=n1; k<=n2; k+=(fast_opt)?3:1)
        {
            //get the map
            for(j=0; j<ylen; j++)
            {
                i=j+k;
                if(i>=0 && i<L1) y2x_[j]=ifr[i];
                else             y2x_[j]=-1;
            }

            //evaluate the map quickly in three iterations
            tmscore=get_score_fast(r1, r2, xtm, ytm, x, y, xlen, ylen, y2x_,
                d0, d0_search, t, u);

            if(tmscore>=tmscore_max)
            {
                tmscore_max=tmscore;
                for(j=0; j<ylen; j++) y2x[j]=y2x_[j];
            }
        }
    }
    else
    {
        int L2=L_fr;
        int min_len=getmin(xlen, L2);    
        int min_ali= (int) (min_len/2.5);              //minimum size of considered fragment 
        if(min_ali<=fra_min1)  min_ali=fra_min1;    
        int n1, n2;
        n1 = -L2+min_ali; 
        n2 = xlen-min_ali;

        int i, j, k;    

        for(k=n1; k<=n2; k++)
        {
            //get the map
            for(j=0; j<ylen; j++) y2x_[j]=-1;

            for(j=0; j<L2; j++)
            {
                i=j+k;
                if(i>=0 && i<xlen) y2x_[ifr[j]]=i;
            }
        
            //evaluate the map quickly in three iterations
            tmscore=get_score_fast(r1, r2, xtm, ytm,
                x, y, xlen, ylen, y2x_, d0,d0_search, t, u);
            if(tmscore>=tmscore_max)
            {
                tmscore_max=tmscore;
                for(j=0; j<ylen; j++) y2x[j]=y2x_[j];
            }
        }
    }    


    delete [] ifr;
    delete [] y2x_;
    return tmscore_max;
}

//heuristic run of dynamic programing iteratively to find the best alignment
//input: initial rotation matrix t, u
//       vectors x and y, d0
//output: best alignment that maximizes the TMscore, will be stored in invmap
double DP_iter(double **r1, double **r2, double **xtm, double **ytm,
    double **xt, bool **path, double **val, double **x, double **y,
    int xlen, int ylen, double t[3], double u[3][3], int invmap0[],
    int g1, int g2, int iteration_max, double local_d0_search,
    double D0_MIN, double Lnorm, double d0, double score_d8)
{
    double gap_open[2]={-0.6, 0};
    double rmsd; 
    int *invmap=new int[ylen+1];
    
    int iteration, i, j, k;
    double tmscore, tmscore_max, tmscore_old=0;    
    int score_sum_method=8, simplify_step=40;
    tmscore_max=-1;

    //double d01=d0+1.5;
    double d02=d0*d0;
    for(int g=g1; g<g2; g++)
    {
        for(iteration=0; iteration<iteration_max; iteration++)
        {           
            NWDP_TM(path, val, x, y, xlen, ylen,
                t, u, d02, gap_open[g], invmap);
            
            k=0;
            for(j=0; j<ylen; j++) 
            {
                i=invmap[j];

                if(i>=0) //aligned
                {
                    xtm[k][0]=x[i][0];
                    xtm[k][1]=x[i][1];
                    xtm[k][2]=x[i][2];
                    
                    ytm[k][0]=y[j][0];
                    ytm[k][1]=y[j][1];
                    ytm[k][2]=y[j][2];
                    k++;
                }
            }

            tmscore = TMscore8_search(r1, r2, xtm, ytm, xt, k, t, u,
                simplify_step, score_sum_method, &rmsd, local_d0_search,
                Lnorm, score_d8, d0);

           
            if(tmscore>tmscore_max)
            {
                tmscore_max=tmscore;
                for(i=0; i<ylen; i++) invmap0[i]=invmap[i];
            }
    
            if(iteration>0)
            {
                if(fabs(tmscore_old-tmscore)<0.000001) break;       
            }
            tmscore_old=tmscore;
        }// for iteration           
        
    }//for gapopen
    
    
    delete []invmap;
    return tmscore_max;
}


void output_superpose(const string filename, const char *fname_super,
    double t[3], double u[3][3], const int ter_opt, const int mirror_opt)
{
    int compress_type=0; // uncompressed file
    ifstream fin;
    redi::ipstream fin_gz; // if file is compressed
    if (filename.size()>=3 && 
        filename.substr(filename.size()-3,3)==".gz")
    {
        fin_gz.open("zcat "+filename);
        compress_type=1;
    }
    else if (filename.size()>=4 && 
        filename.substr(filename.size()-4,4)==".bz2")
    {
        fin_gz.open("bzcat "+filename);
        compress_type=2;
    }
    else fin.open(filename.c_str());

    stringstream buf;
    string line;
    double x[3];  // before transform
    double x1[3]; // after transform

    /* for PDBx/mmCIF only */
    map<string,int> _atom_site;
    int atom_site_pos;
    vector<string> line_vec;

    while (compress_type?fin_gz.good():fin.good())
    {
        if (compress_type) getline(fin_gz, line);
        else               getline(fin, line);
        if (line.compare(0, 6, "ATOM  ")==0 || 
            line.compare(0, 6, "HETATM")==0) // PDB format
        {
            x[0]=atof(line.substr(30,8).c_str());
            x[1]=atof(line.substr(38,8).c_str());
            x[2]=atof(line.substr(46,8).c_str());
            if (mirror_opt) x[2]=-x[2];
            transform(t, u, x, x1);
            buf<<line.substr(0,30)<<setiosflags(ios::fixed)
                <<setprecision(3)
                <<setw(8)<<x1[0] <<setw(8)<<x1[1] <<setw(8)<<x1[2]
                <<line.substr(54)<<'\n';
        }
        else if (line.compare(0,5,"loop_")==0) // PDBx/mmCIF
        {
            buf<<line<<'\n';
            while(1)
            {
                if (compress_type) 
                {
                    if (fin_gz.good()) getline(fin_gz, line);
                    else PrintErrorAndQuit("ERROR! Unexpected end of "+filename);
                }
                else
                {
                    if (fin.good()) getline(fin, line);
                    else PrintErrorAndQuit("ERROR! Unexpected end of "+filename);
                }
                if (line.size()) break;
            }
            buf<<line<<'\n';
            if (line.compare(0,11,"_atom_site.")) continue;
            _atom_site.clear();
            atom_site_pos=0;
            _atom_site[line.substr(11,line.size()-12)]=atom_site_pos;
            while(1)
            {
                while(1)
                {
                    if (compress_type) 
                    {
                        if (fin_gz.good()) getline(fin_gz, line);
                        else PrintErrorAndQuit("ERROR! Unexpected end of "+filename);
                    }
                    else
                    {
                        if (fin.good()) getline(fin, line);
                        else PrintErrorAndQuit("ERROR! Unexpected end of "+filename);
                    }
                    if (line.size()) break;
                }
                if (line.compare(0,11,"_atom_site.")) break;
                _atom_site[line.substr(11,line.size()-12)]=++atom_site_pos;
                buf<<line<<'\n';
            }

            if (_atom_site.count("group_PDB")*
                _atom_site.count("Cartn_x")*
                _atom_site.count("Cartn_y")*
                _atom_site.count("Cartn_z")==0)
            {
                buf<<line<<'\n';
                cerr<<"Warning! Missing one of the following _atom_site data items: group_PDB, Cartn_x, Cartn_y, Cartn_z"<<endl;
                continue;
            }

            while(1)
            {
                line_vec.clear();
                split(line,line_vec);
                if (line_vec[_atom_site["group_PDB"]]!="ATOM" &&
                    line_vec[_atom_site["group_PDB"]]!="HETATM") break;

                x[0]=atof(line_vec[_atom_site["Cartn_x"]].c_str());
                x[1]=atof(line_vec[_atom_site["Cartn_y"]].c_str());
                x[2]=atof(line_vec[_atom_site["Cartn_z"]].c_str());
                if (mirror_opt) x[2]=-x[2];
                transform(t, u, x, x1);

                for (atom_site_pos=0; atom_site_pos<_atom_site.size(); atom_site_pos++)
                {
                    if (atom_site_pos==_atom_site["Cartn_x"])
                        buf<<setiosflags(ios::fixed)<<setprecision(3)
                           <<setw(8)<<x1[0]<<' ';
                    else if (atom_site_pos==_atom_site["Cartn_y"])
                        buf<<setiosflags(ios::fixed)<<setprecision(3)
                           <<setw(8)<<x1[1]<<' ';
                    else if (atom_site_pos==_atom_site["Cartn_z"])
                        buf<<setiosflags(ios::fixed)<<setprecision(3)
                           <<setw(8)<<x1[2]<<' ';
                    else buf<<line_vec[atom_site_pos]<<' ';
                }
                buf<<'\n';

                if (compress_type && fin_gz.good()) getline(fin_gz, line);
                else if (!compress_type && fin.good()) getline(fin, line);
                else break;
            }
            if (compress_type?fin_gz.good():fin.good()) buf<<line<<'\n';
        }
        else if (line.size())
        {
            buf<<line<<'\n';
            if (ter_opt>=1 && line.compare(0,3,"END")==0) break;
        }
    }
    if (compress_type) fin_gz.close();
    else               fin.close();

    ofstream fp(fname_super);
    fp<<buf.str();
    fp.close();
    buf.str(string()); // clear stream
}

void output_pymol(const string xname, const string yname,
    const string fname_super, double t[3], double u[3][3], const int ter_opt, 
    const int mm_opt, const int split_opt, const int mirror_opt,
    const char *seqM, const char *seqxA, const char *seqyA,
    const vector<string>&resi_vec1, const vector<string>&resi_vec2,
    const string chainID1, const string chainID2)
{
    int compress_type=0; // uncompressed file
    ifstream fin;
    redi::ipstream fin_gz; // if file is compressed
    if (xname.size()>=3 && 
        xname.substr(xname.size()-3,3)==".gz")
    {
        fin_gz.open("zcat "+xname);
        compress_type=1;
    }
    else if (xname.size()>=4 && 
        xname.substr(xname.size()-4,4)==".bz2")
    {
        fin_gz.open("bzcat "+xname);
        compress_type=2;
    }
    else fin.open(xname.c_str());

    stringstream buf;
    stringstream buf_pymol;
    string line;
    double x[3];  // before transform
    double x1[3]; // after transform

    /* for PDBx/mmCIF only */
    map<string,int> _atom_site;
    size_t atom_site_pos;
    vector<string> line_vec;
    int infmt=-1; // 0 - PDB, 3 - PDBx/mmCIF

    while (compress_type?fin_gz.good():fin.good())
    {
        if (compress_type) getline(fin_gz, line);
        else               getline(fin, line);
        if (line.compare(0, 6, "ATOM  ")==0 || 
            line.compare(0, 6, "HETATM")==0) // PDB format
        {
            infmt=0;
            x[0]=atof(line.substr(30,8).c_str());
            x[1]=atof(line.substr(38,8).c_str());
            x[2]=atof(line.substr(46,8).c_str());
            if (mirror_opt) x[2]=-x[2];
            transform(t, u, x, x1);
            buf<<line.substr(0,30)<<setiosflags(ios::fixed)
                <<setprecision(3)
                <<setw(8)<<x1[0] <<setw(8)<<x1[1] <<setw(8)<<x1[2]
                <<line.substr(54)<<'\n';
        }
        else if (line.compare(0,5,"loop_")==0) // PDBx/mmCIF
        {
            infmt=3;
            buf<<line<<'\n';
            while(1)
            {
                if (compress_type) 
                {
                    if (fin_gz.good()) getline(fin_gz, line);
                    else PrintErrorAndQuit("ERROR! Unexpected end of "+xname);
                }
                else
                {
                    if (fin.good()) getline(fin, line);
                    else PrintErrorAndQuit("ERROR! Unexpected end of "+xname);
                }
                if (line.size()) break;
            }
            buf<<line<<'\n';
            if (line.compare(0,11,"_atom_site.")) continue;
            _atom_site.clear();
            atom_site_pos=0;
            _atom_site[line.substr(11,line.size()-12)]=atom_site_pos;
            while(1)
            {
                while(1)
                {
                    if (compress_type) 
                    {
                        if (fin_gz.good()) getline(fin_gz, line);
                        else PrintErrorAndQuit("ERROR! Unexpected end of "+xname);
                    }
                    else
                    {
                        if (fin.good()) getline(fin, line);
                        else PrintErrorAndQuit("ERROR! Unexpected end of "+xname);
                    }
                    if (line.size()) break;
                }
                if (line.compare(0,11,"_atom_site.")) break;
                _atom_site[line.substr(11,line.size()-12)]=++atom_site_pos;
                buf<<line<<'\n';
            }

            if (_atom_site.count("group_PDB")*
                _atom_site.count("Cartn_x")*
                _atom_site.count("Cartn_y")*
                _atom_site.count("Cartn_z")==0)
            {
                buf<<line<<'\n';
                cerr<<"Warning! Missing one of the following _atom_site data items: group_PDB, Cartn_x, Cartn_y, Cartn_z"<<endl;
                continue;
            }

            while(1)
            {
                line_vec.clear();
                split(line,line_vec);
                if (line_vec[_atom_site["group_PDB"]]!="ATOM" &&
                    line_vec[_atom_site["group_PDB"]]!="HETATM") break;

                x[0]=atof(line_vec[_atom_site["Cartn_x"]].c_str());
                x[1]=atof(line_vec[_atom_site["Cartn_y"]].c_str());
                x[2]=atof(line_vec[_atom_site["Cartn_z"]].c_str());
                if (mirror_opt) x[2]=-x[2];
                transform(t, u, x, x1);

                for (atom_site_pos=0; atom_site_pos<_atom_site.size(); atom_site_pos++)
                {
                    if (atom_site_pos==_atom_site["Cartn_x"])
                        buf<<setiosflags(ios::fixed)<<setprecision(3)
                           <<setw(8)<<x1[0]<<' ';
                    else if (atom_site_pos==_atom_site["Cartn_y"])
                        buf<<setiosflags(ios::fixed)<<setprecision(3)
                           <<setw(8)<<x1[1]<<' ';
                    else if (atom_site_pos==_atom_site["Cartn_z"])
                        buf<<setiosflags(ios::fixed)<<setprecision(3)
                           <<setw(8)<<x1[2]<<' ';
                    else buf<<line_vec[atom_site_pos]<<' ';
                }
                buf<<'\n';

                if (compress_type && fin_gz.good()) getline(fin_gz, line);
                else if (!compress_type && fin.good()) getline(fin, line);
                else break;
            }
            if (compress_type?fin_gz.good():fin.good()) buf<<line<<'\n';
        }
        else if (line.size())
        {
            buf<<line<<'\n';
            if (ter_opt>=1 && line.compare(0,3,"END")==0) break;
        }
    }
    if (compress_type) fin_gz.close();
    else               fin.close();

    string fname_super_full=fname_super;
    if (infmt==0)      fname_super_full+=".pdb";
    else if (infmt==3) fname_super_full+=".cif";
    ofstream fp;
    fp.open(fname_super_full.c_str());
    fp<<buf.str();
    fp.close();
    buf.str(string()); // clear stream

    string chain1_sele;
    string chain2_sele;
    int i;
    if (!mm_opt)
    {
        if (split_opt==2 && ter_opt>=1) // align one chain from model 1
        {
            chain1_sele=" and c. "+chainID1.substr(1);
            chain2_sele=" and c. "+chainID2.substr(1);
        }
        else if (split_opt==2 && ter_opt==0) // align one chain from each model
        {
            for (i=1;i<chainID1.size();i++) if (chainID1[i]==',') break;
            chain1_sele=" and c. "+chainID1.substr(i+1);
            for (i=1;i<chainID2.size();i++) if (chainID2[i]==',') break;
            chain2_sele=" and c. "+chainID2.substr(i+1);
        }
    }

    /* extract aligned region */
    int i1=-1;
    int i2=-1;
    string resi1_sele;
    string resi2_sele;
    string resi1_bond;
    string resi2_bond;
    string prev_resi1;
    string prev_resi2;
    string curr_resi1;
    string curr_resi2;
    if (mm_opt)
    {
        ;
    }
    else
    {
        for (i=0;i<strlen(seqM);i++)
        {
            i1+=(seqxA[i]!='-' && seqxA[i]!='*');
            i2+=(seqyA[i]!='-');
            if (seqM[i]==' ' || seqxA[i]=='*') continue;
            curr_resi1=resi_vec1[i1].substr(0,4);
            curr_resi2=resi_vec2[i2].substr(0,4);
            if (resi1_sele.size()==0)
                resi1_sele =    "i. "+curr_resi1;
            else
            {
                resi1_sele+=" or i. "+curr_resi1;
                resi1_bond+="bond structure1 and i. "+prev_resi1+
                                              ", i. "+curr_resi1+"\n";
            }
            if (resi2_sele.size()==0)
                resi2_sele =    "i. "+curr_resi2;
            else
            {
                resi2_sele+=" or i. "+curr_resi2;
                resi2_bond+="bond structure2 and i. "+prev_resi2+
                                              ", i. "+curr_resi2+"\n";
            }
            prev_resi1=curr_resi1;
            prev_resi2=curr_resi2;
            //if (seqM[i]!=':') continue;
        }
        if (resi1_sele.size()) resi1_sele=" and ( "+resi1_sele+")";
        if (resi2_sele.size()) resi2_sele=" and ( "+resi2_sele+")";
    }

    /* write pymol script */
    vector<string> pml_list;
    pml_list.push_back(fname_super+"");
    pml_list.push_back(fname_super+"_atm");
    pml_list.push_back(fname_super+"_all");
    pml_list.push_back(fname_super+"_all_atm");
    pml_list.push_back(fname_super+"_all_atm_lig");

    for (int p=0;p<pml_list.size();p++)
    {
        if (mm_opt && p<=1) continue;
        buf_pymol
            <<"#!/usr/bin/env pymol\n"
            <<"cmd.load(\""<<fname_super_full<<"\", \"structure1\")\n"
            <<"cmd.load(\""<<yname<<"\", \"structure2\")\n"
            <<"hide all\n"
            <<"set all_states, "<<((ter_opt==0)?"on":"off")<<'\n';
        if (p==0) // .pml
        {
            if (chain1_sele.size()) buf_pymol
                <<"remove structure1 and not "<<chain1_sele.substr(4)<<"\n";
            if (chain2_sele.size()) buf_pymol
                <<"remove structure2 and not "<<chain2_sele.substr(4)<<"\n";
            buf_pymol
                <<"remove not n. CA and not n. C3'\n"
                <<resi1_bond
                <<resi2_bond
                <<"show stick, structure1"<<chain1_sele<<resi1_sele<<"\n"
                <<"show stick, structure2"<<chain2_sele<<resi2_sele<<"\n";
        }
        else if (p==1) // _atm.pml
        {
            buf_pymol
                <<"show cartoon, structure1"<<chain1_sele<<resi1_sele<<"\n"
                <<"show cartoon, structure2"<<chain2_sele<<resi2_sele<<"\n";
        }
        else if (p==2) // _all.pml
        {
            buf_pymol
                <<"show ribbon, structure1"<<chain1_sele<<"\n"
                <<"show ribbon, structure2"<<chain2_sele<<"\n";
        }
        else if (p==3) // _all_atm.pml
        {
            buf_pymol
                <<"show cartoon, structure1"<<chain1_sele<<"\n"
                <<"show cartoon, structure2"<<chain2_sele<<"\n";
        }
        else if (p==4) // _all_atm_lig.pml
        {
            buf_pymol
                <<"show cartoon, structure1\n"
                <<"show cartoon, structure2\n"
                <<"show stick, not polymer\n"
                <<"show sphere, not polymer\n";
        }
        buf_pymol
            <<"color blue, structure1\n"
            <<"color red, structure2\n"
            <<"set ribbon_width, 6\n"
            <<"set stick_radius, 0.3\n"
            <<"set sphere_scale, 0.25\n"
            <<"set ray_shadow, 0\n"
            <<"bg_color white\n"
            <<"set transparency=0.2\n"
            <<"zoom polymer and ((structure1"<<chain1_sele
            <<") or (structure2"<<chain2_sele<<"))\n"
            <<endl;

        fp.open((pml_list[p]+".pml").c_str());
        fp<<buf_pymol.str();
        fp.close();
        buf_pymol.str(string());
    }

    /* clean up */
    pml_list.clear();
    
    resi1_sele.clear();
    resi2_sele.clear();
    
    resi1_bond.clear();
    resi2_bond.clear();
    
    prev_resi1.clear();
    prev_resi2.clear();

    curr_resi1.clear();
    curr_resi2.clear();

    chain1_sele.clear();
    chain2_sele.clear();
}

void output_rasmol(const string xname, const string yname,
    const string fname_super, double t[3], double u[3][3], const int ter_opt,
    const int mm_opt, const int split_opt, const int mirror_opt,
    const char *seqM, const char *seqxA, const char *seqyA,
    const vector<string>&resi_vec1, const vector<string>&resi_vec2,
    const string chainID1, const string chainID2,
    const int xlen, const int ylen, const double d0A, const int n_ali8,
    const double rmsd, const double TM1, const double Liden)
{
    stringstream buf;
    stringstream buf_all;
    stringstream buf_atm;
    stringstream buf_all_atm;
    stringstream buf_all_atm_lig;
    //stringstream buf_pdb;
    stringstream buf_tm;
    string line;
    double x[3];  // before transform
    double x1[3]; // after transform
    bool after_ter; // true if passed the "TER" line in PDB
    string asym_id; // chain ID

    buf_tm<<"REMARK US-align"
        <<"\nREMARK Structure 1:"<<setw(11)<<left<<xname+chainID1<<" Size= "<<xlen
        <<"\nREMARK Structure 2:"<<setw(11)<<yname+chainID2<<right<<" Size= "<<ylen
        <<" (TM-score is normalized by "<<setw(4)<<ylen<<", d0="
        <<setiosflags(ios::fixed)<<setprecision(2)<<setw(6)<<d0A<<")"
        <<"\nREMARK Aligned length="<<setw(4)<<n_ali8<<", RMSD="
        <<setw(6)<<setiosflags(ios::fixed)<<setprecision(2)<<rmsd
        <<", TM-score="<<setw(7)<<setiosflags(ios::fixed)<<setprecision(5)<<TM1
        <<", ID="<<setw(5)<<setiosflags(ios::fixed)<<setprecision(3)
        <<((n_ali8>0)?Liden/n_ali8:0)<<endl;
    string rasmol_CA_header="load inline\nselect *A\nwireframe .45\nselect *B\nwireframe .20\nselect all\ncolor white\n";
    string rasmol_cartoon_header="load inline\nselect all\ncartoon\nselect *A\ncolor blue\nselect *B\ncolor red\nselect ligand\nwireframe 0.25\nselect solvent\nspacefill 0.25\nselect all\nexit\n"+buf_tm.str();
    if (!mm_opt) buf<<rasmol_CA_header;
    buf_all<<rasmol_CA_header;
    if (!mm_opt) buf_atm<<rasmol_cartoon_header;
    buf_all_atm<<rasmol_cartoon_header;
    buf_all_atm_lig<<rasmol_cartoon_header;

    /* selecting chains for -mol */
    string chain1_sele;
    string chain2_sele;
    int i;
    if (!mm_opt)
    {
        if (split_opt==2 && ter_opt>=1) // align one chain from model 1
        {
            chain1_sele=chainID1.substr(1);
            chain2_sele=chainID2.substr(1);
        }
        else if (split_opt==2 && ter_opt==0) // align one chain from each model
        {
            for (i=1;i<chainID1.size();i++) if (chainID1[i]==',') break;
            chain1_sele=chainID1.substr(i+1);
            for (i=1;i<chainID2.size();i++) if (chainID2[i]==',') break;
            chain2_sele=chainID2.substr(i+1);
        }
    }


    /* for PDBx/mmCIF only */
    map<string,int> _atom_site;
    int atom_site_pos;
    vector<string> line_vec;
    string atom; // 4-character atom name
    string AA;   // 3-character residue name
    string resi; // 4-character residue sequence number
    string inscode; // 1-character insertion code
    string model_index; // model index
    bool is_mmcif=false;

    /* used for CONECT record of chain1 */
    int ca_idx1=0; // all CA atoms
    int lig_idx1=0; // all atoms
    vector <int> idx_vec;

    /* used for CONECT record of chain2 */
    int ca_idx2=0; // all CA atoms
    int lig_idx2=0; // all atoms

    /* extract aligned region */
    vector<string> resi_aln1;
    vector<string> resi_aln2;
    int i1=-1;
    int i2=-1;
    if (!mm_opt)
    {
        for (i=0;i<strlen(seqM);i++)
        {
            i1+=(seqxA[i]!='-');
            i2+=(seqyA[i]!='-');
            if (seqM[i]==' ') continue;
            resi_aln1.push_back(resi_vec1[i1].substr(0,4));
            resi_aln2.push_back(resi_vec2[i2].substr(0,4));
            if (seqM[i]!=':') continue;
            buf    <<"select "<<resi_aln1.back()<<":A,"
                   <<resi_aln2.back()<<":B\ncolor red\n";
            buf_all<<"select "<<resi_aln1.back()<<":A,"
                   <<resi_aln2.back()<<":B\ncolor red\n";
        }
        buf<<"select all\nexit\n"<<buf_tm.str();
    }
    buf_all<<"select all\nexit\n"<<buf_tm.str();

    ifstream fin;
    /* read first file */
    after_ter=false;
    asym_id="";
    fin.open(xname.c_str());
    while (fin.good())
    {
        getline(fin, line);
        if (ter_opt>=3 && line.compare(0,3,"TER")==0) after_ter=true;
        if (is_mmcif==false && line.size()>=54 &&
           (line.compare(0, 6, "ATOM  ")==0 ||
            line.compare(0, 6, "HETATM")==0)) // PDB format
        {
            if (line[16]!='A' && line[16]!=' ') continue;
            x[0]=atof(line.substr(30,8).c_str());
            x[1]=atof(line.substr(38,8).c_str());
            x[2]=atof(line.substr(46,8).c_str());
            if (mirror_opt) x[2]=-x[2];
            transform(t, u, x, x1);
            //buf_pdb<<line.substr(0,30)<<setiosflags(ios::fixed)
                //<<setprecision(3)
                //<<setw(8)<<x1[0] <<setw(8)<<x1[1] <<setw(8)<<x1[2]
                //<<line.substr(54)<<'\n';

            if (after_ter && line.compare(0,6,"ATOM  ")==0) continue;
            lig_idx1++;
            buf_all_atm_lig<<line.substr(0,6)<<setw(5)<<lig_idx1
                <<line.substr(11,9)<<" A"<<line.substr(22,8)
                <<setiosflags(ios::fixed)<<setprecision(3)
                <<setw(8)<<x1[0]<<setw(8)<<x1[1] <<setw(8)<<x1[2]<<'\n';
            if (chain1_sele.size() && line[21]!=chain1_sele[0]) continue;
            if (after_ter || line.compare(0,6,"ATOM  ")) continue;
            if (ter_opt>=2)
            {
                if (ca_idx1 && asym_id.size() && asym_id!=line.substr(21,1)) 
                {
                    after_ter=true;
                    continue;
                }
                asym_id=line[21];
            }
            buf_all_atm<<"ATOM  "<<setw(5)<<lig_idx1
                <<line.substr(11,9)<<" A"<<line.substr(22,8)
                <<setiosflags(ios::fixed)<<setprecision(3)
                <<setw(8)<<x1[0]<<setw(8)<<x1[1] <<setw(8)<<x1[2]<<'\n';
            if (!mm_opt && find(resi_aln1.begin(),resi_aln1.end(),
                line.substr(22,4))!=resi_aln1.end())
            {
                buf_atm<<"ATOM  "<<setw(5)<<lig_idx1
                    <<line.substr(11,9)<<" A"<<line.substr(22,8)
                    <<setiosflags(ios::fixed)<<setprecision(3)
                    <<setw(8)<<x1[0]<<setw(8)<<x1[1] <<setw(8)<<x1[2]<<'\n';
            }
            if (line.substr(12,4)!=" CA " && line.substr(12,4)!=" C3'") continue;
            ca_idx1++;
            buf_all<<"ATOM  "<<setw(5)<<ca_idx1<<' '
                <<line.substr(12,4)<<' '<<line.substr(17,3)<<" A"<<line.substr(22,8)
                <<setiosflags(ios::fixed)<<setprecision(3)
                <<setw(8)<<x1[0]<<setw(8)<<x1[1]<<setw(8)<<x1[2]<<'\n';
            if (find(resi_aln1.begin(),resi_aln1.end(),
                line.substr(22,4))==resi_aln1.end()) continue;
            if (!mm_opt) buf<<"ATOM  "<<setw(5)<<ca_idx1<<' '
                <<line.substr(12,4)<<' '<<line.substr(17,3)<<" A"<<line.substr(22,8)
                <<setiosflags(ios::fixed)<<setprecision(3)
                <<setw(8)<<x1[0]<<setw(8)<<x1[1]<<setw(8)<<x1[2]<<'\n';
            idx_vec.push_back(ca_idx1);
        }
        else if (line.compare(0,5,"loop_")==0) // PDBx/mmCIF
        {
            while(1)
            {
                if (fin.good()) getline(fin, line);
                else PrintErrorAndQuit("ERROR! Unexpected end of "+xname);
                if (line.size()) break;
            }
            if (line.compare(0,11,"_atom_site.")) continue;
            _atom_site.clear();
            atom_site_pos=0;
            _atom_site[line.substr(11,line.size()-12)]=atom_site_pos;
            while(1)
            {
                if (fin.good()) getline(fin, line);
                else PrintErrorAndQuit("ERROR! Unexpected end of "+xname);
                if (line.size()==0) continue;
                if (line.compare(0,11,"_atom_site.")) break;
                _atom_site[line.substr(11,line.size()-12)]=++atom_site_pos;
            }

            if (is_mmcif==false)
            {
                //buf_pdb.str(string());
                is_mmcif=true;
            }

            while(1)
            {
                line_vec.clear();
                split(line,line_vec);
                if (line_vec[_atom_site["group_PDB"]]!="ATOM" &&
                    line_vec[_atom_site["group_PDB"]]!="HETATM") break;
                if (_atom_site.count("pdbx_PDB_model_num"))
                {
                    if (model_index.size() && model_index!=
                        line_vec[_atom_site["pdbx_PDB_model_num"]])
                        break;
                    model_index=line_vec[_atom_site["pdbx_PDB_model_num"]];
                }

                x[0]=atof(line_vec[_atom_site["Cartn_x"]].c_str());
                x[1]=atof(line_vec[_atom_site["Cartn_y"]].c_str());
                x[2]=atof(line_vec[_atom_site["Cartn_z"]].c_str());
                if (mirror_opt) x[2]=-x[2];
                transform(t, u, x, x1);

                if (_atom_site.count("label_alt_id")==0 || 
                    line_vec[_atom_site["label_alt_id"]]=="." ||
                    line_vec[_atom_site["label_alt_id"]]=="A")
                {
                    atom=line_vec[_atom_site["label_atom_id"]];
                    if (atom[0]=='"') atom=atom.substr(1);
                    if (atom.size() && atom[atom.size()-1]=='"')
                        atom=atom.substr(0,atom.size()-1);
                    if      (atom.size()==0) atom="    ";
                    else if (atom.size()==1) atom=" "+atom+"  ";
                    else if (atom.size()==2) atom=" "+atom+" ";
                    else if (atom.size()==3) atom=" "+atom;
                    else if (atom.size()>=5) atom=atom.substr(0,4);
            
                    AA=line_vec[_atom_site["label_comp_id"]]; // residue name
                    if      (AA.size()==1) AA="  "+AA;
                    else if (AA.size()==2) AA=" " +AA;
                    else if (AA.size()>=4) AA=AA.substr(0,3);
                
                    if (_atom_site.count("auth_seq_id"))
                        resi=line_vec[_atom_site["auth_seq_id"]];
                    else resi=line_vec[_atom_site["label_seq_id"]];
                    while (resi.size()<4) resi=' '+resi;
                    if (resi.size()>4) resi=resi.substr(0,4);
                
                    inscode=' ';
                    if (_atom_site.count("pdbx_PDB_ins_code") && 
                        line_vec[_atom_site["pdbx_PDB_ins_code"]]!="?")
                        inscode=line_vec[_atom_site["pdbx_PDB_ins_code"]][0];

                    if (_atom_site.count("auth_asym_id"))
                    {
                        if (chain1_sele.size()) after_ter
                            =line_vec[_atom_site["auth_asym_id"]]!=chain1_sele;
                        else if (ter_opt>=2 && ca_idx1 && asym_id.size() && 
                            asym_id!=line_vec[_atom_site["auth_asym_id"]])
                            after_ter=true;
                        asym_id=line_vec[_atom_site["auth_asym_id"]];
                    }
                    else if (_atom_site.count("label_asym_id"))
                    {
                        if (chain1_sele.size()) after_ter
                            =line_vec[_atom_site["label_asym_id"]]!=chain1_sele;
                        if (ter_opt>=2 && ca_idx1 && asym_id.size() && 
                            asym_id!=line_vec[_atom_site["label_asym_id"]])
                            after_ter=true;
                        asym_id=line_vec[_atom_site["label_asym_id"]];
                    }
                    //buf_pdb<<left<<setw(6)
                        //<<line_vec[_atom_site["group_PDB"]]<<right
                        //<<setw(5)<<lig_idx1%100000<<' '<<atom<<' '
                        //<<AA<<" "<<asym_id[asym_id.size()-1]
                        //<<resi<<inscode<<"   "
                        //<<setiosflags(ios::fixed)<<setprecision(3)
                        //<<setw(8)<<x1[0]
                        //<<setw(8)<<x1[1]
                        //<<setw(8)<<x1[2]<<'\n';

                    if (after_ter==false ||
                        line_vec[_atom_site["group_pdb"]]=="HETATM")
                    {
                        lig_idx1++;
                        buf_all_atm_lig<<left<<setw(6)
                            <<line_vec[_atom_site["group_PDB"]]<<right
                            <<setw(5)<<lig_idx1%100000<<' '<<atom<<' '
                            <<AA<<" A"<<resi<<inscode<<"   "
                            <<setiosflags(ios::fixed)<<setprecision(3)
                            <<setw(8)<<x1[0]
                            <<setw(8)<<x1[1]
                            <<setw(8)<<x1[2]<<'\n';
                        if (after_ter==false &&
                            line_vec[_atom_site["group_PDB"]]=="ATOM")
                        {
                            buf_all_atm<<"ATOM  "<<setw(6)
                                <<setw(5)<<lig_idx1%100000<<' '<<atom<<' '
                                <<AA<<" A"<<resi<<inscode<<"   "
                                <<setiosflags(ios::fixed)<<setprecision(3)
                                <<setw(8)<<x1[0]
                                <<setw(8)<<x1[1]
                                <<setw(8)<<x1[2]<<'\n';
                            if (!mm_opt && find(resi_aln1.begin(),
                                resi_aln1.end(),resi)!=resi_aln1.end())
                            {
                                buf_atm<<"ATOM  "<<setw(6)
                                    <<setw(5)<<lig_idx1%100000<<' '
                                    <<atom<<' '<<AA<<" A"<<resi<<inscode<<"   "
                                    <<setiosflags(ios::fixed)<<setprecision(3)
                                    <<setw(8)<<x1[0]
                                    <<setw(8)<<x1[1]
                                    <<setw(8)<<x1[2]<<'\n';
                            }
                            if (atom==" CA " || atom==" C3'")
                            {
                                ca_idx1++;
            //mm_opt, split_opt, mirror_opt, chainID1,chainID2);
                                buf_all<<"ATOM  "<<setw(6)
                                    <<setw(5)<<ca_idx1%100000<<' '<<atom<<' '
                                    <<AA<<" A"<<resi<<inscode<<"   "
                                    <<setiosflags(ios::fixed)<<setprecision(3)
                                    <<setw(8)<<x1[0]
                                    <<setw(8)<<x1[1]
                                    <<setw(8)<<x1[2]<<'\n';
                                if (!mm_opt && find(resi_aln1.begin(),
                                    resi_aln1.end(),resi)!=resi_aln1.end())
                                {
                                    buf<<"ATOM  "<<setw(6)
                                    <<setw(5)<<ca_idx1%100000<<' '<<atom<<' '
                                    <<AA<<" A"<<resi<<inscode<<"   "
                                    <<setiosflags(ios::fixed)<<setprecision(3)
                                    <<setw(8)<<x1[0]
                                    <<setw(8)<<x1[1]
                                    <<setw(8)<<x1[2]<<'\n';
                                    idx_vec.push_back(ca_idx1);
                                }
                            }
                        }
                    }
                }

                while(1)
                {
                    if (fin.good()) getline(fin, line);
                    else break;
                    if (line.size()) break;
                }
            }
        }
        else if (line.size() && is_mmcif==false)
        {
            //buf_pdb<<line<<'\n';
            if (ter_opt>=1 && line.compare(0,3,"END")==0) break;
        }
    }
    fin.close();
    if (!mm_opt) buf<<"TER\n";
    buf_all<<"TER\n";
    if (!mm_opt) buf_atm<<"TER\n";
    buf_all_atm<<"TER\n";
    buf_all_atm_lig<<"TER\n";
    for (i=1;i<ca_idx1;i++) buf_all<<"CONECT"
        <<setw(5)<<i%100000<<setw(5)<<(i+1)%100000<<'\n';
    if (!mm_opt) for (i=1;i<idx_vec.size();i++) buf<<"CONECT"
        <<setw(5)<<idx_vec[i-1]%100000<<setw(5)<<idx_vec[i]%100000<<'\n';
    idx_vec.clear();

    /* read second file */
    after_ter=false;
    asym_id="";
    fin.open(yname.c_str());
    while (fin.good())
    {
        getline(fin, line);
        if (ter_opt>=3 && line.compare(0,3,"TER")==0) after_ter=true;
        if (line.size()>=54 && (line.compare(0, 6, "ATOM  ")==0 ||
            line.compare(0, 6, "HETATM")==0)) // PDB format
        {
            if (line[16]!='A' && line[16]!=' ') continue;
            if (after_ter && line.compare(0,6,"ATOM  ")==0) continue;
            lig_idx2++;
            buf_all_atm_lig<<line.substr(0,6)<<setw(5)<<lig_idx1+lig_idx2
                <<line.substr(11,9)<<" B"<<line.substr(22,32)<<'\n';
            if (chain1_sele.size() && line[21]!=chain1_sele[0]) continue;
            if (after_ter || line.compare(0,6,"ATOM  ")) continue;
            if (ter_opt>=2)
            {
                if (ca_idx2 && asym_id.size() && asym_id!=line.substr(21,1))
                {
                    after_ter=true;
                    continue;
                }
                asym_id=line[21];
            }
            buf_all_atm<<"ATOM  "<<setw(5)<<lig_idx1+lig_idx2
                <<line.substr(11,9)<<" B"<<line.substr(22,32)<<'\n';
            if (!mm_opt && find(resi_aln2.begin(),resi_aln2.end(),
                line.substr(22,4))!=resi_aln2.end())
            {
                buf_atm<<"ATOM  "<<setw(5)<<lig_idx1+lig_idx2
                    <<line.substr(11,9)<<" B"<<line.substr(22,32)<<'\n';
            }
            if (line.substr(12,4)!=" CA " && line.substr(12,4)!=" C3'") continue;
            ca_idx2++;
            buf_all<<"ATOM  "<<setw(5)<<ca_idx1+ca_idx2<<' '<<line.substr(12,4)
                <<' '<<line.substr(17,3)<<" B"<<line.substr(22,32)<<'\n';
            if (find(resi_aln2.begin(),resi_aln2.end(),line.substr(22,4)
                )==resi_aln2.end()) continue;
            if (!mm_opt) buf<<"ATOM  "<<setw(5)<<ca_idx1+ca_idx2<<' '
                <<line.substr(12,4)<<' '<<line.substr(17,3)<<" B"
                <<line.substr(22,32)<<'\n';
            idx_vec.push_back(ca_idx1+ca_idx2);
        }
        else if (line.compare(0,5,"loop_")==0) // PDBx/mmCIF
        {
            while(1)
            {
                if (fin.good()) getline(fin, line);
                else PrintErrorAndQuit("ERROR! Unexpected end of "+yname);
                if (line.size()) break;
            }
            if (line.compare(0,11,"_atom_site.")) continue;
            _atom_site.clear();
            atom_site_pos=0;
            _atom_site[line.substr(11,line.size()-12)]=atom_site_pos;
            while(1)
            {
                if (fin.good()) getline(fin, line);
                else PrintErrorAndQuit("ERROR! Unexpected end of "+yname);
                if (line.size()==0) continue;
                if (line.compare(0,11,"_atom_site.")) break;
                _atom_site[line.substr(11,line.size()-12)]=++atom_site_pos;
            }

            while(1)
            {
                line_vec.clear();
                split(line,line_vec);
                if (line_vec[_atom_site["group_PDB"]]!="ATOM" &&
                    line_vec[_atom_site["group_PDB"]]!="HETATM") break;
                if (_atom_site.count("pdbx_PDB_model_num"))
                {
                    if (model_index.size() && model_index!=
                        line_vec[_atom_site["pdbx_PDB_model_num"]])
                        break;
                    model_index=line_vec[_atom_site["pdbx_PDB_model_num"]];
                }

                if (_atom_site.count("label_alt_id")==0 || 
                    line_vec[_atom_site["label_alt_id"]]=="." ||
                    line_vec[_atom_site["label_alt_id"]]=="A")
                {
                    atom=line_vec[_atom_site["label_atom_id"]];
                    if (atom[0]=='"') atom=atom.substr(1);
                    if (atom.size() && atom[atom.size()-1]=='"')
                        atom=atom.substr(0,atom.size()-1);
                    if      (atom.size()==0) atom="    ";
                    else if (atom.size()==1) atom=" "+atom+"  ";
                    else if (atom.size()==2) atom=" "+atom+" ";
                    else if (atom.size()==3) atom=" "+atom;
                    else if (atom.size()>=5) atom=atom.substr(0,4);
            
                    AA=line_vec[_atom_site["label_comp_id"]]; // residue name
                    if      (AA.size()==1) AA="  "+AA;
                    else if (AA.size()==2) AA=" " +AA;
                    else if (AA.size()>=4) AA=AA.substr(0,3);
                
                    if (_atom_site.count("auth_seq_id"))
                        resi=line_vec[_atom_site["auth_seq_id"]];
                    else resi=line_vec[_atom_site["label_seq_id"]];
                    while (resi.size()<4) resi=' '+resi;
                    if (resi.size()>4) resi=resi.substr(0,4);
                
                    inscode=' ';
                    if (_atom_site.count("pdbx_PDB_ins_code") && 
                        line_vec[_atom_site["pdbx_PDB_ins_code"]]!="?")
                        inscode=line_vec[_atom_site["pdbx_PDB_ins_code"]][0];
                    
                    if (_atom_site.count("auth_asym_id"))
                    {
                        if (chain2_sele.size()) after_ter
                            =line_vec[_atom_site["auth_asym_id"]]!=chain2_sele;
                        if (ter_opt>=2 && ca_idx2 && asym_id.size() && 
                            asym_id!=line_vec[_atom_site["auth_asym_id"]])
                            after_ter=true;
                        asym_id=line_vec[_atom_site["auth_asym_id"]];
                    }
                    else if (_atom_site.count("label_asym_id"))
                    {
                        if (chain2_sele.size()) after_ter
                            =line_vec[_atom_site["label_asym_id"]]!=chain2_sele;
                        if (ter_opt>=2 && ca_idx2 && asym_id.size() && 
                            asym_id!=line_vec[_atom_site["label_asym_id"]])
                            after_ter=true;
                        asym_id=line_vec[_atom_site["label_asym_id"]];
                    }
                    if (after_ter==false || 
                        line_vec[_atom_site["group_PDB"]]=="HETATM")
                    {
                        lig_idx2++;
                        buf_all_atm_lig<<left<<setw(6)
                            <<line_vec[_atom_site["group_PDB"]]<<right
                            <<setw(5)<<(lig_idx1+lig_idx2)%100000<<' '
                            <<atom<<' '<<AA<<" B"<<resi<<inscode<<"   "
                            <<setw(8)<<line_vec[_atom_site["Cartn_x"]]
                            <<setw(8)<<line_vec[_atom_site["Cartn_y"]]
                            <<setw(8)<<line_vec[_atom_site["Cartn_z"]]
                            <<'\n';
                        if (after_ter==false &&
                            line_vec[_atom_site["group_PDB"]]=="ATOM")
                        {
                            buf_all_atm<<"ATOM  "<<setw(6)
                                <<setw(5)<<(lig_idx1+lig_idx2)%100000<<' '
                                <<atom<<' '<<AA<<" B"<<resi<<inscode<<"   "
                                <<setw(8)<<line_vec[_atom_site["Cartn_x"]]
                                <<setw(8)<<line_vec[_atom_site["Cartn_y"]]
                                <<setw(8)<<line_vec[_atom_site["Cartn_z"]]
                                <<'\n';
                            if (!mm_opt && find(resi_aln2.begin(),
                                resi_aln2.end(),resi)!=resi_aln2.end())
                            {
                                buf_atm<<"ATOM  "<<setw(6)
                                    <<setw(5)<<(lig_idx1+lig_idx2)%100000<<' '
                                    <<atom<<' '<<AA<<" B"<<resi<<inscode<<"   "
                                    <<setw(8)<<line_vec[_atom_site["Cartn_x"]]
                                    <<setw(8)<<line_vec[_atom_site["Cartn_y"]]
                                    <<setw(8)<<line_vec[_atom_site["Cartn_z"]]
                                    <<'\n';
                            }
                            if (atom==" CA " || atom==" C3'")
                            {
                                ca_idx2++;
                                buf_all<<"ATOM  "<<setw(6)
                                    <<setw(5)<<(ca_idx1+ca_idx2)%100000
                                    <<' '<<atom<<' '<<AA<<" B"<<resi<<inscode<<"   "
                                    <<setw(8)<<line_vec[_atom_site["Cartn_x"]]
                                    <<setw(8)<<line_vec[_atom_site["Cartn_y"]]
                                    <<setw(8)<<line_vec[_atom_site["Cartn_z"]]
                                    <<'\n';
                                if (!mm_opt && find(resi_aln2.begin(),
                                    resi_aln2.end(),resi)!=resi_aln2.end())
                                {
                                    buf<<"ATOM  "<<setw(6)
                                    <<setw(5)<<(ca_idx1+ca_idx2)%100000
                                    <<' '<<atom<<' '<<AA<<" B"<<resi<<inscode<<"   "
                                    <<setw(8)<<line_vec[_atom_site["Cartn_x"]]
                                    <<setw(8)<<line_vec[_atom_site["Cartn_y"]]
                                    <<setw(8)<<line_vec[_atom_site["Cartn_z"]]
                                    <<'\n';
                                    idx_vec.push_back(ca_idx1+ca_idx2);
                                }
                            }
                        }
                    }
                }

                if (fin.good()) getline(fin, line);
                else break;
            }
        }
        else if (line.size())
        {
            if (ter_opt>=1 && line.compare(0,3,"END")==0) break;
        }
    }
    fin.close();
    if (!mm_opt) buf<<"TER\n";
    buf_all<<"TER\n";
    if (!mm_opt) buf_atm<<"TER\n";
    buf_all_atm<<"TER\n";
    buf_all_atm_lig<<"TER\n";
    for (i=ca_idx1+1;i<ca_idx1+ca_idx2;i++) buf_all<<"CONECT"
        <<setw(5)<<i%100000<<setw(5)<<(i+1)%100000<<'\n';
    for (i=1;i<idx_vec.size();i++) buf<<"CONECT"
        <<setw(5)<<idx_vec[i-1]%100000<<setw(5)<<idx_vec[i]%100000<<'\n';
    idx_vec.clear();

    /* write pymol script */
    ofstream fp;
    /*
    stringstream buf_pymol;
    vector<string> pml_list;
    pml_list.push_back(fname_super+"");
    pml_list.push_back(fname_super+"_atm");
    pml_list.push_back(fname_super+"_all");
    pml_list.push_back(fname_super+"_all_atm");
    pml_list.push_back(fname_super+"_all_atm_lig");
    for (i=0;i<pml_list.size();i++)
    {
        buf_pymol<<"#!/usr/bin/env pymol\n"
            <<"load "<<pml_list[i]<<"\n"
            <<"hide all\n"
            <<((i==0 || i==2)?("show stick\n"):("show cartoon\n"))
            <<"color blue, chain A\n"
            <<"color red, chain B\n"
            <<"set ray_shadow, 0\n"
            <<"set stick_radius, 0.3\n"
            <<"set sphere_scale, 0.25\n"
            <<"show stick, not polymer\n"
            <<"show sphere, not polymer\n"
            <<"bg_color white\n"
            <<"set transparency=0.2\n"
            <<"zoom polymer\n"
            <<endl;
        fp.open((pml_list[i]+".pml").c_str());
        fp<<buf_pymol.str();
        fp.close();
        buf_pymol.str(string());
        pml_list[i].clear();
    }
    pml_list.clear();
    */
    
    /* write rasmol script */
    if (!mm_opt)
    {
        fp.open((fname_super).c_str());
        fp<<buf.str();
        fp.close();
    }
    fp.open((fname_super+"_all").c_str());
    fp<<buf_all.str();
    fp.close();
    if (!mm_opt)
    {
        fp.open((fname_super+"_atm").c_str());
        fp<<buf_atm.str();
        fp.close();
    }
    fp.open((fname_super+"_all_atm").c_str());
    fp<<buf_all_atm.str();
    fp.close();
    fp.open((fname_super+"_all_atm_lig").c_str());
    fp<<buf_all_atm_lig.str();
    fp.close();
    //fp.open((fname_super+".pdb").c_str());
    //fp<<buf_pdb.str();
    //fp.close();

    /* clear stream */
    buf.str(string());
    buf_all.str(string());
    buf_atm.str(string());
    buf_all_atm.str(string());
    buf_all_atm_lig.str(string());
    //buf_pdb.str(string());
    buf_tm.str(string());
    resi_aln1.clear();
    resi_aln2.clear();
    asym_id.clear();
    line_vec.clear();
    atom.clear();
    AA.clear();
    resi.clear();
    inscode.clear();
    model_index.clear();
}

/* extract rotation matrix based on TMscore8 */
void output_rotation_matrix(const char* fname_matrix,
    const double t[3], const double u[3][3])
{
    fstream fout;
    fout.open(fname_matrix, ios::out | ios::trunc);
    if (fout)// succeed
    {
        fout << "------ The rotation matrix to rotate Structure_1 to Structure_2 ------\n";
        char dest[1000];
        sprintf(dest, "m %18s %14s %14s %14s\n", "t[m]", "u[m][0]", "u[m][1]", "u[m][2]");
        fout << string(dest);
        for (int k = 0; k < 3; k++)
        {
            sprintf(dest, "%d %18.10f %14.10f %14.10f %14.10f\n", k, t[k], u[k][0], u[k][1], u[k][2]);
            fout << string(dest);
        }
        fout << "\nCode for rotating Structure 1 from (x,y,z) to (X,Y,Z):\n"
                "for(i=0; i<L; i++)\n"
                "{\n"
                "   X[i] = t[0] + u[0][0]*x[i] + u[0][1]*y[i] + u[0][2]*z[i];\n"
                "   Y[i] = t[1] + u[1][0]*x[i] + u[1][1]*y[i] + u[1][2]*z[i];\n"
                "   Z[i] = t[2] + u[2][0]*x[i] + u[2][1]*y[i] + u[2][2]*z[i];\n"
                "}\n";
        fout.close();
    }
    else
        cout << "Open file to output rotation matrix fail.\n";
}

//output the final results
void output_results(const string xname, const string yname,
    const string chainID1, const string chainID2,
    const int xlen, const int ylen, double t[3], double u[3][3],
    const double TM1, const double TM2,
    const double TM3, const double TM4, const double TM5,
    const double rmsd, const double d0_out, const char *seqM,
    const char *seqxA, const char *seqyA, const double Liden,
    const int n_ali8, const int L_ali, const double TM_ali,
    const double rmsd_ali, const double TM_0, const double d0_0,
    const double d0A, const double d0B, const double Lnorm_ass,
    const double d0_scale, const double d0a, const double d0u,
    const char* fname_matrix, const int outfmt_opt, const int ter_opt,
    const int mm_opt, const int split_opt, const int o_opt,
    const string fname_super, const int i_opt, const int a_opt,
    const bool u_opt, const bool d_opt, const int mirror_opt,
    const vector<string>&resi_vec1, const vector<string>&resi_vec2)
{
    if (outfmt_opt<=0)
    {
        printf("\nName of Structure_1: %s%s (to be superimposed onto Structure_2)\n",
            xname.c_str(), chainID1.c_str());
        printf("Name of Structure_2: %s%s\n", yname.c_str(), chainID2.c_str());
        printf("Length of Structure_1: %d residues\n", xlen);
        printf("Length of Structure_2: %d residues\n\n", ylen);

        if (i_opt)
            printf("User-specified initial alignment: TM/Lali/rmsd = %7.5lf, %4d, %6.3lf\n", TM_ali, L_ali, rmsd_ali);

        printf("Aligned length= %d, RMSD= %6.2f, Seq_ID=n_identical/n_aligned= %4.3f\n", n_ali8, rmsd, (n_ali8>0)?Liden/n_ali8:0);
        printf("TM-score= %6.5f (normalized by length of Structure_1: L=%d, d0=%.2f)\n", TM2, xlen, d0B);
        printf("TM-score= %6.5f (normalized by length of Structure_2: L=%d, d0=%.2f)\n", TM1, ylen, d0A);

        if (a_opt==1)
            printf("TM-score= %6.5f (if normalized by average length of two structures: L=%.1f, d0=%.2f)\n", TM3, (xlen+ylen)*0.5, d0a);
        if (u_opt)
            printf("TM-score= %6.5f (normalized by user-specified L=%.2f and d0=%.2f)\n", TM4, Lnorm_ass, d0u);
        if (d_opt)
            printf("TM-score= %6.5f (scaled by user-specified d0=%.2f, and L=%d)\n", TM5, d0_scale, ylen);
        printf("(You should use TM-score normalized by length of the reference structure)\n");
    
        //output alignment
        printf("\n(\":\" denotes residue pairs of d <%4.1f Angstrom, ", d0_out);
        printf("\".\" denotes other aligned residues)\n");
        printf("%s\n", seqxA);
        printf("%s\n", seqM);
        printf("%s\n", seqyA);
    }
    else if (outfmt_opt==1)
    {
        printf(">%s%s\tL=%d\td0=%.2f\tseqID=%.3f\tTM-score=%.5f\n",
            xname.c_str(), chainID1.c_str(), xlen, d0B, Liden/xlen, TM2);
        printf("%s\n", seqxA);
        printf(">%s%s\tL=%d\td0=%.2f\tseqID=%.3f\tTM-score=%.5f\n",
            yname.c_str(), chainID2.c_str(), ylen, d0A, Liden/ylen, TM1);
        printf("%s\n", seqyA);

        printf("# Lali=%d\tRMSD=%.2f\tseqID_ali=%.3f\n",
            n_ali8, rmsd, (n_ali8>0)?Liden/n_ali8:0);

        if (i_opt)
            printf("# User-specified initial alignment: TM=%.5lf\tLali=%4d\trmsd=%.3lf\n", TM_ali, L_ali, rmsd_ali);

        if(a_opt)
            printf("# TM-score=%.5f (normalized by average length of two structures: L=%.1f\td0=%.2f)\n", TM3, (xlen+ylen)*0.5, d0a);

        if(u_opt)
            printf("# TM-score=%.5f (normalized by user-specified L=%.2f\td0=%.2f)\n", TM4, Lnorm_ass, d0u);

        if(d_opt)
            printf("# TM-score=%.5f (scaled by user-specified d0=%.2f\tL=%d)\n", TM5, d0_scale, ylen);

        printf("$$$$\n");
    }
    else if (outfmt_opt==2)
    {
        printf("%s%s\t%s%s\t%.4f\t%.4f\t%.2f\t%4.3f\t%4.3f\t%4.3f\t%d\t%d\t%d",
            xname.c_str(), chainID1.c_str(), yname.c_str(), chainID2.c_str(),
            TM2, TM1, rmsd, Liden/xlen, Liden/ylen, (n_ali8>0)?Liden/n_ali8:0,
            xlen, ylen, n_ali8);
    }
    cout << endl;

    if (strlen(fname_matrix)) output_rotation_matrix(fname_matrix, t, u);

    if (fname_super.size())
        output_superpose(xname, fname_super.c_str(), t, u, ter_opt, mirror_opt);
    //if (o_opt==1)
        //output_pymol(xname, yname, fname_super, t, u, ter_opt,
            //mm_opt, split_opt, mirror_opt, seqM, seqxA, seqyA,
            //resi_vec1, resi_vec2, chainID1, chainID2);
    //else if (o_opt==2)
        //output_rasmol(xname, yname, fname_super, t, u, ter_opt,
            //mm_opt, split_opt, mirror_opt, seqM, seqxA, seqyA,
            //resi_vec1, resi_vec2, chainID1, chainID2,
            //xlen, ylen, d0A, n_ali8, rmsd, TM1, Liden);
}

double standard_TMscore(double **r1, double **r2, double **xtm, double **ytm,
    double **xt, double **x, double **y, int xlen, int ylen, int invmap[],
    int& L_ali, double& RMSD, double D0_MIN, double Lnorm, double d0,
    double d0_search, double score_d8, double t[3], double u[3][3],
    const int mol_type)
{
    D0_MIN = 0.5;
    Lnorm = ylen;
    if (mol_type>0) // RNA
    {
        if     (Lnorm<=11) d0=0.3; 
        else if(Lnorm>11 && Lnorm<=15) d0=0.4;
        else if(Lnorm>15 && Lnorm<=19) d0=0.5;
        else if(Lnorm>19 && Lnorm<=23) d0=0.6;
        else if(Lnorm>23 && Lnorm<30)  d0=0.7;
        else d0=(0.6*pow((Lnorm*1.0-0.5), 1.0/2)-2.5);
    }
    else
    {
        if (Lnorm > 21) d0=(1.24*pow((Lnorm*1.0-15), 1.0/3) -1.8);
        else d0 = D0_MIN;
        if (d0 < D0_MIN) d0 = D0_MIN;
    }
    double d0_input = d0;// Scaled by seq_min

    double tmscore;// collected alined residues from invmap
    int n_al = 0;
    int i;
    for (int j = 0; j<ylen; j++)
    {
        i = invmap[j];
        if (i >= 0)
        {
            xtm[n_al][0] = x[i][0];
            xtm[n_al][1] = x[i][1];
            xtm[n_al][2] = x[i][2];

            ytm[n_al][0] = y[j][0];
            ytm[n_al][1] = y[j][1];
            ytm[n_al][2] = y[j][2];

            r1[n_al][0] = x[i][0];
            r1[n_al][1] = x[i][1];
            r1[n_al][2] = x[i][2];

            r2[n_al][0] = y[j][0];
            r2[n_al][1] = y[j][1];
            r2[n_al][2] = y[j][2];

            n_al++;
        }
        else if (i != -1) PrintErrorAndQuit("Wrong map!\n");
    }
    L_ali = n_al;

    Kabsch(r1, r2, n_al, 0, &RMSD, t, u);
    RMSD = sqrt( RMSD/(1.0*n_al) );
    
    int temp_simplify_step = 1;
    int temp_score_sum_method = 0;
    d0_search = d0_input;
    double rms = 0.0;
    tmscore = TMscore8_search_standard(r1, r2, xtm, ytm, xt, n_al, t, u,
        temp_simplify_step, temp_score_sum_method, &rms, d0_input,
        score_d8, d0);
    tmscore = tmscore * n_al / (1.0*Lnorm);

    return tmscore;
}

/* copy the value of t and u into t0,u0 */
void copy_t_u(double t[3], double u[3][3], double t0[3], double u0[3][3])
{
    int i,j;
    for (i=0;i<3;i++)
    {
        t0[i]=t[i];
        for (j=0;j<3;j++) u0[i][j]=u[i][j];
    }
}

/* calculate approximate TM-score given rotation matrix */
double approx_TM(const int xlen, const int ylen, const int a_opt,
    double **xa, double **ya, double t[3], double u[3][3],
    const int invmap0[], const int mol_type)
{
    double Lnorm_0=ylen; // normalized by the second protein
    if (a_opt==-2 && xlen>ylen) Lnorm_0=xlen;      // longer
    else if (a_opt==-1 && xlen<ylen) Lnorm_0=xlen; // shorter
    else if (a_opt==1) Lnorm_0=(xlen+ylen)/2.;     // average
    
    double D0_MIN;
    double Lnorm;
    double d0;
    double d0_search;
    parameter_set4final(Lnorm_0, D0_MIN, Lnorm, d0, d0_search, mol_type);
    double TMtmp=0;
    double d;
    double xtmp[3]={0,0,0};

    for(int i=0,j=0; j<ylen; j++)
    {
        i=invmap0[j];
        if(i>=0)//aligned
        {
            transform(t, u, &xa[i][0], &xtmp[0]);
            d=sqrt(dist(&xtmp[0], &ya[j][0]));
            TMtmp+=1/(1+(d/d0)*(d/d0));
            //if (d <= score_d8) TMtmp+=1/(1+(d/d0)*(d/d0));
        }
    }
    TMtmp/=Lnorm_0;
    return TMtmp;
}

void clean_up_after_approx_TM(int *invmap0, int *invmap,
    double **score, bool **path, double **val, double **xtm, double **ytm,
    double **xt, double **r1, double **r2, const int xlen, const int minlen)
{
    delete [] invmap0;
    delete [] invmap;
    DeleteArray(&score, xlen+1);
    DeleteArray(&path, xlen+1);
    DeleteArray(&val, xlen+1);
    DeleteArray(&xtm, minlen);
    DeleteArray(&ytm, minlen);
    DeleteArray(&xt, xlen);
    DeleteArray(&r1, minlen);
    DeleteArray(&r2, minlen);
    return;
}

/* Entry function for TM-align. Return TM-score calculation status:
 * 0   - full TM-score calculation 
 * 1   - terminated due to exception
 * 2-7 - pre-terminated due to low TM-score */
int TMalign_main(double **xa, double **ya,
    const char *seqx, const char *seqy, const char *secx, const char *secy,
    double t0[3], double u0[3][3],
    double &TM1, double &TM2, double &TM3, double &TM4, double &TM5,
    double &d0_0, double &TM_0,
    double &d0A, double &d0B, double &d0u, double &d0a, double &d0_out,
    string &seqM, string &seqxA, string &seqyA,
    double &rmsd0, int &L_ali, double &Liden,
    double &TM_ali, double &rmsd_ali, int &n_ali, int &n_ali8,
    const int xlen, const int ylen,
    const vector<string> sequence, const double Lnorm_ass,
    const double d0_scale, const int i_opt, const int a_opt,
    const bool u_opt, const bool d_opt, const bool fast_opt,
    const int mol_type, const double TMcut=-1)
{
    double D0_MIN;        //for d0
    double Lnorm;         //normalization length
    double score_d8,d0,d0_search,dcu0;//for TMscore search
    double t[3], u[3][3]; //Kabsch translation vector and rotation matrix
    double **score;       // Input score table for dynamic programming
    bool   **path;        // for dynamic programming  
    double **val;         // for dynamic programming  
    double **xtm, **ytm;  // for TMscore search engine
    double **xt;          //for saving the superposed version of r_1 or xtm
    double **r1, **r2;    // for Kabsch rotation

    /***********************/
    /* allocate memory     */
    /***********************/
    int minlen = min(xlen, ylen);
    NewArray(&score, xlen+1, ylen+1);
    NewArray(&path, xlen+1, ylen+1);
    NewArray(&val, xlen+1, ylen+1);
    NewArray(&xtm, minlen, 3);
    NewArray(&ytm, minlen, 3);
    NewArray(&xt, xlen, 3);
    NewArray(&r1, minlen, 3);
    NewArray(&r2, minlen, 3);

    /***********************/
    /*    parameter set    */
    /***********************/
    parameter_set4search(xlen, ylen, D0_MIN, Lnorm, 
        score_d8, d0, d0_search, dcu0);
    int simplify_step    = 40; //for simplified search engine
    int score_sum_method = 8;  //for scoring method, whether only sum over pairs with dis<score_d8

    int i;
    int *invmap0         = new int[ylen+1];
    int *invmap          = new int[ylen+1];
    double TM, TMmax=-1;
    for(i=0; i<ylen; i++) invmap0[i]=-1;

    double ddcc=0.4;
    if (Lnorm <= 40) ddcc=0.1;   //Lnorm was setted in parameter_set4search
    double local_d0_search = d0_search;

    //************************************************//
    //    get initial alignment from user's input:    //
    //    Stick to the initial alignment              //
    //************************************************//
    bool bAlignStick = false;
    if (i_opt==3)// if input has set parameter for "-I"
    {
        // In the original code, this loop starts from 1, which is
        // incorrect. Fortran starts from 1 but C++ should starts from 0.
        for (int j = 0; j < ylen; j++)// Set aligned position to be "-1"
            invmap[j] = -1;

        int i1 = -1;// in C version, index starts from zero, not from one
        int i2 = -1;
        int L1 = sequence[0].size();
        int L2 = sequence[1].size();
        int L = min(L1, L2);// Get positions for aligned residues
        for (int kk1 = 0; kk1 < L; kk1++)
        {
            if (sequence[0][kk1] != '-') i1++;
            if (sequence[1][kk1] != '-')
            {
                i2++;
                if (i2 >= ylen || i1 >= xlen) kk1 = L;
                else if (sequence[0][kk1] != '-') invmap[i2] = i1;
            }
        }

        //--------------- 2. Align proteins from original alignment
        double prevD0_MIN = D0_MIN;// stored for later use
        int prevLnorm = Lnorm;
        double prevd0 = d0;
        TM_ali = standard_TMscore(r1, r2, xtm, ytm, xt, xa, ya, xlen, ylen,
            invmap, L_ali, rmsd_ali, D0_MIN, Lnorm, d0, d0_search, score_d8,
            t, u, mol_type);
        D0_MIN = prevD0_MIN;
        Lnorm = prevLnorm;
        d0 = prevd0;
        TM = detailed_search_standard(r1, r2, xtm, ytm, xt, xa, ya, xlen, ylen,
            invmap, t, u, 40, 8, local_d0_search, true, Lnorm, score_d8, d0);
        if (TM > TMmax)
        {
            TMmax = TM;
            for (i = 0; i<ylen; i++) invmap0[i] = invmap[i];
        }
        bAlignStick = true;
    }

    /******************************************************/
    /*    get initial alignment with gapless threading    */
    /******************************************************/
    if (!bAlignStick)
    {
        get_initial(r1, r2, xtm, ytm, xa, ya, xlen, ylen, invmap0, d0,
            d0_search, fast_opt, t, u);
        TM = detailed_search(r1, r2, xtm, ytm, xt, xa, ya, xlen, ylen, invmap0,
            t, u, simplify_step, score_sum_method, local_d0_search, Lnorm,
            score_d8, d0);
        if (TM>TMmax) TMmax = TM;
        if (TMcut>0) copy_t_u(t, u, t0, u0);
        //run dynamic programing iteratively to find the best alignment
        TM = DP_iter(r1, r2, xtm, ytm, xt, path, val, xa, ya, xlen, ylen,
             t, u, invmap, 0, 2, (fast_opt)?2:30, local_d0_search,
             D0_MIN, Lnorm, d0, score_d8);
        if (TM>TMmax)
        {
            TMmax = TM;
            for (int i = 0; i<ylen; i++) invmap0[i] = invmap[i];
            if (TMcut>0) copy_t_u(t, u, t0, u0);
        }

        if (TMcut>0) // pre-terminate if TM-score is too low
        {
            double TMtmp=approx_TM(xlen, ylen, a_opt,
                xa, ya, t0, u0, invmap0, mol_type);

            if (TMtmp<0.5*TMcut)
            {
                TM1=TM2=TM3=TM4=TM5=TMtmp;
                clean_up_after_approx_TM(invmap0, invmap, score, path, val,
                    xtm, ytm, xt, r1, r2, xlen, minlen);
                return 2;
            }
        }

        /************************************************************/
        /*    get initial alignment based on secondary structure    */
        /************************************************************/
        get_initial_ss(path, val, secx, secy, xlen, ylen, invmap);
        TM = detailed_search(r1, r2, xtm, ytm, xt, xa, ya, xlen, ylen, invmap,
            t, u, simplify_step, score_sum_method, local_d0_search, Lnorm,
            score_d8, d0);
        if (TM>TMmax)
        {
            TMmax = TM;
            for (int i = 0; i<ylen; i++) invmap0[i] = invmap[i];
            if (TMcut>0) copy_t_u(t, u, t0, u0);
        }
        if (TM > TMmax*0.2)
        {
            TM = DP_iter(r1, r2, xtm, ytm, xt, path, val, xa, ya,
                xlen, ylen, t, u, invmap, 0, 2, (fast_opt)?2:30,
                local_d0_search, D0_MIN, Lnorm, d0, score_d8);
            if (TM>TMmax)
            {
                TMmax = TM;
                for (int i = 0; i<ylen; i++) invmap0[i] = invmap[i];
                if (TMcut>0) copy_t_u(t, u, t0, u0);
            }
        }

        if (TMcut>0) // pre-terminate if TM-score is too low
        {
            double TMtmp=approx_TM(xlen, ylen, a_opt,
                xa, ya, t0, u0, invmap0, mol_type);

            if (TMtmp<0.52*TMcut)
            {
                TM1=TM2=TM3=TM4=TM5=TMtmp;
                clean_up_after_approx_TM(invmap0, invmap, score, path, val,
                    xtm, ytm, xt, r1, r2, xlen, minlen);
                return 3;
            }
        }

        /************************************************************/
        /*    get initial alignment based on local superposition    */
        /************************************************************/
        //=initial5 in original TM-align
        if (get_initial5( r1, r2, xtm, ytm, path, val, xa, ya,
            xlen, ylen, invmap, d0, d0_search, fast_opt, D0_MIN))
        {
            TM = detailed_search(r1, r2, xtm, ytm, xt, xa, ya, xlen, ylen,
                invmap, t, u, simplify_step, score_sum_method,
                local_d0_search, Lnorm, score_d8, d0);
            if (TM>TMmax)
            {
                TMmax = TM;
                for (int i = 0; i<ylen; i++) invmap0[i] = invmap[i];
                if (TMcut>0) copy_t_u(t, u, t0, u0);
            }
            if (TM > TMmax*ddcc)
            {
                TM = DP_iter(r1, r2, xtm, ytm, xt, path, val, xa, ya,
                    xlen, ylen, t, u, invmap, 0, 2, 2, local_d0_search,
                    D0_MIN, Lnorm, d0, score_d8);
                if (TM>TMmax)
                {
                    TMmax = TM;
                    for (int i = 0; i<ylen; i++) invmap0[i] = invmap[i];
                    if (TMcut>0) copy_t_u(t, u, t0, u0);
                }
            }
        }
        else
            cerr << "\n\nWarning: initial alignment from local superposition fail!\n\n" << endl;

        if (TMcut>0) // pre-terminate if TM-score is too low
        {
            double TMtmp=approx_TM(xlen, ylen, a_opt,
                xa, ya, t0, u0, invmap0, mol_type);

            if (TMtmp<0.54*TMcut)
            {
                TM1=TM2=TM3=TM4=TM5=TMtmp;
                clean_up_after_approx_TM(invmap0, invmap, score, path, val,
                    xtm, ytm, xt, r1, r2, xlen, minlen);
                return 4;
            }
        }

        /********************************************************************/
        /* get initial alignment by local superposition+secondary structure */
        /********************************************************************/
        //=initial3 in original TM-align
        get_initial_ssplus(r1, r2, score, path, val, secx, secy, xa, ya,
            xlen, ylen, invmap0, invmap, D0_MIN, d0);
        TM = detailed_search(r1, r2, xtm, ytm, xt, xa, ya, xlen, ylen, invmap,
             t, u, simplify_step, score_sum_method, local_d0_search, Lnorm,
             score_d8, d0);
        if (TM>TMmax)
        {
            TMmax = TM;
            for (i = 0; i<ylen; i++) invmap0[i] = invmap[i];
            if (TMcut>0) copy_t_u(t, u, t0, u0);
        }
        if (TM > TMmax*ddcc)
        {
            TM = DP_iter(r1, r2, xtm, ytm, xt, path, val, xa, ya,
                xlen, ylen, t, u, invmap, 0, 2, (fast_opt)?2:30,
                local_d0_search, D0_MIN, Lnorm, d0, score_d8);
            if (TM>TMmax)
            {
                TMmax = TM;
                for (i = 0; i<ylen; i++) invmap0[i] = invmap[i];
                if (TMcut>0) copy_t_u(t, u, t0, u0);
            }
        }

        if (TMcut>0) // pre-terminate if TM-score is too low
        {
            double TMtmp=approx_TM(xlen, ylen, a_opt,
                xa, ya, t0, u0, invmap0, mol_type);

            if (TMtmp<0.56*TMcut)
            {
                TM1=TM2=TM3=TM4=TM5=TMtmp;
                clean_up_after_approx_TM(invmap0, invmap, score, path, val,
                    xtm, ytm, xt, r1, r2, xlen, minlen);
                return 5;
            }
        }

        /*******************************************************************/
        /*    get initial alignment based on fragment gapless threading    */
        /*******************************************************************/
        //=initial4 in original TM-align
        get_initial_fgt(r1, r2, xtm, ytm, xa, ya, xlen, ylen,
            invmap, d0, d0_search, dcu0, fast_opt, t, u);
        TM = detailed_search(r1, r2, xtm, ytm, xt, xa, ya, xlen, ylen, invmap,
            t, u, simplify_step, score_sum_method, local_d0_search, Lnorm,
            score_d8, d0);
        if (TM>TMmax)
        {
            TMmax = TM;
            for (i = 0; i<ylen; i++) invmap0[i] = invmap[i];
            if (TMcut>0) copy_t_u(t, u, t0, u0);
        }
        if (TM > TMmax*ddcc)
        {
            TM = DP_iter(r1, r2, xtm, ytm, xt, path, val, xa, ya,
                xlen, ylen, t, u, invmap, 1, 2, 2, local_d0_search, D0_MIN,
                Lnorm, d0, score_d8);
            if (TM>TMmax)
            {
                TMmax = TM;
                for (i = 0; i<ylen; i++) invmap0[i] = invmap[i];
                if (TMcut>0) copy_t_u(t, u, t0, u0);
            }
        }

        if (TMcut>0) // pre-terminate if TM-score is too low
        {
            double TMtmp=approx_TM(xlen, ylen, a_opt,
                xa, ya, t0, u0, invmap0, mol_type);

            if (TMtmp<0.58*TMcut)
            {
                TM1=TM2=TM3=TM4=TM5=TMtmp;
                clean_up_after_approx_TM(invmap0, invmap, score, path, val,
                    xtm, ytm, xt, r1, r2, xlen, minlen);
                return 6;
            }
        }

        //************************************************//
        //    get initial alignment from user's input:    //
        //************************************************//
        if (i_opt==1)// if input has set parameter for "-i"
        {
            for (int j = 0; j < ylen; j++)// Set aligned position to be "-1"
                invmap[j] = -1;

            int i1 = -1;// in C version, index starts from zero, not from one
            int i2 = -1;
            int L1 = sequence[0].size();
            int L2 = sequence[1].size();
            int L = min(L1, L2);// Get positions for aligned residues
            for (int kk1 = 0; kk1 < L; kk1++)
            {
                if (sequence[0][kk1] != '-')
                    i1++;
                if (sequence[1][kk1] != '-')
                {
                    i2++;
                    if (i2 >= ylen || i1 >= xlen) kk1 = L;
                    else if (sequence[0][kk1] != '-') invmap[i2] = i1;
                }
            }

            //--------------- 2. Align proteins from original alignment
            double prevD0_MIN = D0_MIN;// stored for later use
            int prevLnorm = Lnorm;
            double prevd0 = d0;
            TM_ali = standard_TMscore(r1, r2, xtm, ytm, xt, xa, ya,
                xlen, ylen, invmap, L_ali, rmsd_ali, D0_MIN, Lnorm, d0,
                d0_search, score_d8, t, u, mol_type);
            D0_MIN = prevD0_MIN;
            Lnorm = prevLnorm;
            d0 = prevd0;

            TM = detailed_search_standard(r1, r2, xtm, ytm, xt, xa, ya,
                xlen, ylen, invmap, t, u, 40, 8, local_d0_search, true, Lnorm,
                score_d8, d0);
            if (TM > TMmax)
            {
                TMmax = TM;
                for (i = 0; i<ylen; i++) invmap0[i] = invmap[i];
            }
            // Different from get_initial, get_initial_ss and get_initial_ssplus
            TM = DP_iter(r1, r2, xtm, ytm, xt, path, val, xa, ya,
                xlen, ylen, t, u, invmap, 0, 2, (fast_opt)?2:30,
                local_d0_search, D0_MIN, Lnorm, d0, score_d8);
            if (TM>TMmax)
            {
                TMmax = TM;
                for (i = 0; i<ylen; i++) invmap0[i] = invmap[i];
            }
        }
    }



    //*******************************************************************//
    //    The alignment will not be changed any more in the following    //
    //*******************************************************************//
    //check if the initial alignment is generated appropriately
    bool flag=false;
    for(i=0; i<ylen; i++)
    {
        if(invmap0[i]>=0)
        {
            flag=true;
            break;
        }
    }
    if(!flag)
    {
        cout << "There is no alignment between the two proteins! "
             << "Program stop with no result!" << endl;
        TM1=TM2=TM3=TM4=TM5=0;
        return 1;
    }

    /* last TM-score pre-termination */
    if (TMcut>0)
    {
        double TMtmp=approx_TM(xlen, ylen, a_opt,
            xa, ya, t0, u0, invmap0, mol_type);

        if (TMtmp<0.6*TMcut)
        {
            TM1=TM2=TM3=TM4=TM5=TMtmp;
            clean_up_after_approx_TM(invmap0, invmap, score, path, val,
                xtm, ytm, xt, r1, r2, xlen, minlen);
            return 7;
        }
    }

    //********************************************************************//
    //    Detailed TMscore search engine --> prepare for final TMscore    //
    //********************************************************************//
    //run detailed TMscore search engine for the best alignment, and
    //extract the best rotation matrix (t, u) for the best alignment
    simplify_step=1;
    if (fast_opt) simplify_step=40;
    score_sum_method=8;
    TM = detailed_search_standard(r1, r2, xtm, ytm, xt, xa, ya, xlen, ylen,
        invmap0, t, u, simplify_step, score_sum_method, local_d0_search,
        false, Lnorm, score_d8, d0);

    //select pairs with dis<d8 for final TMscore computation and output alignment
    int k=0;
    int *m1, *m2;
    double d;
    m1=new int[xlen]; //alignd index in x
    m2=new int[ylen]; //alignd index in y
    do_rotation(xa, xt, xlen, t, u);
    k=0;
    for(int j=0; j<ylen; j++)
    {
        i=invmap0[j];
        if(i>=0)//aligned
        {
            n_ali++;
            d=sqrt(dist(&xt[i][0], &ya[j][0]));
            if (d <= score_d8 || (i_opt == 3))
            {
                m1[k]=i;
                m2[k]=j;

                xtm[k][0]=xa[i][0];
                xtm[k][1]=xa[i][1];
                xtm[k][2]=xa[i][2];

                ytm[k][0]=ya[j][0];
                ytm[k][1]=ya[j][1];
                ytm[k][2]=ya[j][2];

                r1[k][0] = xt[i][0];
                r1[k][1] = xt[i][1];
                r1[k][2] = xt[i][2];
                r2[k][0] = ya[j][0];
                r2[k][1] = ya[j][1];
                r2[k][2] = ya[j][2];

                k++;
            }
        }
    }
    n_ali8=k;

    Kabsch(r1, r2, n_ali8, 0, &rmsd0, t, u);// rmsd0 is used for final output, only recalculate rmsd0, not t & u
    rmsd0 = sqrt(rmsd0 / n_ali8);


    //****************************************//
    //              Final TMscore             //
    //    Please set parameters for output    //
    //****************************************//
    double rmsd;
    simplify_step=1;
    score_sum_method=0;
    double Lnorm_0=ylen;


    //normalized by length of structure A
    parameter_set4final(Lnorm_0, D0_MIN, Lnorm, d0, d0_search, mol_type);
    d0A=d0;
    d0_0=d0A;
    local_d0_search = d0_search;
    TM1 = TMscore8_search(r1, r2, xtm, ytm, xt, n_ali8, t0, u0, simplify_step,
        score_sum_method, &rmsd, local_d0_search, Lnorm, score_d8, d0);
    TM_0 = TM1;

    //normalized by length of structure B
    parameter_set4final(xlen+0.0, D0_MIN, Lnorm, d0, d0_search, mol_type);
    d0B=d0;
    local_d0_search = d0_search;
    TM2 = TMscore8_search(r1, r2, xtm, ytm, xt, n_ali8, t, u, simplify_step,
        score_sum_method, &rmsd, local_d0_search, Lnorm, score_d8, d0);

    double Lnorm_d0;
    if (a_opt>0)
    {
        //normalized by average length of structures A, B
        Lnorm_0=(xlen+ylen)*0.5;
        parameter_set4final(Lnorm_0, D0_MIN, Lnorm, d0, d0_search, mol_type);
        d0a=d0;
        d0_0=d0a;
        local_d0_search = d0_search;

        TM3 = TMscore8_search(r1, r2, xtm, ytm, xt, n_ali8, t0, u0,
            simplify_step, score_sum_method, &rmsd, local_d0_search, Lnorm,
            score_d8, d0);
        TM_0=TM3;
    }
    if (u_opt)
    {
        //normalized by user assigned length
        parameter_set4final(Lnorm_ass, D0_MIN, Lnorm,
            d0, d0_search, mol_type);
        d0u=d0;
        d0_0=d0u;
        Lnorm_0=Lnorm_ass;
        local_d0_search = d0_search;
        TM4 = TMscore8_search(r1, r2, xtm, ytm, xt, n_ali8, t0, u0,
            simplify_step, score_sum_method, &rmsd, local_d0_search, Lnorm,
            score_d8, d0);
        TM_0=TM4;
    }
    if (d_opt)
    {
        //scaled by user assigned d0
        parameter_set4scale(ylen, d0_scale, Lnorm, d0, d0_search);
        d0_out=d0_scale;
        d0_0=d0_scale;
        //Lnorm_0=ylen;
        Lnorm_d0=Lnorm_0;
        local_d0_search = d0_search;
        TM5 = TMscore8_search(r1, r2, xtm, ytm, xt, n_ali8, t0, u0,
            simplify_step, score_sum_method, &rmsd, local_d0_search, Lnorm,
            score_d8, d0);
        TM_0=TM5;
    }

    /* derive alignment from superposition */
    int ali_len=xlen+ylen; //maximum length of alignment
    seqxA.assign(ali_len,'-');
    seqM.assign( ali_len,' ');
    seqyA.assign(ali_len,'-');
    
    //do_rotation(xa, xt, xlen, t, u);
    do_rotation(xa, xt, xlen, t0, u0);

    int kk=0, i_old=0, j_old=0;
    d=0;
    Liden=0;
    for(int k=0; k<n_ali8; k++)
    {
        for(int i=i_old; i<m1[k]; i++)
        {
            //align x to gap
            seqxA[kk]=seqx[i];
            seqyA[kk]='-';
            seqM[kk]=' ';                    
            kk++;
        }

        for(int j=j_old; j<m2[k]; j++)
        {
            //align y to gap
            seqxA[kk]='-';
            seqyA[kk]=seqy[j];
            seqM[kk]=' ';
            kk++;
        }

        seqxA[kk]=seqx[m1[k]];
        seqyA[kk]=seqy[m2[k]];
        Liden+=(seqxA[kk]==seqyA[kk]);
        d=sqrt(dist(&xt[m1[k]][0], &ya[m2[k]][0]));
        if(d<d0_out) seqM[kk]=':';
        else         seqM[kk]='.';
        kk++;  
        i_old=m1[k]+1;
        j_old=m2[k]+1;
    }

    //tail
    for(int i=i_old; i<xlen; i++)
    {
        //align x to gap
        seqxA[kk]=seqx[i];
        seqyA[kk]='-';
        seqM[kk]=' ';
        kk++;
    }    
    for(int j=j_old; j<ylen; j++)
    {
        //align y to gap
        seqxA[kk]='-';
        seqyA[kk]=seqy[j];
        seqM[kk]=' ';
        kk++;
    }
    seqxA=seqxA.substr(0,kk);
    seqyA=seqyA.substr(0,kk);
    seqM =seqM.substr(0,kk);

    /* free memory */
    clean_up_after_approx_TM(invmap0, invmap, score, path, val,
        xtm, ytm, xt, r1, r2, xlen, minlen);
    delete [] m1;
    delete [] m2;
    return 0; // zero for no exception
}

/* entry function for TM-align with circular permutation
 * i_opt, a_opt, u_opt, d_opt, TMcut are not implemented yet */
int CPalign_main(double **xa, double **ya,
    const char *seqx, const char *seqy, const char *secx, const char *secy,
    double t0[3], double u0[3][3],
    double &TM1, double &TM2, double &TM3, double &TM4, double &TM5,
    double &d0_0, double &TM_0,
    double &d0A, double &d0B, double &d0u, double &d0a, double &d0_out,
    string &seqM, string &seqxA, string &seqyA,
    double &rmsd0, int &L_ali, double &Liden,
    double &TM_ali, double &rmsd_ali, int &n_ali, int &n_ali8,
    const int xlen, const int ylen,
    const vector<string> sequence, const double Lnorm_ass,
    const double d0_scale, const int i_opt, const int a_opt,
    const bool u_opt, const bool d_opt, const bool fast_opt,
    const int mol_type, const double TMcut=-1)
{
    char   *seqx_cp; // for the protein sequence 
    char   *secx_cp; // for the secondary structure 
    double **xa_cp;   // coordinates
    string seqxA_cp,seqyA_cp;  // alignment
    int    i,r;
    int    cp_point=0;    // position of circular permutation
    int    cp_aln_best=0; // amount of aligned residue in sliding window
    int    cp_aln_current;// amount of aligned residue in sliding window

    /* duplicate structure */
    NewArray(&xa_cp, xlen*2, 3);
    seqx_cp = new char[xlen*2 + 1];
    secx_cp = new char[xlen*2 + 1];
    for (r=0;r<xlen;r++)
    {
        xa_cp[r+xlen][0]=xa_cp[r][0]=xa[r][0];
        xa_cp[r+xlen][1]=xa_cp[r][1]=xa[r][1];
        xa_cp[r+xlen][2]=xa_cp[r][2]=xa[r][2];
        seqx_cp[r+xlen]=seqx_cp[r]=seqx[r];
        secx_cp[r+xlen]=secx_cp[r]=secx[r];
    }
    seqx_cp[2*xlen]=0;
    secx_cp[2*xlen]=0;
    
    /* fTM-align alignment */
    double TM1_cp,TM2_cp,TM4_cp;
    const double Lnorm_tmp=getmin(xlen,ylen);
    TMalign_main(xa_cp, ya, seqx_cp, seqy, secx_cp, secy,
        t0, u0, TM1_cp, TM2_cp, TM3, TM4_cp, TM5,
        d0_0, TM_0, d0A, d0B, d0u, d0a, d0_out, seqM, seqxA_cp, seqyA_cp,
        rmsd0, L_ali, Liden, TM_ali, rmsd_ali, n_ali, n_ali8,
        xlen*2, ylen, sequence, Lnorm_tmp, d0_scale,
        0, false, true, false, true, mol_type, -1);

    /* delete gap in seqxA_cp */
    r=0;
    seqxA=seqxA_cp;
    seqyA=seqyA_cp;
    for (i=0;i<seqxA_cp.size();i++)
    {
        if (seqxA_cp[i]!='-')
        {
            seqxA[r]=seqxA_cp[i];
            seqyA[r]=seqyA_cp[i];
            r++;
        }
    }
    seqxA=seqxA.substr(0,r);
    seqyA=seqyA.substr(0,r);

    /* count the number of aligned residues in each window
     * r - residue index in the original unaligned sequence 
     * i - position in the alignment */
    for (r=0;r<xlen-1;r++)
    {
        cp_aln_current=0;
        for (i=r;i<r+xlen;i++) cp_aln_current+=(seqyA[i]!='-');

        if (cp_aln_current>cp_aln_best)
        {
            cp_aln_best=cp_aln_current;
            cp_point=r;
        }
    }
    seqM.clear();
    seqxA.clear();
    seqyA.clear();
    seqxA_cp.clear();
    seqyA_cp.clear();
    rmsd0=Liden=n_ali=n_ali8=0;

    /* fTM-align alignment */
    TMalign_main(xa, ya, seqx, seqy, secx, secy,
        t0, u0, TM1, TM2, TM3, TM4, TM5,
        d0_0, TM_0, d0A, d0B, d0u, d0a, d0_out, seqM, seqxA, seqyA,
        rmsd0, L_ali, Liden, TM_ali, rmsd_ali, n_ali, n_ali8,
        xlen, ylen, sequence, Lnorm_tmp, d0_scale,
        0, false, true, false, true, mol_type, -1);

    /* do not use circular permutation of number of aligned residues is not
     * larger than sequence-order dependent alignment */
    //cout<<"cp: aln="<<cp_aln_best<<"\tTM="<<TM4_cp<<endl;
    //cout<<"TM: aln="<<n_ali8<<"\tTM="<<TM4<<endl;
    if (n_ali8>=cp_aln_best || TM4>=TM4_cp) cp_point=0;

    /* prepare structure for final alignment */
    seqM.clear();
    seqxA.clear();
    seqyA.clear();
    rmsd0=Liden=n_ali=n_ali8=0;
    if (cp_point!=0)
    {
        for (r=0;r<xlen;r++)
        {
            xa_cp[r][0]=xa_cp[r+cp_point][0];
            xa_cp[r][1]=xa_cp[r+cp_point][1];
            xa_cp[r][2]=xa_cp[r+cp_point][2];
            seqx_cp[r]=seqx_cp[r+cp_point];
            secx_cp[r]=secx_cp[r+cp_point];
        }
    }
    seqx_cp[xlen]=0;
    secx_cp[xlen]=0;

    /* test another round of alignment as concatenated alignment can
     * inflate the number of aligned residues and TM-score. e.g. 1yadA 2duaA */
    if (cp_point!=0)
    {
        TMalign_main(xa_cp, ya, seqx_cp, seqy, secx_cp, secy,
            t0, u0, TM1_cp, TM2_cp, TM3, TM4_cp, TM5,
            d0_0, TM_0, d0A, d0B, d0u, d0a, d0_out, seqM, seqxA_cp, seqyA_cp,
            rmsd0, L_ali, Liden, TM_ali, rmsd_ali, n_ali, cp_aln_best,
            xlen, ylen, sequence, Lnorm_tmp, d0_scale,
            0, false, true, false, true, mol_type, -1);
        //cout<<"cp: aln="<<cp_aln_best<<"\tTM="<<TM4_cp<<endl;
        if (n_ali8>=cp_aln_best || TM4>=TM4_cp)
        {
            cp_point=0;
            for (r=0;r<xlen;r++)
            {
                xa_cp[r][0]=xa[r][0];
                xa_cp[r][1]=xa[r][1];
                xa_cp[r][2]=xa[r][2];
                seqx_cp[r]=seqx[r];
                secx_cp[r]=secx[r];
            }
        }
    }

    /* full TM-align */
    TMalign_main(xa_cp, ya, seqx_cp, seqy, secx_cp, secy,
        t0, u0, TM1, TM2, TM3, TM4, TM5,
        d0_0, TM_0, d0A, d0B, d0u, d0a, d0_out, seqM, seqxA_cp, seqyA_cp,
        rmsd0, L_ali, Liden, TM_ali, rmsd_ali, n_ali, n_ali8,
        xlen, ylen, sequence, Lnorm_ass, d0_scale,
        i_opt, a_opt, u_opt, d_opt, fast_opt, mol_type, TMcut);

    /* correct alignment
     * r - residue index in the original unaligned sequence 
     * i - position in the alignment */
    if (cp_point>0)
    {
        r=0;
        for (i=0;i<seqxA_cp.size();i++)
        {
            r+=(seqxA_cp[i]!='-');
            if (r>=(xlen-cp_point)) 
            {
                i++;
                break;
            }
        }
        seqxA=seqxA_cp.substr(0,i)+'*'+seqxA_cp.substr(i);
        seqM =seqM.substr(0,i)    +' '+seqM.substr(i);
        seqyA=seqyA_cp.substr(0,i)+'-'+seqyA_cp.substr(i);
    }
    else
    {
        seqxA=seqxA_cp;
        seqyA=seqyA_cp;
    }

    /* clean up */
    delete[]seqx_cp;
    delete[]secx_cp;
    DeleteArray(&xa_cp,xlen*2);
    seqxA_cp.clear();
    seqyA_cp.clear();
    return cp_point;
}

/* entry function for se
 * outfmt_opt>=2 should not parse sequence alignment 
 * u_opt corresponds to option -L
 *       if u_opt==2, use d0 from Lnorm_ass for alignment
 * */
int se_main(
    double **xa, double **ya, const char *seqx, const char *seqy,
    double &TM1, double &TM2, double &TM3, double &TM4, double &TM5,
    double &d0_0, double &TM_0,
    double &d0A, double &d0B, double &d0u, double &d0a, double &d0_out,
    string &seqM, string &seqxA, string &seqyA,
    double &rmsd0, int &L_ali, double &Liden,
    double &TM_ali, double &rmsd_ali, int &n_ali, int &n_ali8,
    const int xlen, const int ylen, const vector<string> &sequence,
    const double Lnorm_ass, const double d0_scale, const bool i_opt,
    const bool a_opt, const int u_opt, const bool d_opt, const int mol_type,
    const int outfmt_opt, int *invmap)
{
    double D0_MIN;        //for d0
    double Lnorm;         //normalization length
    double score_d8,d0,d0_search,dcu0;//for TMscore search
    double **score;       // Input score table for dynamic programming
    bool   **path;        // for dynamic programming  
    double **val;         // for dynamic programming  

    int *m1=NULL;
    int *m2=NULL;
    double d;
    if (outfmt_opt<2)
    {
        m1=new int[xlen]; //alignd index in x
        m2=new int[ylen]; //alignd index in y
    }

    /***********************/
    /* allocate memory     */
    /***********************/
    NewArray(&score, xlen+1, ylen+1);
    NewArray(&path, xlen+1, ylen+1);
    NewArray(&val, xlen+1, ylen+1);
    //int *invmap          = new int[ylen+1];

    /* set d0 */
    parameter_set4search(xlen, ylen, D0_MIN, Lnorm,
        score_d8, d0, d0_search, dcu0); // set score_d8
    parameter_set4final(xlen, D0_MIN, Lnorm,
        d0B, d0_search, mol_type); // set d0B
    parameter_set4final(ylen, D0_MIN, Lnorm,
        d0A, d0_search, mol_type); // set d0A
    if (a_opt)
        parameter_set4final((xlen+ylen)*0.5, D0_MIN, Lnorm,
            d0a, d0_search, mol_type); // set d0a
    if (u_opt)
    {
        parameter_set4final(Lnorm_ass, D0_MIN, Lnorm,
            d0u, d0_search, mol_type); // set d0u
        if (u_opt==2)
        {
            parameter_set4search(Lnorm_ass, Lnorm_ass, D0_MIN, Lnorm,
                score_d8, d0, d0_search, dcu0); // set score_d8
        }
    }

    /* perform alignment */
    for(int j=0; j<ylen; j++) invmap[j]=-1;
    if (!i_opt) NWDP_SE(path, val, xa, ya, xlen, ylen, d0*d0, 0, invmap);
    else
    {
        int i1 = -1;// in C version, index starts from zero, not from one
        int i2 = -1;
        int L1 = sequence[0].size();
        int L2 = sequence[1].size();
        int L = min(L1, L2);// Get positions for aligned residues
        for (int kk1 = 0; kk1 < L; kk1++)
        {
            if (sequence[0][kk1] != '-') i1++;
            if (sequence[1][kk1] != '-')
            {
                i2++;
                if (i2 >= ylen || i1 >= xlen) kk1 = L;
                else if (sequence[0][kk1] != '-') invmap[i2] = i1;
            }
        }
    }

    rmsd0=TM1=TM2=TM3=TM4=TM5=0;
    int k=0;
    n_ali=0;
    n_ali8=0;
    for(int i=0,j=0; j<ylen; j++)
    {
        i=invmap[j];
        if(i>=0)//aligned
        {
            n_ali++;
            d=sqrt(dist(&xa[i][0], &ya[j][0]));
            if (d <= score_d8 || i_opt)
            {
                if (outfmt_opt<2)
                {
                    m1[k]=i;
                    m2[k]=j;
                }
                k++;
                TM2+=1/(1+(d/d0B)*(d/d0B)); // chain_1
                TM1+=1/(1+(d/d0A)*(d/d0A)); // chain_2
                if (a_opt) TM3+=1/(1+(d/d0a)*(d/d0a)); // -a
                if (u_opt) TM4+=1/(1+(d/d0u)*(d/d0u)); // -u
                if (d_opt) TM5+=1/(1+(d/d0_scale)*(d/d0_scale)); // -d
                rmsd0+=d*d;
            }
        }
    }
    n_ali8=k;
    TM2/=xlen;
    TM1/=ylen;
    TM3/=(xlen+ylen)*0.5;
    TM4/=Lnorm_ass;
    TM5/=ylen;
    if (n_ali8) rmsd0=sqrt(rmsd0/n_ali8);

    if (outfmt_opt>=2)
    {
        DeleteArray(&score, xlen+1);
        DeleteArray(&path, xlen+1);
        DeleteArray(&val, xlen+1);
        return 0;
    }

    /* extract aligned sequence */
    int ali_len=xlen+ylen; //maximum length of alignment
    seqxA.assign(ali_len,'-');
    seqM.assign( ali_len,' ');
    seqyA.assign(ali_len,'-');
    
    int kk=0, i_old=0, j_old=0;
    d=0;
    Liden=0;
    for(int k=0; k<n_ali8; k++)
    {
        for(int i=i_old; i<m1[k]; i++)
        {
            //align x to gap
            seqxA[kk]=seqx[i];
            seqyA[kk]='-';
            seqM[kk]=' ';                    
            kk++;
        }

        for(int j=j_old; j<m2[k]; j++)
        {
            //align y to gap
            seqxA[kk]='-';
            seqyA[kk]=seqy[j];
            seqM[kk]=' ';
            kk++;
        }

        seqxA[kk]=seqx[m1[k]];
        seqyA[kk]=seqy[m2[k]];
        Liden+=(seqxA[kk]==seqyA[kk]);
        d=sqrt(dist(&xa[m1[k]][0], &ya[m2[k]][0]));
        if(d<d0_out) seqM[kk]=':';
        else         seqM[kk]='.';
        kk++;  
        i_old=m1[k]+1;
        j_old=m2[k]+1;
    }

    //tail
    for(int i=i_old; i<xlen; i++)
    {
        //align x to gap
        seqxA[kk]=seqx[i];
        seqyA[kk]='-';
        seqM[kk]=' ';
        kk++;
    }    
    for(int j=j_old; j<ylen; j++)
    {
        //align y to gap
        seqxA[kk]='-';
        seqyA[kk]=seqy[j];
        seqM[kk]=' ';
        kk++;
    }
    seqxA=seqxA.substr(0,kk);
    seqyA=seqyA.substr(0,kk);
    seqM =seqM.substr(0,kk);

    /* free memory */
    //delete [] invmap;
    delete [] m1;
    delete [] m2;
    DeleteArray(&score, xlen+1);
    DeleteArray(&path, xlen+1);
    DeleteArray(&val, xlen+1);
    return 0; // zero for no exception
}

/* count the number of nucleic acid chains (na_chain_num) and
 * protein chains (aa_chain_num) in a complex */
int count_na_aa_chain_num(int &na_chain_num,int &aa_chain_num,
    const vector<int>&mol_vec)
{
    na_chain_num=0;
    aa_chain_num=0;
    for (size_t i=0;i<mol_vec.size();i++)
    {
        if (mol_vec[i]>0) na_chain_num++;
        else              aa_chain_num++;
    }
    return na_chain_num+aa_chain_num;
}

/* adjust chain assignment for dimer-dimer alignment 
 * return true if assignment is adjusted */
bool adjust_dimer_assignment(        
    const vector<vector<vector<double> > >&xa_vec,
    const vector<vector<vector<double> > >&ya_vec,
    const vector<int>&xlen_vec, const vector<int>&ylen_vec,
    const vector<int>&mol_vec1, const vector<int>&mol_vec2,
    int *assign1_list, int *assign2_list,
    const vector<vector<string> >&seqxA_mat,
    const vector<vector<string> >&seqyA_mat)
{
    /* check currently assigned chains */
    int i1,i2,j1,j2;
    i1=i2=j1=j2=-1;    
    int chain1_num=xa_vec.size();
    int i,j;
    for (i=0;i<chain1_num;i++)
    {
        if (assign1_list[i]>=0)
        {
            if (i1<0)
            {
                i1=i;
                j1=assign1_list[i1];
            }
            else
            {
                i2=i;
                j2=assign1_list[i2];
            }
        }
    }

    /* normalize d0 by L */
    int xlen=xlen_vec[i1]+xlen_vec[i2];
    int ylen=ylen_vec[j1]+ylen_vec[j2];
    int mol_type=mol_vec1[i1]+mol_vec1[i2]+
                 mol_vec2[j1]+mol_vec2[j2];
    double D0_MIN, d0, d0_search;
    double Lnorm=getmin(xlen,ylen);
    parameter_set4final(getmin(xlen,ylen), D0_MIN, Lnorm, d0, 
        d0_search, mol_type);

    double **xa,**ya, **xt;
    NewArray(&xa, xlen, 3);
    NewArray(&ya, ylen, 3);
    NewArray(&xt, xlen, 3);

    double RMSD = 0;
    double dd   = 0;
    double t[3];
    double u[3][3];
    size_t L_ali=0; // index of residue in aligned region
    size_t r=0;     // index of residue in full alignment

    /* total score using current assignment */
    L_ali=0;
    i=j=-1;
    for (r=0;r<seqxA_mat[i1][j1].size();r++)
    {
        i+=(seqxA_mat[i1][j1][r]!='-');
        j+=(seqyA_mat[i1][j1][r]!='-');
        if (seqxA_mat[i1][j1][r]=='-' || seqyA_mat[i1][j1][r]=='-') continue;
        xa[L_ali][0]=xa_vec[i1][i][0];
        xa[L_ali][1]=xa_vec[i1][i][1];
        xa[L_ali][2]=xa_vec[i1][i][2];
        ya[L_ali][0]=ya_vec[j1][j][0];
        ya[L_ali][1]=ya_vec[j1][j][1];
        ya[L_ali][2]=ya_vec[j1][j][2];
        L_ali++;
    }
    i=j=-1;
    for (r=0;r<seqxA_mat[i2][j2].size();r++)
    {
        i+=(seqxA_mat[i2][j2][r]!='-');
        j+=(seqyA_mat[i2][j2][r]!='-');
        if (seqxA_mat[i2][j2][r]=='-' || seqyA_mat[i2][j2][r]=='-') continue;
        xa[L_ali][0]=xa_vec[i2][i][0];
        xa[L_ali][1]=xa_vec[i2][i][1];
        xa[L_ali][2]=xa_vec[i2][i][2];
        ya[L_ali][0]=ya_vec[j2][j][0];
        ya[L_ali][1]=ya_vec[j2][j][1];
        ya[L_ali][2]=ya_vec[j2][j][2];
        L_ali++;
    }

    Kabsch(xa, ya, L_ali, 1, &RMSD, t, u);
    do_rotation(xa, xt, L_ali, t, u);

    double total_score1=0;
    for (r=0;r<L_ali;r++)
    {
        dd=dist(xt[r],ya[r]);
        total_score1+=1/(1+dd/d0*d0);
    }
    total_score1/=Lnorm;

    /* total score using reversed assignment */
    L_ali=0;
    i=j=-1;
    for (r=0;r<seqxA_mat[i1][j2].size();r++)
    {
        i+=(seqxA_mat[i1][j2][r]!='-');
        j+=(seqyA_mat[i1][j2][r]!='-');
        if (seqxA_mat[i1][j2][r]=='-' || seqyA_mat[i1][j2][r]=='-') continue;
        xa[L_ali][0]=xa_vec[i1][i][0];
        xa[L_ali][1]=xa_vec[i1][i][1];
        xa[L_ali][2]=xa_vec[i1][i][2];
        ya[L_ali][0]=ya_vec[j2][j][0];
        ya[L_ali][1]=ya_vec[j2][j][1];
        ya[L_ali][2]=ya_vec[j2][j][2];
        L_ali++;
    }
    i=j=-1;
    for (r=0;r<seqxA_mat[i2][j1].size();r++)
    {
        i+=(seqxA_mat[i2][j1][r]!='-');
        j+=(seqyA_mat[i2][j1][r]!='-');
        if (seqxA_mat[i2][j1][r]=='-' || seqyA_mat[i2][j1][r]=='-') continue;
        xa[L_ali][0]=xa_vec[i2][i][0];
        xa[L_ali][1]=xa_vec[i2][i][1];
        xa[L_ali][2]=xa_vec[i2][i][2];
        ya[L_ali][0]=ya_vec[j1][j][0];
        ya[L_ali][1]=ya_vec[j1][j][1];
        ya[L_ali][2]=ya_vec[j1][j][2];
        L_ali++;
    }

    Kabsch(xa, ya, L_ali, 1, &RMSD, t, u);
    do_rotation(xa, xt, L_ali, t, u);

    double total_score2=0;
    for (r=0;r<L_ali;r++)
    {
        dd=dist(xt[r],ya[r]);
        total_score2+=1/(1+dd/d0*d0);
    }
    total_score2/=Lnorm;

    /* swap chain assignment */
    if (total_score1<total_score2)
    {
        assign1_list[i1]=j2;
        assign1_list[i2]=j1;
        assign2_list[j1]=i2;
        assign2_list[j2]=i1;
    }

    /* clean up */
    DeleteArray(&xa, xlen);
    DeleteArray(&ya, ylen);
    DeleteArray(&xt, xlen);
    return total_score1<total_score2;
}

/* count how many chains are paired */
int count_assign_pair(int *assign1_list,const int chain1_num)
{
    int pair_num=0;
    int i;
    for (i=0;i<chain1_num;i++) pair_num+=(assign1_list[i]>=0);
    return pair_num;
}


/* assign chain-chain correspondence */
double enhanced_greedy_search(double **TMave_mat,int *assign1_list,
    int *assign2_list, const int chain1_num, const int chain2_num)
{
    double total_score=0;
    double tmp_score=0;
    int i,j;
    int maxi=0;
    int maxj=0;

    /* initialize parameters */
    for (i=0;i<chain1_num;i++) assign1_list[i]=-1;
    for (j=0;j<chain2_num;j++) assign2_list[j]=-1;

    /* greedy assignment: in each iteration, the highest chain pair is
     * assigned, until no assignable chain is left */
    while(1)
    {
        tmp_score=-1;
        for (i=0;i<chain1_num;i++)
        {
            if (assign1_list[i]>=0) continue;
            for (j=0;j<chain2_num;j++)
            {
                if (assign2_list[j]>=0 || TMave_mat[i][j]<=0) continue;
                if (TMave_mat[i][j]>tmp_score) 
                {
                    maxi=i;
                    maxj=j;
                    tmp_score=TMave_mat[i][j];
                }
            }
        }
        if (tmp_score<=0) break; // error: no assignable chain
        assign1_list[maxi]=maxj;
        assign2_list[maxj]=maxi;
        total_score+=tmp_score;
    }
    if (total_score<=0) return total_score; // error: no assignable chain
    //cout<<"assign1_list={";
    //for (i=0;i<chain1_num;i++) cout<<assign1_list[i]<<","; cout<<"}"<<endl;
    //cout<<"assign2_list={";
    //for (j=0;j<chain2_num;j++) cout<<assign2_list[j]<<","; cout<<"}"<<endl;

    /* iterative refinemnt */
    double delta_score;
    int *assign1_tmp=new int [chain1_num];
    int *assign2_tmp=new int [chain2_num];
    for (i=0;i<chain1_num;i++) assign1_tmp[i]=assign1_list[i];
    for (j=0;j<chain2_num;j++) assign2_tmp[j]=assign2_list[j];
    int old_i=-1;
    int old_j=-1;

    for (int iter=0;iter<getmin(chain1_num,chain2_num)*5;iter++)
    {
        delta_score=-1;
        for (i=0;i<chain1_num;i++)
        {
            old_j=assign1_list[i];
            for (j=0;j<chain2_num;j++)
            {
                // attempt to swap (i,old_j=assign1_list[i]) with (i,j)
                if (j==assign1_list[i] || TMave_mat[i][j]<=0) continue;
                old_i=assign2_list[j];

                assign1_tmp[i]=j;
                if (old_i>=0) assign1_tmp[old_i]=old_j;
                assign2_tmp[j]=i;
                if (old_j>=0) assign2_tmp[old_j]=old_i;

                delta_score=TMave_mat[i][j];
                if (old_j>=0) delta_score-=TMave_mat[i][old_j];
                if (old_i>=0) delta_score-=TMave_mat[old_i][j];
                if (old_i>=0 && old_j>=0) delta_score+=TMave_mat[old_i][old_j];

                if (delta_score>0) // successful swap
                {
                    assign1_list[i]=j;
                    if (old_i>=0) assign1_list[old_i]=old_j;
                    assign2_list[j]=i;
                    if (old_j>=0) assign2_list[old_j]=old_i;
                    total_score+=delta_score;
                    break;
                }
                else
                {
                    assign1_tmp[i]=assign1_list[i];
                    if (old_i>=0) assign1_tmp[old_i]=assign1_list[old_i];
                    assign2_tmp[j]=assign2_list[j];
                    if (old_j>=0) assign2_tmp[old_j]=assign2_list[old_j];
                }
            }
            if (delta_score>0) break;
        }
        if (delta_score<=0) break; // cannot swap any chain pair
    }

    /* clean up */
    delete[]assign1_tmp;
    delete[]assign2_tmp;
    return total_score;
}

double calculate_centroids(const vector<vector<vector<double> > >&a_vec,
    const int chain_num, double ** centroids)
{
    int L=0;
    int c,r; // index of chain and residue
    for (c=0; c<chain_num; c++)
    {
        centroids[c][0]=0;
        centroids[c][1]=0;
        centroids[c][2]=0;
        L=a_vec[c].size();
        for (r=0; r<L; r++)
        {
            centroids[c][0]+=a_vec[c][r][0];
            centroids[c][1]+=a_vec[c][r][1];
            centroids[c][2]+=a_vec[c][r][2];
        }
        centroids[c][0]/=L;
        centroids[c][1]/=L;
        centroids[c][2]/=L;
        //cout<<centroids[c][0]<<'\t'
            //<<centroids[c][1]<<'\t'
            //<<centroids[c][2]<<endl;
    }

    vector<double> d0_vec(chain_num,-1);
    int c2=0;
    double d0MM=0;
    for (c=0; c<chain_num; c++)
    {
        for (c2=0; c2<chain_num; c2++)
        {
            if (c2==c) continue;
            d0MM=sqrt(dist(centroids[c],centroids[c2]));
            if (d0_vec[c]<=0) d0_vec[c]=d0MM;
            else d0_vec[c]=getmin(d0_vec[c], d0MM);
        }
    }
    d0MM=0;
    for (c=0; c<chain_num; c++) d0MM+=d0_vec[c];
    d0MM/=chain_num;
    d0_vec.clear();
    //cout<<d0MM<<endl;
    return d0MM;
}

/* calculate MMscore of aligned chains
 * MMscore = sum(TMave_mat[i][j]) * sum(1/(1+dij^2/d0MM^2)) 
 *         / (L* getmin(chain1_num,chain2_num))
 * dij is the centroid distance between chain pair i and j
 * d0MM is scaling factor. TMave_mat[i][j] is the TM-score between
 * chain pair i and j multiple by getmin(Li*Lj) */
double calMMscore(double **TMave_mat,int *assign1_list,
    const int chain1_num, const int chain2_num, double **xcentroids,
    double **ycentroids, const double d0MM, double **r1, double **r2,
    double **xt, double t[3], double u[3][3], const int L)
{
    int Nali=0; // number of aligned chain
    int i,j;
    double MMscore=0;
    for (i=0;i<chain1_num;i++)
    {
        j=assign1_list[i];
        if (j<0) continue;

        r1[Nali][0]=xcentroids[i][0];
        r1[Nali][1]=xcentroids[i][1];
        r1[Nali][2]=xcentroids[i][2];

        r2[Nali][0]=ycentroids[j][0];
        r2[Nali][1]=ycentroids[j][1];
        r2[Nali][2]=ycentroids[j][2];

        Nali++;
        MMscore+=TMave_mat[i][j];
    }
    MMscore/=L;

    double RMSD = 0;
    double TMscore=0;
    if (Nali>=3)
    {
        /* Kabsch superposition */
        Kabsch(r1, r2, Nali, 1, &RMSD, t, u);
        do_rotation(r1, xt, Nali, t, u);

        /* calculate pseudo-TMscore */
        double dd=0;
        for (i=0;i<Nali;i++)
        {
            dd=dist(xt[i], r2[i]);
            TMscore+=1/(1+dd/(d0MM*d0MM));
        }
    }
    else if (Nali==2)
    {
        double dd=dist(r1[0],r2[0]);
        TMscore=1/(1+dd/(d0MM*d0MM));
    }
    else TMscore=1; // only one aligned chain.
    TMscore/=getmin(chain1_num,chain2_num);
    MMscore*=TMscore;
    return MMscore;
}

/* reassign chain-chain correspondence, specific for homooligomer */
double homo_refined_greedy_search(double **TMave_mat,int *assign1_list,
    int *assign2_list, const int chain1_num, const int chain2_num,
    double **xcentroids, double **ycentroids, const double d0MM,
    const int L, double **ut_mat)
{
    double MMscore_max=0;
    double MMscore=0;
    int i,j;
    int c1,c2;
    int max_i=-1; // the chain pair whose monomer u t yields highest MMscore
    int max_j=-1;

    int chain_num=getmin(chain1_num,chain2_num);
    int *assign1_tmp=new int [chain1_num];
    int *assign2_tmp=new int [chain2_num];
    double **xt;
    NewArray(&xt, chain1_num, 3);
    double t[3];
    double u[3][3];
    int ui,uj,ut_idx;
    double TMscore=0; // pseudo TM-score
    double TMsum  =0;
    double TMnow  =0;
    double TMmax  =0;
    double dd=0;

    size_t  total_pair=chain1_num*chain2_num; // total pair
    double *ut_tmc_mat=new double [total_pair]; // chain level TM-score
    vector<pair<double,int> > ut_tm_vec(total_pair,make_pair(0.0,0)); // product of both

    for (c1=0;c1<chain1_num;c1++)
    {
        for (c2=0;c2<chain2_num;c2++)
        {
            if (TMave_mat[c1][c2]<=0) continue;
            ut_idx=c1*chain2_num+c2;
            for (ui=0;ui<3;ui++)
                for (uj=0;uj<3;uj++) u[ui][uj]=ut_mat[ut_idx][ui*3+uj];
            for (uj=0;uj<3;uj++) t[uj]=ut_mat[ut_idx][9+uj];
            
            do_rotation(xcentroids, xt, chain1_num, t, u);

            for (i=0;i<chain1_num;i++) assign1_tmp[i]=-1;
            for (j=0;j<chain2_num;j++) assign2_tmp[j]=-1;


            for (i=0;i<chain1_num;i++)
            {
                for (j=0;j<chain2_num;j++)
                {
                    ut_idx=i*chain2_num+j;
                    ut_tmc_mat[ut_idx]=0;
                    ut_tm_vec[ut_idx].first=-1;
                    ut_tm_vec[ut_idx].second=ut_idx;
                    if (TMave_mat[i][j]<=0) continue;
                    dd=dist(xt[i],ycentroids[j]);
                    ut_tmc_mat[ut_idx]=1/(1+dd/(d0MM*d0MM));
                    ut_tm_vec[ut_idx].first=
                        ut_tmc_mat[ut_idx]*TMave_mat[i][j];
                    //cout<<"TM["<<ut_idx<<"]="<<ut_tm_vec[ut_idx].first<<endl;
                }
            }
            //cout<<"sorting "<<total_pair<<" chain pairs"<<endl;

            /* initial assignment */
            assign1_tmp[c1]=c2;
            assign2_tmp[c2]=c1;
            TMsum=TMave_mat[c1][c2];
            TMscore=ut_tmc_mat[c1*chain2_num+c2];

            /* further assignment */
            sort(ut_tm_vec.begin(), ut_tm_vec.end()); // sort in ascending order
            for (ut_idx=total_pair-1;ut_idx>=0;ut_idx--)
            {
                j=ut_tm_vec[ut_idx].second % chain2_num;
                i=int(ut_tm_vec[ut_idx].second / chain2_num);
                if (TMave_mat[i][j]<=0) break;
                if (assign1_tmp[i]>=0 || assign2_tmp[j]>=0) continue;
                assign1_tmp[i]=j;
                assign2_tmp[j]=i;
                TMsum+=TMave_mat[i][j];
                TMscore+=ut_tmc_mat[i*chain2_num+j];
                //cout<<"ut_idx="<<ut_tm_vec[ut_idx].second
                    //<<"\ti="<<i<<"\tj="<<j<<"\ttm="<<ut_tm_vec[ut_idx].first<<endl;
            }

            /* final MMscore */
            MMscore=(TMsum/L)*(TMscore/chain_num);
            if (max_i<0 || max_j<0 || MMscore>MMscore_max)
            {
                max_i=c1;
                max_j=c2;
                MMscore_max=MMscore;
                for (i=0;i<chain1_num;i++) assign1_list[i]=assign1_tmp[i];
                for (j=0;j<chain2_num;j++) assign2_list[j]=assign2_tmp[j];
                //cout<<"TMsum/L="<<TMsum/L<<endl;
                //cout<<"TMscore/chain_num="<<TMscore/chain_num<<endl;
                //cout<<"MMscore="<<MMscore<<endl;
                //cout<<"assign1_list={";
                //for (i=0;i<chain1_num;i++) 
                    //cout<<assign1_list[i]<<","; cout<<"}"<<endl;
                //cout<<"assign2_list={";
                //for (j=0;j<chain2_num;j++)
                    //cout<<assign2_list[j]<<","; cout<<"}"<<endl;
            }
        }
    }

    /* clean up */
    delete[]assign1_tmp;
    delete[]assign2_tmp;
    delete[]ut_tmc_mat;
    ut_tm_vec.clear();
    DeleteArray(&xt, chain1_num);
    return MMscore;
}

/* reassign chain-chain correspondence, specific for heterooligomer */
double hetero_refined_greedy_search(double **TMave_mat,int *assign1_list,
    int *assign2_list, const int chain1_num, const int chain2_num,
    double **xcentroids, double **ycentroids, const double d0MM, const int L)
{
    double MMscore_old=0;
    double MMscore=0;
    int i,j;

    double **r1;
    double **r2;
    double **xt;
    int chain_num=getmin(chain1_num,chain2_num);
    NewArray(&r1, chain_num, 3);
    NewArray(&r2, chain_num, 3);
    NewArray(&xt, chain_num, 3);
    double t[3];
    double u[3][3];

    /* calculate MMscore */
    MMscore=MMscore_old=calMMscore(TMave_mat, assign1_list, chain1_num,
        chain2_num, xcentroids, ycentroids, d0MM, r1, r2, xt, t, u, L);
    //cout<<"MMscore="<<MMscore<<endl;

    /* iteratively refine chain assignment. in each iteration, attempt
     * to swap (i,old_j=assign1_list[i]) with (i,j) */
    double delta_score=-1;
    int *assign1_tmp=new int [chain1_num];
    int *assign2_tmp=new int [chain2_num];
    for (i=0;i<chain1_num;i++) assign1_tmp[i]=assign1_list[i];
    for (j=0;j<chain2_num;j++) assign2_tmp[j]=assign2_list[j];
    int old_i=-1;
    int old_j=-1;
    for (int iter=0;iter<chain1_num*chain2_num;iter++)
    {
        delta_score=-1;
        for (i=0;i<chain1_num;i++)
        {
            old_j=assign1_list[i];
            for (j=0;j<chain2_num;j++)
            {
                if (j==assign1_list[i] || TMave_mat[i][j]<=0) continue;
                old_i=assign2_list[j];

                assign1_tmp[i]=j;
                if (old_i>=0) assign1_tmp[old_i]=old_j;
                assign2_tmp[j]=i;
                if (old_j>=0) assign2_tmp[old_j]=old_i;
                
                MMscore=calMMscore(TMave_mat, assign1_tmp, chain1_num,
                    chain2_num, xcentroids, ycentroids, d0MM,
                    r1, r2, xt, t, u, L);

                //cout<<"(i,j,old_i,old_j,MMscore)=("<<i<<","<<j<<","
                    //<<old_i<<","<<old_j<<","<<MMscore<<")"<<endl;

                if (MMscore>MMscore_old) // successful swap
                {
                    assign1_list[i]=j;
                    if (old_i>=0) assign1_list[old_i]=old_j;
                    assign2_list[j]=i;
                    if (old_j>=0) assign2_list[old_j]=old_i;
                    delta_score=(MMscore-MMscore_old);
                    MMscore_old=MMscore;
                    //cout<<"MMscore="<<MMscore<<endl;
                    break;
                }
                else
                {
                    assign1_tmp[i]=assign1_list[i];
                    if (old_i>=0) assign1_tmp[old_i]=assign1_list[old_i];
                    assign2_tmp[j]=assign2_list[j];
                    if (old_j>=0) assign2_tmp[old_j]=assign2_list[old_j];
                }
            }
        }
        //cout<<"iter="<<iter<<endl;
        //cout<<"assign1_list={";
        //for (i=0;i<chain1_num;i++) cout<<assign1_list[i]<<","; cout<<"}"<<endl;
        //cout<<"assign2_list={";
        //for (j=0;j<chain2_num;j++) cout<<assign2_list[j]<<","; cout<<"}"<<endl;
        if (delta_score<=0) break; // cannot swap any chain pair
    }
    MMscore=MMscore_old;
    //cout<<"MMscore="<<MMscore<<endl;

    /* clean up */
    delete[]assign1_tmp;
    delete[]assign2_tmp;
    DeleteArray(&r1, chain_num);
    DeleteArray(&r2, chain_num);
    DeleteArray(&xt, chain_num);
    return MMscore;
}

void copy_chain_data(const vector<vector<double> >&a_vec_i,
    const vector<char>&seq_vec_i,const vector<char>&sec_vec_i,
    const int len,double **a,char *seq,char *sec)
{
    int r;
    for (r=0;r<len;r++)
    {
        a[r][0]=a_vec_i[r][0];
        a[r][1]=a_vec_i[r][1];
        a[r][2]=a_vec_i[r][2];
        seq[r]=seq_vec_i[r];
        sec[r]=sec_vec_i[r];
    }
    seq[len]=0;
    sec[len]=0;
}

/* clear chains with L<3 */
void clear_full_PDB_lines(vector<vector<string> > PDB_lines,const string atom_opt)
{
    int chain_i;
    int Lch;
    int a;
    bool select_atom;
    string line;
    for (chain_i=0;chain_i<PDB_lines.size();chain_i++)
    {
        Lch=0;
        for (a=0;a<PDB_lines[chain_i].size();a++)
        {
            line=PDB_lines[chain_i][a];
            if (atom_opt=="auto")
            {
                if (line[17]==' ' && (line[18]=='D'||line[18]==' '))
                     select_atom=(line.compare(12,4," C3'")==0);
                else select_atom=(line.compare(12,4," CA ")==0);
            }
            else     select_atom=(line.compare(12,4,atom_opt)==0);
            Lch+=select_atom;
        }
        if (Lch<3)
        {
            for (a=0;a<PDB_lines[chain_i].size();a++)
                PDB_lines[chain_i][a].clear();
            PDB_lines[chain_i].clear();
        }
    }
    line.clear();
}

size_t get_full_PDB_lines(const string filename,
    vector<vector<string> >&PDB_lines, const int ter_opt,
    const int infmt_opt, const int split_opt, const int het_opt)
{
    size_t i=0; // resi i.e. atom index
    string line;
    char chainID=0;
    vector<string> tmp_str_vec;
    
    int compress_type=0; // uncompressed file
    ifstream fin;
    redi::ipstream fin_gz; // if file is compressed
    if (filename.size()>=3 && 
        filename.substr(filename.size()-3,3)==".gz")
    {
        fin_gz.open("zcat '"+filename+"'");
        compress_type=1;
    }
    else if (filename.size()>=4 && 
        filename.substr(filename.size()-4,4)==".bz2")
    {
        fin_gz.open("bzcat '"+filename+"'");
        compress_type=2;
    }
    else fin.open(filename.c_str());

    if (infmt_opt==0||infmt_opt==-1) // PDB format
    {
        while (compress_type?fin_gz.good():fin.good())
        {
            if (compress_type) getline(fin_gz, line);
            else               getline(fin, line);
            if (infmt_opt==-1 && line.compare(0,5,"loop_")==0) // PDBx/mmCIF
                return get_full_PDB_lines(filename,PDB_lines,
                    ter_opt, 3, split_opt,het_opt);
            if (i > 0)
            {
                if      (ter_opt>=1 && line.compare(0,3,"END")==0) break;
                else if (ter_opt>=3 && line.compare(0,3,"TER")==0) break;
            }
            if (split_opt && line.compare(0,3,"END")==0) chainID=0;
            if (line.size()>=54 && (line[16]==' ' || line[16]=='A') && (
                (line.compare(0, 6, "ATOM  ")==0) || 
                (line.compare(0, 6, "HETATM")==0 && het_opt==1) ||
                (line.compare(0, 6, "HETATM")==0 && het_opt==2 && 
                 line.compare(17,3, "MSE")==0)))
            {
                if (!chainID)
                {
                    chainID=line[21];
                    PDB_lines.push_back(tmp_str_vec);
                }
                else if (ter_opt>=2 && chainID!=line[21]) break;
                if (split_opt==2 && chainID!=line[21])
                {
                    chainID=line[21];
                    PDB_lines.push_back(tmp_str_vec);
                } 
               
                PDB_lines.back().push_back(line);
                i++;
            }
        }
    }
    else if (infmt_opt==1) // SPICKER format
    {
        size_t L=0;
        float x,y,z;
        stringstream i8_stream;
        while (compress_type?fin_gz.good():fin.good())
        {
            if (compress_type) fin_gz>>L>>x>>y>>z;
            else               fin   >>L>>x>>y>>z;
            if (compress_type) getline(fin_gz, line);
            else               getline(fin, line);
            if (!(compress_type?fin_gz.good():fin.good())) break;
            for (i=0;i<L;i++)
            {
                if (compress_type) fin_gz>>x>>y>>z;
                else               fin   >>x>>y>>z;
                i8_stream<<"ATOM   "<<setw(4)<<i+1<<"  CA  UNK  "<<setw(4)
                    <<i+1<<"    "<<setiosflags(ios::fixed)<<setprecision(3)
                    <<setw(8)<<x<<setw(8)<<y<<setw(8)<<z;
                line=i8_stream.str();
                i8_stream.str(string());
                PDB_lines.back().push_back(line);
            }
            if (compress_type) getline(fin_gz, line);
            else               getline(fin, line);
        }
    }
    else if (infmt_opt==2) // xyz format
    {
        size_t L=0;
        stringstream i8_stream;
        while (compress_type?fin_gz.good():fin.good())
        {
            if (compress_type) getline(fin_gz, line);
            else               getline(fin, line);
            L=atoi(line.c_str());
            if (compress_type) getline(fin_gz, line);
            else               getline(fin, line);
            for (i=0;i<line.size();i++)
                if (line[i]==' '||line[i]=='\t') break;
            if (!(compress_type?fin_gz.good():fin.good())) break;
            PDB_lines.push_back(tmp_str_vec);
            for (i=0;i<L;i++)
            {
                if (compress_type) getline(fin_gz, line);
                else               getline(fin, line);
                i8_stream<<"ATOM   "<<setw(4)<<i+1<<"  CA  "
                    <<AAmap(line[0])<<"  "<<setw(4)<<i+1<<"    "
                    <<line.substr(2,8)<<line.substr(11,8)<<line.substr(20,8);
                line=i8_stream.str();
                i8_stream.str(string());
                PDB_lines.back().push_back(line);
            }
        }
    }
    else if (infmt_opt==3) // PDBx/mmCIF format
    {
        bool loop_ = false; // not reading following content
        map<string,int> _atom_site;
        int atom_site_pos;
        vector<string> line_vec;
        string alt_id=".";  // alternative location indicator
        string asym_id="."; // this is similar to chainID, except that
                            // chainID is char while asym_id is a string
                            // with possibly multiple char
        string prev_asym_id="";
        string AA="";       // residue name
        string atom="";
        string resi="";
        string model_index=""; // the same as model_idx but type is string
        stringstream i8_stream;
        while (compress_type?fin_gz.good():fin.good())
        {
            if (compress_type) getline(fin_gz, line);
            else               getline(fin, line);
            if (line.size()==0) continue;
            if (loop_) loop_ = line.compare(0,2,"# ");
            if (!loop_)
            {
                if (line.compare(0,5,"loop_")) continue;
                while(1)
                {
                    if (compress_type)
                    {
                        if (fin_gz.good()) getline(fin_gz, line);
                        else PrintErrorAndQuit("ERROR! Unexpected end of "+filename);
                    }
                    else
                    {
                        if (fin.good()) getline(fin, line);
                        else PrintErrorAndQuit("ERROR! Unexpected end of "+filename);
                    }
                    if (line.size()) break;
                }
                if (line.compare(0,11,"_atom_site.")) continue;

                loop_=true;
                _atom_site.clear();
                atom_site_pos=0;
                _atom_site[line.substr(11,line.size()-12)]=atom_site_pos;

                while(1)
                {
                    if (compress_type) getline(fin_gz, line);
                    else               getline(fin, line);
                    if (line.size()==0) continue;
                    if (line.compare(0,11,"_atom_site.")) break;
                    _atom_site[line.substr(11,line.size()-12)]=++atom_site_pos;
                }


                if (_atom_site.count("group_PDB")*
                    _atom_site.count("label_atom_id")*
                    _atom_site.count("label_comp_id")*
                   (_atom_site.count("auth_asym_id")+
                    _atom_site.count("label_asym_id"))*
                   (_atom_site.count("auth_seq_id")+
                    _atom_site.count("label_seq_id"))*
                    _atom_site.count("Cartn_x")*
                    _atom_site.count("Cartn_y")*
                    _atom_site.count("Cartn_z")==0)
                {
                    loop_ = false;
                    cerr<<"Warning! Missing one of the following _atom_site data items: group_PDB, label_atom_id, label_comp_id, auth_asym_id/label_asym_id, auth_seq_id/label_seq_id, Cartn_x, Cartn_y, Cartn_z"<<endl;
                    continue;
                }
            }

            line_vec.clear();
            split(line,line_vec);
            if ((line_vec[_atom_site["group_PDB"]]!="ATOM" &&
                 line_vec[_atom_site["group_PDB"]]!="HETATM") ||
                (line_vec[_atom_site["group_PDB"]]=="HETATM" &&
                 (het_opt==0 || 
                 (het_opt==2 && line_vec[_atom_site["label_comp_id"]]!="MSE")))
                ) continue;
            
            alt_id=".";
            if (_atom_site.count("label_alt_id")) // in 39.4 % of entries
                alt_id=line_vec[_atom_site["label_alt_id"]];
            if (alt_id!="." && alt_id!="A") continue;

            atom=line_vec[_atom_site["label_atom_id"]];
            if (atom[0]=='"') atom=atom.substr(1);
            if (atom.size() && atom[atom.size()-1]=='"')
                atom=atom.substr(0,atom.size()-1);
            if (atom.size()==0) continue;
            if      (atom.size()==1) atom=" "+atom+"  ";
            else if (atom.size()==2) atom=" "+atom+" "; // wrong for sidechain H
            else if (atom.size()==3) atom=" "+atom;
            else if (atom.size()>=5) continue;

            AA=line_vec[_atom_site["label_comp_id"]]; // residue name
            if      (AA.size()==1) AA="  "+AA;
            else if (AA.size()==2) AA=" " +AA;
            else if (AA.size()>=4) continue;

            if (_atom_site.count("auth_asym_id"))
                 asym_id=line_vec[_atom_site["auth_asym_id"]];
            else asym_id=line_vec[_atom_site["label_asym_id"]];
            if (asym_id==".") asym_id=" ";
            
            if (_atom_site.count("pdbx_PDB_model_num") && 
                model_index!=line_vec[_atom_site["pdbx_PDB_model_num"]])
            {
                model_index=line_vec[_atom_site["pdbx_PDB_model_num"]];
                if (PDB_lines.size() && ter_opt>=1) break;
                if (PDB_lines.size()==0 || split_opt>=1)
                {
                    PDB_lines.push_back(tmp_str_vec);
                    prev_asym_id=asym_id;
                }
            }

            if (prev_asym_id!=asym_id)
            {
                if (prev_asym_id!="" && ter_opt>=2) break;
                if (split_opt>=2) PDB_lines.push_back(tmp_str_vec);
            }
            if (prev_asym_id!=asym_id) prev_asym_id=asym_id;

            if (_atom_site.count("auth_seq_id"))
                 resi=line_vec[_atom_site["auth_seq_id"]];
            else resi=line_vec[_atom_site["label_seq_id"]];
            if (_atom_site.count("pdbx_PDB_ins_code") && 
                line_vec[_atom_site["pdbx_PDB_ins_code"]]!="?")
                resi+=line_vec[_atom_site["pdbx_PDB_ins_code"]][0];
            else resi+=" ";

            i++;
            i8_stream<<"ATOM  "
                <<setw(5)<<i<<" "<<atom<<" "<<AA<<setw(2)<<asym_id.substr(0,2)
                <<setw(5)<<resi.substr(0,5)<<"   "
                <<setw(8)<<line_vec[_atom_site["Cartn_x"]].substr(0,8)
                <<setw(8)<<line_vec[_atom_site["Cartn_y"]].substr(0,8)
                <<setw(8)<<line_vec[_atom_site["Cartn_z"]].substr(0,8);
            PDB_lines.back().push_back(i8_stream.str());
            i8_stream.str(string());
        }
        _atom_site.clear();
        line_vec.clear();
        alt_id.clear();
        asym_id.clear();
        AA.clear();
    }

    if (compress_type) fin_gz.close();
    else               fin.close();
    line.clear();
    return PDB_lines.size();
}

void parse_chain_list(const vector<string>&chain_list,
    vector<vector<vector<double> > >&a_vec, vector<vector<char> >&seq_vec,
    vector<vector<char> >&sec_vec, vector<int>&mol_vec, vector<int>&len_vec,
    vector<string>&chainID_list, const int ter_opt, const int split_opt,
    const string mol_opt, const int infmt_opt, const string atom_opt,
    const int mirror_opt, const int het_opt, int &len_aa, int &len_na,  
    const int o_opt, vector<string>&resi_vec)
{
    size_t i;
    int chain_i,r;
    string name;
    int chainnum;
    double **xa;
    int len;
    char *seq,*sec;

    vector<vector<string> >PDB_lines;
    vector<double> tmp_atom_array(3,0);
    vector<vector<double> > tmp_chain_array;
    vector<char>tmp_seq_array;
    vector<char>tmp_sec_array;
    //vector<string> resi_vec;
    int read_resi=0;
    if (o_opt) read_resi=2;

    for (i=0;i<chain_list.size();i++)
    {
        name=chain_list[i];
        chainnum=get_PDB_lines(name, PDB_lines, chainID_list,
            mol_vec, ter_opt, infmt_opt, atom_opt, split_opt, het_opt);
        if (!chainnum)
        {
            cerr<<"Warning! Cannot parse file: "<<name
                <<". Chain number 0."<<endl;
            continue;
        }
        for (chain_i=0;chain_i<chainnum;chain_i++)
        {
            len=PDB_lines[chain_i].size();
            if (!len)
            {
                cerr<<"Warning! Cannot parse file: "<<name
                    <<". Chain length 0."<<endl;
                continue;
            }
            else if (len<3)
            {
                cerr<<"Sequence is too short <3!: "<<name<<endl;
                continue;
            }
            NewArray(&xa, len, 3);
            seq = new char[len + 1];
            sec = new char[len + 1];
            len = read_PDB(PDB_lines[chain_i], xa, seq, resi_vec, read_resi);
            if (mirror_opt) for (r=0;r<len;r++) xa[r][2]=-xa[r][2];
            if (mol_vec[chain_i]>0 || mol_opt=="RNA")
                make_sec(seq, xa, len, sec,atom_opt);
            else make_sec(xa, len, sec); // secondary structure assignment
            
            /* store in vector */
            tmp_chain_array.assign(len,tmp_atom_array);
            vector<char>tmp_seq_array(len+1,0);
            vector<char>tmp_sec_array(len+1,0);
            for (r=0;r<len;r++)
            {
                tmp_chain_array[r][0]=xa[r][0];
                tmp_chain_array[r][1]=xa[r][1];
                tmp_chain_array[r][2]=xa[r][2];
                tmp_seq_array[r]=seq[r];
                tmp_sec_array[r]=sec[r];
            }
            a_vec.push_back(tmp_chain_array);
            seq_vec.push_back(tmp_seq_array);
            sec_vec.push_back(tmp_sec_array);
            len_vec.push_back(len);

            /* clean up */
            tmp_chain_array.clear();
            tmp_seq_array.clear();
            tmp_sec_array.clear();
            PDB_lines[chain_i].clear();
            DeleteArray(&xa, len);
            delete [] seq;
            delete [] sec;
        } // chain_i
        name.clear();
        PDB_lines.clear();
        mol_vec.clear();
    } // i
    tmp_atom_array.clear();

    if (mol_opt=="RNA") mol_vec.assign(a_vec.size(),1);
    else if (mol_opt=="protein") mol_vec.assign(a_vec.size(),-1);
    else
    {
        mol_vec.assign(a_vec.size(),0);
        for (i=0;i<a_vec.size();i++)
        {
            for (r=0;r<len_vec[i];r++)
            {
                if (seq_vec[i][r]>='a' && seq_vec[i][r]<='z') mol_vec[i]++;
                else mol_vec[i]--;
            }
        }
    }

    len_aa=0;
    len_na=0;
    for (i=0;i<a_vec.size();i++)
    {
        if (mol_vec[i]>0) len_na+=len_vec[i];
        else              len_aa+=len_vec[i];
    }
}

int copy_chain_pair_data(
    const vector<vector<vector<double> > >&xa_vec,
    const vector<vector<vector<double> > >&ya_vec,
    const vector<vector<char> >&seqx_vec, const vector<vector<char> >&seqy_vec,
    const vector<vector<char> >&secx_vec, const vector<vector<char> >&secy_vec,
    const vector<int> &mol_vec1, const vector<int> &mol_vec2,
    const vector<int> &xlen_vec, const vector<int> &ylen_vec,
    double **xa, double **ya, char *seqx, char *seqy, char *secx, char *secy,
    int chain1_num, int chain2_num,
    vector<vector<string> >&seqxA_mat, vector<vector<string> >&seqyA_mat,
    int *assign1_list, int *assign2_list, vector<string>&sequence)
{
    int i,j,r;
    for (i=0;i<sequence.size();i++) sequence[i].clear();
    sequence.clear();
    sequence.push_back("");
    sequence.push_back("");
    int mol_type=0;
    int xlen=0;
    int ylen=0;
    for (i=0;i<chain1_num;i++)
    {
        j=assign1_list[i];
        if (j<0) continue;
        for (r=0;r<xlen_vec[i];r++)
        {
            seqx[xlen]=seqx_vec[i][r];
            secx[xlen]=secx_vec[i][r];
            xa[xlen][0]= xa_vec[i][r][0];
            xa[xlen][1]= xa_vec[i][r][1];
            xa[xlen][2]= xa_vec[i][r][2];
            xlen++;
        }
        sequence[0]+=seqxA_mat[i][j];
        for (r=0;r<ylen_vec[j];r++)
        {
            seqy[ylen]=seqy_vec[j][r];
            secy[ylen]=secy_vec[j][r];
            ya[ylen][0]= ya_vec[j][r][0];
            ya[ylen][1]= ya_vec[j][r][1];
            ya[ylen][2]= ya_vec[j][r][2];
            ylen++;
        }
        sequence[1]+=seqyA_mat[i][j];
        mol_type+=mol_vec1[i]+mol_vec2[j];
    }
    seqx[xlen]=0;
    secx[xlen]=0;
    seqy[ylen]=0;
    secy[ylen]=0;
    return mol_type;
}

double MMalign_search(
    const vector<vector<vector<double> > >&xa_vec,
    const vector<vector<vector<double> > >&ya_vec,
    const vector<vector<char> >&seqx_vec, const vector<vector<char> >&seqy_vec,
    const vector<vector<char> >&secx_vec, const vector<vector<char> >&secy_vec,
    const vector<int> &mol_vec1, const vector<int> &mol_vec2,
    const vector<int> &xlen_vec, const vector<int> &ylen_vec,
    double **xa, double **ya, char *seqx, char *seqy, char *secx, char *secy,
    int len_aa, int len_na, int chain1_num, int chain2_num, double **TMave_mat,
    vector<vector<string> >&seqxA_mat, vector<vector<string> >&seqyA_mat,
    int *assign1_list, int *assign2_list, vector<string>&sequence,
    double d0_scale, bool fast_opt, const int i_opt=3)
{
    double total_score=0;
    int i,j;
    int xlen=0;
    int ylen=0;
    for (i=0;i<chain1_num;i++)
    {
        if (assign1_list[i]<0) continue;
        xlen+=xlen_vec[i];
        ylen+=ylen_vec[assign1_list[i]];
    }
    if (xlen<=3 || ylen<=3) return total_score;

    seqx = new char[xlen+1];
    secx = new char[xlen+1];
    NewArray(&xa, xlen, 3);
    seqy = new char[ylen+1];
    secy = new char[ylen+1];
    NewArray(&ya, ylen, 3);

    int mol_type=copy_chain_pair_data(xa_vec, ya_vec, seqx_vec, seqy_vec,
        secx_vec, secy_vec, mol_vec1, mol_vec2, xlen_vec, ylen_vec,
        xa, ya, seqx, seqy, secx, secy, chain1_num, chain2_num,
        seqxA_mat, seqyA_mat, assign1_list, assign2_list, sequence);

    /* declare variable specific to this pair of TMalign */
    double t0[3], u0[3][3];
    double TM1, TM2;
    double TM3, TM4, TM5;     // for a_opt, u_opt, d_opt
    double d0_0, TM_0;
    double d0A, d0B, d0u, d0a;
    double d0_out=5.0;
    string seqM, seqxA, seqyA;// for output alignment
    double rmsd0 = 0.0;
    int L_ali;                // Aligned length in standard_TMscore
    double Liden=0;
    double TM_ali, rmsd_ali;  // TMscore and rmsd in standard_TMscore
    int n_ali=0;
    int n_ali8=0;

    double Lnorm_ass=len_aa+len_na;

    /* entry function for structure alignment */
    TMalign_main(xa, ya, seqx, seqy, secx, secy,
        t0, u0, TM1, TM2, TM3, TM4, TM5,
        d0_0, TM_0, d0A, d0B, d0u, d0a, d0_out, seqM, seqxA, seqyA,
        rmsd0, L_ali, Liden, TM_ali, rmsd_ali, n_ali, n_ali8,
        xlen, ylen, sequence, Lnorm_ass, d0_scale,
        i_opt, false, true, false, fast_opt, mol_type, -1);

    /* clean up */
    delete [] seqx;
    delete [] seqy;
    delete [] secx;
    delete [] secy;
    DeleteArray(&xa,xlen);
    DeleteArray(&ya,ylen);

    /* re-compute chain level alignment */
    for (i=0;i<chain1_num;i++)
    {
        xlen=xlen_vec[i];
        if (xlen<3)
        {
            for (j=0;j<chain2_num;j++) TMave_mat[i][j]=-1;
            continue;
        }
        seqx = new char[xlen+1];
        secx = new char[xlen+1];
        NewArray(&xa, xlen, 3);
        copy_chain_data(xa_vec[i],seqx_vec[i],secx_vec[i],
            xlen,xa,seqx,secx);

        double **xt;
        NewArray(&xt, xlen, 3);
        do_rotation(xa, xt, xlen, t0, u0);

        for (j=0;j<chain2_num;j++)
        {
            if (mol_vec1[i]*mol_vec2[j]<0) //no protein-RNA alignment
            {
                TMave_mat[i][j]=-1;
                continue;
            }

            ylen=ylen_vec[j];
            if (ylen<3)
            {
                TMave_mat[i][j]=-1;
                continue;
            }
            seqy = new char[ylen+1];
            secy = new char[ylen+1];
            NewArray(&ya, ylen, 3);
            copy_chain_data(ya_vec[j],seqy_vec[j],secy_vec[j],
                ylen,ya,seqy,secy);

            /* declare variable specific to this pair of TMalign */
            d0_out=5.0;
            seqM.clear();
            seqxA.clear();
            seqyA.clear();
            rmsd0 = 0.0;
            Liden=0;
            int *invmap = new int[ylen+1];

            double Lnorm_ass=len_aa;
            if (mol_vec1[i]+mol_vec2[j]>0) Lnorm_ass=len_na;

            /* entry function for structure alignment */
            se_main(xt, ya, seqx, seqy, TM1, TM2, TM3, TM4, TM5,
                d0_0, TM_0, d0A, d0B, d0u, d0a, d0_out, seqM, seqxA, seqyA,
                rmsd0, L_ali, Liden, TM_ali, rmsd_ali, n_ali, n_ali8,
                xlen, ylen, sequence, Lnorm_ass, d0_scale,
                0, false, 2, false, mol_vec1[i]+mol_vec2[j], 1, invmap);

            /* print result */
            seqxA_mat[i][j]=seqxA;
            seqyA_mat[i][j]=seqyA;

            TMave_mat[i][j]=TM4*Lnorm_ass;
            if (assign1_list[i]==j) total_score+=TMave_mat[i][j];

            /* clean up */
            seqM.clear();
            seqxA.clear();
            seqyA.clear();

            delete[]seqy;
            delete[]secy;
            DeleteArray(&ya,ylen);
            delete[]invmap;
        }
        delete[]seqx;
        delete[]secx;
        DeleteArray(&xa,xlen);
        DeleteArray(&xt,xlen);
    }
    return total_score;
}

void MMalign_final(
    const string xname, const string yname,
    const vector<string> chainID_list1, const vector<string> chainID_list2,
    string fname_super, string fname_lign, string fname_matrix,
    const vector<vector<vector<double> > >&xa_vec,
    const vector<vector<vector<double> > >&ya_vec,
    const vector<vector<char> >&seqx_vec, const vector<vector<char> >&seqy_vec,
    const vector<vector<char> >&secx_vec, const vector<vector<char> >&secy_vec,
    const vector<int> &mol_vec1, const vector<int> &mol_vec2,
    const vector<int> &xlen_vec, const vector<int> &ylen_vec,
    double **xa, double **ya, char *seqx, char *seqy, char *secx, char *secy,
    int len_aa, int len_na, int chain1_num, int chain2_num,
    double **TMave_mat,
    vector<vector<string> >&seqxA_mat, vector<vector<string> >&seqM_mat,
    vector<vector<string> >&seqyA_mat, int *assign1_list, int *assign2_list,
    vector<string>&sequence, const double d0_scale, const bool m_opt,
    const int o_opt, const int outfmt_opt, const int ter_opt,
    const int split_opt, const bool a_opt, const bool d_opt,
    const bool fast_opt, const bool full_opt, const int mirror_opt,
    const vector<string>&resi_vec1, const vector<string>&resi_vec2)
{
    int i,j;
    int xlen=0;
    int ylen=0;
    for (i=0;i<chain1_num;i++) xlen+=xlen_vec[i];
    for (j=0;j<chain2_num;j++) ylen+=ylen_vec[j];
    if (xlen<=3 || ylen<=3) return;

    seqx = new char[xlen+1];
    secx = new char[xlen+1];
    NewArray(&xa, xlen, 3);
    seqy = new char[ylen+1];
    secy = new char[ylen+1];
    NewArray(&ya, ylen, 3);

    int mol_type=copy_chain_pair_data(xa_vec, ya_vec, seqx_vec, seqy_vec,
        secx_vec, secy_vec, mol_vec1, mol_vec2, xlen_vec, ylen_vec,
        xa, ya, seqx, seqy, secx, secy, chain1_num, chain2_num,
        seqxA_mat, seqyA_mat, assign1_list, assign2_list, sequence);

    /* declare variable specific to this pair of TMalign */
    double t0[3], u0[3][3];
    double TM1, TM2;
    double TM3, TM4, TM5;     // for a_opt, u_opt, d_opt
    double d0_0, TM_0;
    double d0A, d0B, d0u, d0a;
    double d0_out=5.0;
    string seqM, seqxA, seqyA;// for output alignment
    double rmsd0 = 0.0;
    int L_ali;                // Aligned length in standard_TMscore
    double Liden=0;
    double TM_ali, rmsd_ali;  // TMscore and rmsd in standard_TMscore
    int n_ali=0;
    int n_ali8=0;

    double Lnorm_ass=len_aa+len_na;

    /* entry function for structure alignment */
    TMalign_main(xa, ya, seqx, seqy, secx, secy,
        t0, u0, TM1, TM2, TM3, TM4, TM5,
        d0_0, TM_0, d0A, d0B, d0u, d0a, d0_out, seqM, seqxA, seqyA,
        rmsd0, L_ali, Liden, TM_ali, rmsd_ali, n_ali, n_ali8,
        xlen, ylen, sequence, Lnorm_ass, d0_scale,
        3, a_opt, false, d_opt, fast_opt, mol_type, -1);

    /* prepare full complex alignment */
    string chainID1="";
    string chainID2="";
    sequence.clear();
    sequence.push_back(""); // seqxA
    sequence.push_back(""); // seqyA
    sequence.push_back(""); // seqM
    int aln_start=0;
    int aln_end=0;
    for (i=0;i<chain1_num;i++)
    {
        j=assign1_list[i];
        if (j<0) continue;
        chainID1+=chainID_list1[i];
        chainID2+=chainID_list2[j];
        sequence[0]+=seqxA_mat[i][j]+'*';
        sequence[1]+=seqyA_mat[i][j]+'*';

        aln_end+=seqxA_mat[i][j].size();
        seqM_mat[i][j]=seqM.substr(aln_start,aln_end-aln_start);
        sequence[2]+=seqM_mat[i][j]+'*';
        aln_start=aln_end;
    }

    /* prepare unaligned region */
    for (i=0;i<chain1_num;i++)
    {
        if (assign1_list[i]>=0) continue;
        chainID1+=chainID_list1[i];
        chainID2+=':';
        string s(seqx_vec[i].begin(),seqx_vec[i].end());
        sequence[0]+=s.substr(0,xlen_vec[i])+'*';
        sequence[1]+=string(xlen_vec[i],'-')+'*';
        s.clear();
        sequence[2]+=string(xlen_vec[i],' ')+'*';
    }
    for (j=0;j<chain2_num;j++)
    {
        if (assign2_list[j]>=0) continue;
        chainID1+=':';
        chainID2+=chainID_list2[j];
        string s(seqy_vec[j].begin(),seqy_vec[j].end());
        sequence[0]+=string(ylen_vec[j],'-')+'*';
        sequence[1]+=s.substr(0,ylen_vec[j])+'*';
        s.clear();
        sequence[2]+=string(ylen_vec[j],' ')+'*';
    }

    /* print alignment */
    output_results(xname, yname, chainID1.c_str(), chainID2.c_str(),
        xlen, ylen, t0, u0, TM1, TM2, TM3, TM4, TM5, rmsd0, d0_out,
        sequence[2].c_str(), sequence[0].c_str(), sequence[1].c_str(),
        Liden, n_ali8, L_ali, TM_ali, rmsd_ali,
        TM_0, d0_0, d0A, d0B, 0, d0_scale, d0a, d0u, 
        (m_opt?fname_matrix:"").c_str(), outfmt_opt, ter_opt, true,
        split_opt, o_opt, fname_super,
        false, a_opt, false, d_opt, mirror_opt, resi_vec1, resi_vec2);

    /* clean up */
    seqM.clear();
    seqxA.clear();
    seqyA.clear();
    delete [] seqx;
    delete [] seqy;
    delete [] secx;
    delete [] secy;
    DeleteArray(&xa,xlen);
    DeleteArray(&ya,ylen);
    sequence[0].clear();
    sequence[1].clear();
    sequence[2].clear();

    if (!full_opt) return;

    cout<<"# End of alignment for full complex. The following blocks list alignments for individual chains."<<endl;

    /* re-compute chain level alignment */
    for (i=0;i<chain1_num;i++)
    {
        j=assign1_list[i];
        if (j<0) continue;
        xlen=xlen_vec[i];
        seqx = new char[xlen+1];
        secx = new char[xlen+1];
        NewArray(&xa, xlen, 3);
        copy_chain_data(xa_vec[i],seqx_vec[i],secx_vec[i],
            xlen,xa,seqx,secx);

        double **xt;
        NewArray(&xt, xlen, 3);
        do_rotation(xa, xt, xlen, t0, u0);

        ylen=ylen_vec[j];
        if (ylen<3)
        {
            TMave_mat[i][j]=-1;
            continue;
        }
        seqy = new char[ylen+1];
        secy = new char[ylen+1];
        NewArray(&ya, ylen, 3);
        copy_chain_data(ya_vec[j],seqy_vec[j],secy_vec[j],
            ylen,ya,seqy,secy);

        /* declare variable specific to this pair of TMalign */
        d0_out=5.0;
        rmsd0 = 0.0;
        Liden=0;
        int *invmap = new int[ylen+1];
        seqM="";
        seqxA="";
        seqyA="";
        double Lnorm_ass=len_aa;
        if (mol_vec1[i]+mol_vec2[j]>0) Lnorm_ass=len_na;
        sequence[0]=seqxA_mat[i][j];
        sequence[1]=seqyA_mat[i][j];

        /* entry function for structure alignment */
        se_main(xt, ya, seqx, seqy, TM1, TM2, TM3, TM4, TM5,
            d0_0, TM_0, d0A, d0B, d0u, d0a, d0_out, seqM, seqxA, seqyA,
            rmsd0, L_ali, Liden, TM_ali, rmsd_ali, n_ali, n_ali8,
            xlen, ylen, sequence, Lnorm_ass, d0_scale,
            1, a_opt, 2, d_opt, mol_vec1[i]+mol_vec2[j], 1, invmap);

        //TM2=TM4*Lnorm_ass/xlen;
        //TM1=TM4*Lnorm_ass/ylen;
        //d0A=d0u;
        //d0B=d0u;

        /* print result */
        output_results(xname, yname,
            chainID_list1[i].c_str(), chainID_list2[j].c_str(),
            xlen, ylen, t0, u0, TM1, TM2, TM3, TM4, TM5, rmsd0, d0_out,
            seqM_mat[i][j].c_str(), seqxA_mat[i][j].c_str(),
            seqyA_mat[i][j].c_str(), Liden, n_ali8, L_ali, TM_ali, rmsd_ali,
            TM_0, d0_0, d0A, d0B, Lnorm_ass, d0_scale, d0a, d0u, 
            "", outfmt_opt, ter_opt, false, split_opt, 0,
            "", false, a_opt, false, d_opt, 0, resi_vec1, resi_vec2);

        /* clean up */
        seqxA.clear();
        seqM.clear();
        seqyA.clear();
        sequence[0].clear();
        sequence[1].clear();
        delete[]seqy;
        delete[]secy;
        DeleteArray(&ya,ylen);
        delete[]seqx;
        delete[]secx;
        DeleteArray(&xa,xlen);
        DeleteArray(&xt,xlen);
        delete[]invmap;
    }
    sequence.clear();
    return;
}

void copy_chain_assign_data(int chain1_num, int chain2_num, 
    vector<string> &sequence,
    vector<vector<string> >&seqxA_mat, vector<vector<string> >&seqyA_mat,
    int *assign1_list, int *assign2_list, double **TMave_mat,
    vector<vector<string> >&seqxA_tmp, vector<vector<string> >&seqyA_tmp,
    int *assign1_tmp,  int *assign2_tmp,  double **TMave_tmp)
{
    int i,j;
    for (i=0;i<sequence.size();i++) sequence[i].clear();
    sequence.clear();
    sequence.push_back("");
    sequence.push_back("");
    for (i=0;i<chain1_num;i++) assign1_tmp[i]=assign1_list[i];
    for (i=0;i<chain2_num;i++) assign2_tmp[i]=assign2_list[i];
    for (i=0;i<chain1_num;i++)
    {
        for (j=0;j<chain2_num;j++)
        {
            seqxA_tmp[i][j]=seqxA_mat[i][j];
            seqyA_tmp[i][j]=seqyA_mat[i][j];
            TMave_tmp[i][j]=TMave_mat[i][j];
            if (assign1_list[i]==j)
            {
                sequence[0]+=seqxA_mat[i][j];
                sequence[1]+=seqyA_mat[i][j];
            }
        }
    }
    return;
}

void MMalign_iter(double & max_total_score, const int max_iter,
    const vector<vector<vector<double> > >&xa_vec,
    const vector<vector<vector<double> > >&ya_vec,
    const vector<vector<char> >&seqx_vec, const vector<vector<char> >&seqy_vec,
    const vector<vector<char> >&secx_vec, const vector<vector<char> >&secy_vec,
    const vector<int> &mol_vec1, const vector<int> &mol_vec2,
    const vector<int> &xlen_vec, const vector<int> &ylen_vec,
    double **xa, double **ya, char *seqx, char *seqy, char *secx, char *secy,
    int len_aa, int len_na, int chain1_num, int chain2_num, double **TMave_mat,
    vector<vector<string> >&seqxA_mat, vector<vector<string> >&seqyA_mat,
    int *assign1_list, int *assign2_list, vector<string>&sequence,
    double d0_scale, bool fast_opt)
{
    /* tmp assignment */
    double total_score;
    int *assign1_tmp, *assign2_tmp;
    assign1_tmp=new int[chain1_num];
    assign2_tmp=new int[chain2_num];
    double **TMave_tmp;
    NewArray(&TMave_tmp,chain1_num,chain2_num);
    vector<string> tmp_str_vec(chain2_num,"");
    vector<vector<string> >seqxA_tmp(chain1_num,tmp_str_vec);
    vector<vector<string> >seqyA_tmp(chain1_num,tmp_str_vec);
    vector<string> sequence_tmp;
    copy_chain_assign_data(chain1_num, chain2_num, sequence_tmp,
        seqxA_mat, seqyA_mat, assign1_list, assign2_list, TMave_mat,
        seqxA_tmp, seqyA_tmp, assign1_tmp,  assign2_tmp,  TMave_tmp);
    
    for (int iter=0;iter<max_iter;iter++)
    {
        total_score=MMalign_search(xa_vec, ya_vec, seqx_vec, seqy_vec,
            secx_vec, secy_vec, mol_vec1, mol_vec2, xlen_vec, ylen_vec,
            xa, ya, seqx, seqy, secx, secy, len_aa, len_na,
            chain1_num, chain2_num, 
            TMave_tmp, seqxA_tmp, seqyA_tmp, assign1_tmp, assign2_tmp,
            sequence, d0_scale, fast_opt);
        total_score=enhanced_greedy_search(TMave_tmp, assign1_tmp,
            assign2_tmp, chain1_num, chain2_num);
        //if (total_score<=0) PrintErrorAndQuit("ERROR! No assignable chain");
        if (total_score<=max_total_score) break;
        max_total_score=total_score;
        copy_chain_assign_data(chain1_num, chain2_num, sequence,
            seqxA_tmp, seqyA_tmp, assign1_tmp,  assign2_tmp,  TMave_tmp,
            seqxA_mat, seqyA_mat, assign1_list, assign2_list, TMave_mat);
    }

    /* clean up everything */
    delete [] assign1_tmp;
    delete [] assign2_tmp;
    DeleteArray(&TMave_tmp,chain1_num);
    vector<string>().swap(tmp_str_vec);
    vector<vector<string> >().swap(seqxA_tmp);
    vector<vector<string> >().swap(seqyA_tmp);
}


/* Input: vectors x, y, rotation matrix t, u, scale factor d02, and gap_open
 * Output: j2i[1:len2] \in {1:len1} U {-1}
 * path[0:len1, 0:len2]=1,2,3, from diagonal, horizontal, vertical */
void NWDP_TM_dimer(bool **path, double **val, double **x, double **y,
    int len1, int len2, bool **mask,
    double t[3], double u[3][3], double d02, double gap_open, int j2i[])
{
    int i, j;
    double h, v, d;

    //initialization
    for(i=0; i<=len1; i++)
    {
        //val[i][0]=0;
        val[i][0]=i*gap_open;
        path[i][0]=false; //not from diagonal
    }

    for(j=0; j<=len2; j++)
    {
        //val[0][j]=0;
        val[0][j]=j*gap_open;
        path[0][j]=false; //not from diagonal
        j2i[j]=-1;    //all are not aligned, only use j2i[1:len2]
    }      
    double xx[3], dij;


    //decide matrix and path
    for(i=1; i<=len1; i++)
    {
        transform(t, u, &x[i-1][0], xx);
        for(j=1; j<=len2; j++)
        {
            d=FLT_MIN;
            if (mask[i][j])
            {
                dij=dist(xx, &y[j-1][0]);    
                d=val[i-1][j-1] +  1.0/(1+dij/d02);
            } 

            //symbol insertion in horizontal (= a gap in vertical)
            h=val[i-1][j];
            if(path[i-1][j]) h += gap_open; //aligned in last position

            //symbol insertion in vertical
            v=val[i][j-1];
            if(path[i][j-1]) v += gap_open; //aligned in last position


            if(d>=h && d>=v)
            {
                path[i][j]=true; //from diagonal
                val[i][j]=d;
            }
            else 
            {
                path[i][j]=false; //from horizontal
                if(v>=h) val[i][j]=v;
                else val[i][j]=h;
            }
        } //for i
    } //for j

    //trace back to extract the alignment
    i=len1;
    j=len2;
    while(i>0 && j>0)
    {
        if(path[i][j]) //from diagonal
        {
            j2i[j-1]=i-1;
            i--;
            j--;
        }
        else 
        {
            h=val[i-1][j];
            if(path[i-1][j]) h +=gap_open;

            v=val[i][j-1];
            if(path[i][j-1]) v +=gap_open;

            if(v>=h) j--;
            else i--;
        }
    }
}

/* +ss
 * Input: secondary structure secx, secy, and gap_open
 * Output: j2i[1:len2] \in {1:len1} U {-1}
 * path[0:len1, 0:len2]=1,2,3, from diagonal, horizontal, vertical */
void NWDP_TM_dimer(bool **path, double **val, const char *secx, const char *secy,
    const int len1, const int len2, bool **mask, const double gap_open, int j2i[])
{

    int i, j;
    double h, v, d;

    //initialization
    for(i=0; i<=len1; i++)
    {
        //val[i][0]=0;
        val[i][0]=i*gap_open;
        path[i][0]=false; //not from diagonal
    }

    for(j=0; j<=len2; j++)
    {
        //val[0][j]=0;
        val[0][j]=j*gap_open;
        path[0][j]=false; //not from diagonal
        j2i[j]=-1;    //all are not aligned, only use j2i[1:len2]
    }      

    //decide matrix and path
    for(i=1; i<=len1; i++)
    {
        for(j=1; j<=len2; j++)
        {
            d=FLT_MIN;
            if (mask[i][j])
                d=val[i-1][j-1] + 1.0*(secx[i-1]==secy[j-1]);

            //symbol insertion in horizontal (= a gap in vertical)
            h=val[i-1][j];
            if(path[i-1][j]) h += gap_open; //aligned in last position

            //symbol insertion in vertical
            v=val[i][j-1];
            if(path[i][j-1]) v += gap_open; //aligned in last position

            if(d>=h && d>=v)
            {
                path[i][j]=true; //from diagonal
                val[i][j]=d;
            }
            else 
            {
                path[i][j]=false; //from horizontal
                if(v>=h) val[i][j]=v;
                else val[i][j]=h;
            }
        } //for i
    } //for j

    //trace back to extract the alignment
    i=len1;
    j=len2;
    while(i>0 && j>0)
    {
        if(path[i][j]) //from diagonal
        {
            j2i[j-1]=i-1;
            i--;
            j--;
        }
        else 
        {
            h=val[i-1][j];
            if(path[i-1][j]) h +=gap_open;

            v=val[i][j-1];
            if(path[i][j-1]) v +=gap_open;

            if(v>=h) j--;
            else i--;
        }
    }
}

//heuristic run of dynamic programing iteratively to find the best alignment
//input: initial rotation matrix t, u
//       vectors x and y, d0
//output: best alignment that maximizes the TMscore, will be stored in invmap
double DP_iter_dimer(double **r1, double **r2, double **xtm, double **ytm,
    double **xt, bool **path, double **val, double **x, double **y,
    int xlen, int ylen, bool **mask, double t[3], double u[3][3], int invmap0[],
    int g1, int g2, int iteration_max, double local_d0_search,
    double D0_MIN, double Lnorm, double d0, double score_d8)
{
    double gap_open[2]={-0.6, 0};
    double rmsd; 
    int *invmap=new int[ylen+1];
    
    int iteration, i, j, k;
    double tmscore, tmscore_max, tmscore_old=0;    
    int score_sum_method=8, simplify_step=40;
    tmscore_max=-1;

    //double d01=d0+1.5;
    double d02=d0*d0;
    for(int g=g1; g<g2; g++)
    {
        for(iteration=0; iteration<iteration_max; iteration++)
        {           
            NWDP_TM_dimer(path, val, x, y, xlen, ylen, mask,
                t, u, d02, gap_open[g], invmap);
            
            k=0;
            for(j=0; j<ylen; j++) 
            {
                i=invmap[j];

                if(i>=0) //aligned
                {
                    xtm[k][0]=x[i][0];
                    xtm[k][1]=x[i][1];
                    xtm[k][2]=x[i][2];
                    
                    ytm[k][0]=y[j][0];
                    ytm[k][1]=y[j][1];
                    ytm[k][2]=y[j][2];
                    k++;
                }
            }

            tmscore = TMscore8_search(r1, r2, xtm, ytm, xt, k, t, u,
                simplify_step, score_sum_method, &rmsd, local_d0_search,
                Lnorm, score_d8, d0);

           
            if(tmscore>tmscore_max)
            {
                tmscore_max=tmscore;
                for(i=0; i<ylen; i++) invmap0[i]=invmap[i];
            }
    
            if(iteration>0)
            {
                if(fabs(tmscore_old-tmscore)<0.000001) break;       
            }
            tmscore_old=tmscore;
        }// for iteration           
        
    }//for gapopen
    
    
    delete []invmap;
    return tmscore_max;
}

void get_initial_ss_dimer(bool **path, double **val, const char *secx,
    const char *secy, int xlen, int ylen, bool **mask, int *y2x)
{
    double gap_open=-1.0;
    NWDP_TM_dimer(path, val, secx, secy, xlen, ylen, mask, gap_open, y2x);
}

bool get_initial5_dimer( double **r1, double **r2, double **xtm, double **ytm,
    bool **path, double **val, double **x, double **y, int xlen, int ylen,
    bool **mask, int *y2x,
    double d0, double d0_search, const bool fast_opt, const double D0_MIN)
{
    double GL, rmsd;
    double t[3];
    double u[3][3];

    double d01 = d0 + 1.5;
    if (d01 < D0_MIN) d01 = D0_MIN;
    double d02 = d01*d01;

    double GLmax = 0;
    int aL = getmin(xlen, ylen);
    int *invmap = new int[ylen + 1];

    // jump on sequence1-------------->
    int n_jump1 = 0;
    if (xlen > 250)
        n_jump1 = 45;
    else if (xlen > 200)
        n_jump1 = 35;
    else if (xlen > 150)
        n_jump1 = 25;
    else
        n_jump1 = 15;
    if (n_jump1 > (xlen / 3))
        n_jump1 = xlen / 3;

    // jump on sequence2-------------->
    int n_jump2 = 0;
    if (ylen > 250)
        n_jump2 = 45;
    else if (ylen > 200)
        n_jump2 = 35;
    else if (ylen > 150)
        n_jump2 = 25;
    else
        n_jump2 = 15;
    if (n_jump2 > (ylen / 3))
        n_jump2 = ylen / 3;

    // fragment to superimpose-------------->
    int n_frag[2] = { 20, 100 };
    if (n_frag[0] > (aL / 3))
        n_frag[0] = aL / 3;
    if (n_frag[1] > (aL / 2))
        n_frag[1] = aL / 2;

    // start superimpose search-------------->
    if (fast_opt)
    {
        n_jump1*=5;
        n_jump2*=5;
    }
    bool flag = false;
    for (int i_frag = 0; i_frag < 2; i_frag++)
    {
        int m1 = xlen - n_frag[i_frag] + 1;
        int m2 = ylen - n_frag[i_frag] + 1;

        for (int i = 0; i<m1; i = i + n_jump1) //index starts from 0, different from FORTRAN
        {
            for (int j = 0; j<m2; j = j + n_jump2)
            {
                for (int k = 0; k<n_frag[i_frag]; k++) //fragment in y
                {
                    r1[k][0] = x[k + i][0];
                    r1[k][1] = x[k + i][1];
                    r1[k][2] = x[k + i][2];

                    r2[k][0] = y[k + j][0];
                    r2[k][1] = y[k + j][1];
                    r2[k][2] = y[k + j][2];
                }

                // superpose the two structures and rotate it
                Kabsch(r1, r2, n_frag[i_frag], 1, &rmsd, t, u);

                double gap_open = 0.0;
                NWDP_TM_dimer(path, val, x, y, xlen, ylen, mask,
                    t, u, d02, gap_open, invmap);
                GL = get_score_fast(r1, r2, xtm, ytm, x, y, xlen, ylen,
                    invmap, d0, d0_search, t, u);
                if (GL>GLmax)
                {
                    GLmax = GL;
                    for (int ii = 0; ii<ylen; ii++) y2x[ii] = invmap[ii];
                    flag = true;
                }
            }
        }
    }

    delete[] invmap;
    return flag;
}

void get_initial_ssplus_dimer(double **r1, double **r2, double **score,
    bool **path, double **val, const char *secx, const char *secy,
    double **x, double **y, int xlen, int ylen, bool **mask,
    int *y2x0, int *y2x, const double D0_MIN, double d0)
{
    //create score matrix for DP
    score_matrix_rmsd_sec(r1, r2, score, secx, secy, x, y, xlen, ylen,
        y2x0, D0_MIN,d0);

    int i,j;
    for (i=0;i<xlen+1;i++) for (j=0;j<ylen+1;j++) score[i][j]=FLT_MIN;
    
    double gap_open=-1.0;
    NWDP_TM(score, path, val, xlen, ylen, gap_open, y2x);
}

/* Entry function for TM-align. Return TM-score calculation status:
 * 0   - full TM-score calculation 
 * 1   - terminated due to exception
 * 2-7 - pre-terminated due to low TM-score */
int TMalign_dimer_main(double **xa, double **ya,
    const char *seqx, const char *seqy, const char *secx, const char *secy,
    double t0[3], double u0[3][3],
    double &TM1, double &TM2, double &TM3, double &TM4, double &TM5,
    double &d0_0, double &TM_0,
    double &d0A, double &d0B, double &d0u, double &d0a, double &d0_out,
    string &seqM, string &seqxA, string &seqyA,
    double &rmsd0, int &L_ali, double &Liden,
    double &TM_ali, double &rmsd_ali, int &n_ali, int &n_ali8,
    const int xlen, const int ylen,
    bool **mask,
    const vector<string> sequence, const double Lnorm_ass,
    const double d0_scale, const int i_opt, const int a_opt,
    const bool u_opt, const bool d_opt, const bool fast_opt,
    const int mol_type, const double TMcut=-1)
{
    double D0_MIN;        //for d0
    double Lnorm;         //normalization length
    double score_d8,d0,d0_search,dcu0;//for TMscore search
    double t[3], u[3][3]; //Kabsch translation vector and rotation matrix
    double **score;       // Input score table for dynamic programming
    bool   **path;        // for dynamic programming  
    double **val;         // for dynamic programming  
    double **xtm, **ytm;  // for TMscore search engine
    double **xt;          //for saving the superposed version of r_1 or xtm
    double **r1, **r2;    // for Kabsch rotation

    /***********************/
    /* allocate memory     */
    /***********************/
    int minlen = min(xlen, ylen);
    NewArray(&score, xlen+1, ylen+1);
    NewArray(&path, xlen+1, ylen+1);
    NewArray(&val, xlen+1, ylen+1);
    NewArray(&xtm, minlen, 3);
    NewArray(&ytm, minlen, 3);
    NewArray(&xt, xlen, 3);
    NewArray(&r1, minlen, 3);
    NewArray(&r2, minlen, 3);

    /***********************/
    /*    parameter set    */
    /***********************/
    parameter_set4search(xlen, ylen, D0_MIN, Lnorm, 
        score_d8, d0, d0_search, dcu0);
    int simplify_step    = 40; //for simplified search engine
    int score_sum_method = 8;  //for scoring method, whether only sum over pairs with dis<score_d8

    int i;
    int *invmap0         = new int[ylen+1];
    int *invmap          = new int[ylen+1];
    double TM, TMmax=-1;
    for(i=0; i<ylen; i++) invmap0[i]=-1;

    double ddcc=0.4;
    if (Lnorm <= 40) ddcc=0.1;   //Lnorm was setted in parameter_set4search
    double local_d0_search = d0_search;

    //************************************************//
    //    get initial alignment from user's input:    //
    //    Stick to the initial alignment              //
    //************************************************//
    bool bAlignStick = false;
    if (i_opt==3)// if input has set parameter for "-I"
    {
        // In the original code, this loop starts from 1, which is
        // incorrect. Fortran starts from 1 but C++ should starts from 0.
        for (int j = 0; j < ylen; j++)// Set aligned position to be "-1"
            invmap[j] = -1;

        int i1 = -1;// in C version, index starts from zero, not from one
        int i2 = -1;
        int L1 = sequence[0].size();
        int L2 = sequence[1].size();
        int L = min(L1, L2);// Get positions for aligned residues
        for (int kk1 = 0; kk1 < L; kk1++)
        {
            if (sequence[0][kk1] != '-') i1++;
            if (sequence[1][kk1] != '-')
            {
                i2++;
                if (i2 >= ylen || i1 >= xlen) kk1 = L;
                else if (sequence[0][kk1] != '-') invmap[i2] = i1;
            }
        }

        //--------------- 2. Align proteins from original alignment
        double prevD0_MIN = D0_MIN;// stored for later use
        int prevLnorm = Lnorm;
        double prevd0 = d0;
        TM_ali = standard_TMscore(r1, r2, xtm, ytm, xt, xa, ya, xlen, ylen,
            invmap, L_ali, rmsd_ali, D0_MIN, Lnorm, d0, d0_search, score_d8,
            t, u, mol_type);
        D0_MIN = prevD0_MIN;
        Lnorm = prevLnorm;
        d0 = prevd0;
        TM = detailed_search_standard(r1, r2, xtm, ytm, xt, xa, ya, xlen, ylen,
            invmap, t, u, 40, 8, local_d0_search, true, Lnorm, score_d8, d0);
        if (TM > TMmax)
        {
            TMmax = TM;
            for (i = 0; i<ylen; i++) invmap0[i] = invmap[i];
        }
        bAlignStick = true;
    }

    /******************************************************/
    /*    get initial alignment with gapless threading    */
    /******************************************************/
    if (!bAlignStick)
    {
        get_initial(r1, r2, xtm, ytm, xa, ya, xlen, ylen, invmap0, d0,
            d0_search, fast_opt, t, u);
        TM = detailed_search(r1, r2, xtm, ytm, xt, xa, ya, xlen, ylen, invmap0,
            t, u, simplify_step, score_sum_method, local_d0_search, Lnorm,
            score_d8, d0);
        if (TM>TMmax) TMmax = TM;
        if (TMcut>0) copy_t_u(t, u, t0, u0);
        //run dynamic programing iteratively to find the best alignment
        TM = DP_iter_dimer(r1, r2, xtm, ytm, xt, path, val, xa, ya, xlen, ylen,
             mask, t, u, invmap, 0, 2, (fast_opt)?2:30,
             local_d0_search, D0_MIN, Lnorm, d0, score_d8);
        if (TM>TMmax)
        {
            TMmax = TM;
            for (int i = 0; i<ylen; i++) invmap0[i] = invmap[i];
            if (TMcut>0) copy_t_u(t, u, t0, u0);
        }

        if (TMcut>0) // pre-terminate if TM-score is too low
        {
            double TMtmp=approx_TM(xlen, ylen, a_opt,
                xa, ya, t0, u0, invmap0, mol_type);

            if (TMtmp<0.5*TMcut)
            {
                TM1=TM2=TM3=TM4=TM5=TMtmp;
                clean_up_after_approx_TM(invmap0, invmap, score, path, val,
                    xtm, ytm, xt, r1, r2, xlen, minlen);
                return 2;
            }
        }

        /************************************************************/
        /*    get initial alignment based on secondary structure    */
        /************************************************************/
        get_initial_ss_dimer(path, val, secx, secy, xlen, ylen, mask, invmap);
        TM = detailed_search(r1, r2, xtm, ytm, xt, xa, ya, xlen, ylen, invmap,
            t, u, simplify_step, score_sum_method, local_d0_search, Lnorm,
            score_d8, d0);
        if (TM>TMmax)
        {
            TMmax = TM;
            for (int i = 0; i<ylen; i++) invmap0[i] = invmap[i];
            if (TMcut>0) copy_t_u(t, u, t0, u0);
        }
        if (TM > TMmax*0.2)
        {
            TM = DP_iter_dimer(r1, r2, xtm, ytm, xt, path, val, xa, ya,
                xlen, ylen, mask, t, u, invmap, 0, 2,
                (fast_opt)?2:30, local_d0_search, D0_MIN, Lnorm, d0, score_d8);
            if (TM>TMmax)
            {
                TMmax = TM;
                for (int i = 0; i<ylen; i++) invmap0[i] = invmap[i];
                if (TMcut>0) copy_t_u(t, u, t0, u0);
            }
        }

        if (TMcut>0) // pre-terminate if TM-score is too low
        {
            double TMtmp=approx_TM(xlen, ylen, a_opt,
                xa, ya, t0, u0, invmap0, mol_type);

            if (TMtmp<0.52*TMcut)
            {
                TM1=TM2=TM3=TM4=TM5=TMtmp;
                clean_up_after_approx_TM(invmap0, invmap, score, path, val,
                    xtm, ytm, xt, r1, r2, xlen, minlen);
                return 3;
            }
        }

        /************************************************************/
        /*    get initial alignment based on local superposition    */
        /************************************************************/
        //=initial5 in original TM-align
        if (get_initial5_dimer( r1, r2, xtm, ytm, path, val, xa, ya,
            xlen, ylen, mask, invmap, d0, d0_search, fast_opt, D0_MIN))
        {
            TM = detailed_search(r1, r2, xtm, ytm, xt, xa, ya, xlen, ylen,
                invmap, t, u, simplify_step, score_sum_method,
                local_d0_search, Lnorm, score_d8, d0);
            if (TM>TMmax)
            {
                TMmax = TM;
                for (int i = 0; i<ylen; i++) invmap0[i] = invmap[i];
                if (TMcut>0) copy_t_u(t, u, t0, u0);
            }
            if (TM > TMmax*ddcc)
            {
                TM = DP_iter_dimer(r1, r2, xtm, ytm, xt, path, val, xa, ya,
                    xlen, ylen, mask, t, u, invmap, 0, 2, 2,
                    local_d0_search, D0_MIN, Lnorm, d0, score_d8);
                if (TM>TMmax)
                {
                    TMmax = TM;
                    for (int i = 0; i<ylen; i++) invmap0[i] = invmap[i];
                    if (TMcut>0) copy_t_u(t, u, t0, u0);
                }
            }
        }
        else
            cerr << "\n\nWarning: initial alignment from local superposition fail!\n\n" << endl;

        if (TMcut>0) // pre-terminate if TM-score is too low
        {
            double TMtmp=approx_TM(xlen, ylen, a_opt,
                xa, ya, t0, u0, invmap0, mol_type);

            if (TMtmp<0.54*TMcut)
            {
                TM1=TM2=TM3=TM4=TM5=TMtmp;
                clean_up_after_approx_TM(invmap0, invmap, score, path, val,
                    xtm, ytm, xt, r1, r2, xlen, minlen);
                return 4;
            }
        }

        /********************************************************************/
        /* get initial alignment by local superposition+secondary structure */
        /********************************************************************/
        //=initial3 in original TM-align
        get_initial_ssplus_dimer(r1, r2, score, path, val, secx, secy, xa, ya,
            xlen, ylen, mask, invmap0, invmap, D0_MIN, d0);
        TM = detailed_search(r1, r2, xtm, ytm, xt, xa, ya, xlen, ylen, invmap,
             t, u, simplify_step, score_sum_method, local_d0_search, Lnorm,
             score_d8, d0);
        if (TM>TMmax)
        {
            TMmax = TM;
            for (i = 0; i<ylen; i++) invmap0[i] = invmap[i];
            if (TMcut>0) copy_t_u(t, u, t0, u0);
        }
        if (TM > TMmax*ddcc)
        {
            TM = DP_iter_dimer(r1, r2, xtm, ytm, xt, path, val, xa, ya,
                xlen, ylen, mask, t, u, invmap, 0, 2,
                (fast_opt)?2:30, local_d0_search, D0_MIN, Lnorm, d0, score_d8);
            if (TM>TMmax)
            {
                TMmax = TM;
                for (i = 0; i<ylen; i++) invmap0[i] = invmap[i];
                if (TMcut>0) copy_t_u(t, u, t0, u0);
            }
        }

        if (TMcut>0) // pre-terminate if TM-score is too low
        {
            double TMtmp=approx_TM(xlen, ylen, a_opt,
                xa, ya, t0, u0, invmap0, mol_type);

            if (TMtmp<0.56*TMcut)
            {
                TM1=TM2=TM3=TM4=TM5=TMtmp;
                clean_up_after_approx_TM(invmap0, invmap, score, path, val,
                    xtm, ytm, xt, r1, r2, xlen, minlen);
                return 5;
            }
        }

        /*******************************************************************/
        /*    get initial alignment based on fragment gapless threading    */
        /*******************************************************************/
        //=initial4 in original TM-align
        get_initial_fgt(r1, r2, xtm, ytm, xa, ya, xlen, ylen,
            invmap, d0, d0_search, dcu0, fast_opt, t, u);
        TM = detailed_search(r1, r2, xtm, ytm, xt, xa, ya, xlen, ylen, invmap,
            t, u, simplify_step, score_sum_method, local_d0_search, Lnorm,
            score_d8, d0);
        if (TM>TMmax)
        {
            TMmax = TM;
            for (i = 0; i<ylen; i++) invmap0[i] = invmap[i];
            if (TMcut>0) copy_t_u(t, u, t0, u0);
        }
        if (TM > TMmax*ddcc)
        {
            TM = DP_iter_dimer(r1, r2, xtm, ytm, xt, path, val, xa, ya,
                xlen, ylen, mask, t, u, invmap, 1, 2, 2,
                local_d0_search, D0_MIN, Lnorm, d0, score_d8);
            if (TM>TMmax)
            {
                TMmax = TM;
                for (i = 0; i<ylen; i++) invmap0[i] = invmap[i];
                if (TMcut>0) copy_t_u(t, u, t0, u0);
            }
        }

        if (TMcut>0) // pre-terminate if TM-score is too low
        {
            double TMtmp=approx_TM(xlen, ylen, a_opt,
                xa, ya, t0, u0, invmap0, mol_type);

            if (TMtmp<0.58*TMcut)
            {
                TM1=TM2=TM3=TM4=TM5=TMtmp;
                clean_up_after_approx_TM(invmap0, invmap, score, path, val,
                    xtm, ytm, xt, r1, r2, xlen, minlen);
                return 6;
            }
        }

        //************************************************//
        //    get initial alignment from user's input:    //
        //************************************************//
        if (i_opt==1)// if input has set parameter for "-i"
        {
            for (int j = 0; j < ylen; j++)// Set aligned position to be "-1"
                invmap[j] = -1;

            int i1 = -1;// in C version, index starts from zero, not from one
            int i2 = -1;
            int L1 = sequence[0].size();
            int L2 = sequence[1].size();
            int L = min(L1, L2);// Get positions for aligned residues
            for (int kk1 = 0; kk1 < L; kk1++)
            {
                if (sequence[0][kk1] != '-')
                    i1++;
                if (sequence[1][kk1] != '-')
                {
                    i2++;
                    if (i2 >= ylen || i1 >= xlen) kk1 = L;
                    else if (sequence[0][kk1] != '-') invmap[i2] = i1;
                }
            }

            //--------------- 2. Align proteins from original alignment
            double prevD0_MIN = D0_MIN;// stored for later use
            int prevLnorm = Lnorm;
            double prevd0 = d0;
            TM_ali = standard_TMscore(r1, r2, xtm, ytm, xt, xa, ya,
                xlen, ylen, invmap, L_ali, rmsd_ali, D0_MIN, Lnorm, d0,
                d0_search, score_d8, t, u, mol_type);
            D0_MIN = prevD0_MIN;
            Lnorm = prevLnorm;
            d0 = prevd0;

            TM = detailed_search_standard(r1, r2, xtm, ytm, xt, xa, ya,
                xlen, ylen, invmap, t, u, 40, 8, local_d0_search, true, Lnorm,
                score_d8, d0);
            if (TM > TMmax)
            {
                TMmax = TM;
                for (i = 0; i<ylen; i++) invmap0[i] = invmap[i];
            }
            // Different from get_initial, get_initial_ss and get_initial_ssplus
            TM = DP_iter_dimer(r1, r2, xtm, ytm, xt, path, val, xa, ya,
                xlen, ylen, mask, t, u, invmap, 0, 2,
                (fast_opt)?2:30, local_d0_search, D0_MIN, Lnorm, d0, score_d8);
            if (TM>TMmax)
            {
                TMmax = TM;
                for (i = 0; i<ylen; i++) invmap0[i] = invmap[i];
            }
        }
    }



    //*******************************************************************//
    //    The alignment will not be changed any more in the following    //
    //*******************************************************************//
    //check if the initial alignment is generated appropriately
    bool flag=false;
    for(i=0; i<ylen; i++)
    {
        if(invmap0[i]>=0)
        {
            flag=true;
            break;
        }
    }
    if(!flag)
    {
        cout << "There is no alignment between the two proteins! "
             << "Program stop with no result!" << endl;
        TM1=TM2=TM3=TM4=TM5=0;
        return 1;
    }

    /* last TM-score pre-termination */
    if (TMcut>0)
    {
        double TMtmp=approx_TM(xlen, ylen, a_opt,
            xa, ya, t0, u0, invmap0, mol_type);

        if (TMtmp<0.6*TMcut)
        {
            TM1=TM2=TM3=TM4=TM5=TMtmp;
            clean_up_after_approx_TM(invmap0, invmap, score, path, val,
                xtm, ytm, xt, r1, r2, xlen, minlen);
            return 7;
        }
    }

    //********************************************************************//
    //    Detailed TMscore search engine --> prepare for final TMscore    //
    //********************************************************************//
    //run detailed TMscore search engine for the best alignment, and
    //extract the best rotation matrix (t, u) for the best alignment
    simplify_step=1;
    if (fast_opt) simplify_step=40;
    score_sum_method=8;
    TM = detailed_search_standard(r1, r2, xtm, ytm, xt, xa, ya, xlen, ylen,
        invmap0, t, u, simplify_step, score_sum_method, local_d0_search,
        false, Lnorm, score_d8, d0);

    //select pairs with dis<d8 for final TMscore computation and output alignment
    int k=0;
    int *m1, *m2;
    double d;
    m1=new int[xlen]; //alignd index in x
    m2=new int[ylen]; //alignd index in y
    do_rotation(xa, xt, xlen, t, u);
    k=0;
    for(int j=0; j<ylen; j++)
    {
        i=invmap0[j];
        if(i>=0)//aligned
        {
            n_ali++;
            d=sqrt(dist(&xt[i][0], &ya[j][0]));
            if (d <= score_d8 || (i_opt == 3))
            {
                m1[k]=i;
                m2[k]=j;

                xtm[k][0]=xa[i][0];
                xtm[k][1]=xa[i][1];
                xtm[k][2]=xa[i][2];

                ytm[k][0]=ya[j][0];
                ytm[k][1]=ya[j][1];
                ytm[k][2]=ya[j][2];

                r1[k][0] = xt[i][0];
                r1[k][1] = xt[i][1];
                r1[k][2] = xt[i][2];
                r2[k][0] = ya[j][0];
                r2[k][1] = ya[j][1];
                r2[k][2] = ya[j][2];

                k++;
            }
        }
    }
    n_ali8=k;

    Kabsch(r1, r2, n_ali8, 0, &rmsd0, t, u);// rmsd0 is used for final output, only recalculate rmsd0, not t & u
    rmsd0 = sqrt(rmsd0 / n_ali8);


    //****************************************//
    //              Final TMscore             //
    //    Please set parameters for output    //
    //****************************************//
    double rmsd;
    simplify_step=1;
    score_sum_method=0;
    double Lnorm_0=ylen;


    //normalized by length of structure A
    parameter_set4final(Lnorm_0, D0_MIN, Lnorm, d0, d0_search, mol_type);
    d0A=d0;
    d0_0=d0A;
    local_d0_search = d0_search;
    TM1 = TMscore8_search(r1, r2, xtm, ytm, xt, n_ali8, t0, u0, simplify_step,
        score_sum_method, &rmsd, local_d0_search, Lnorm, score_d8, d0);
    TM_0 = TM1;

    //normalized by length of structure B
    parameter_set4final(xlen+0.0, D0_MIN, Lnorm, d0, d0_search, mol_type);
    d0B=d0;
    local_d0_search = d0_search;
    TM2 = TMscore8_search(r1, r2, xtm, ytm, xt, n_ali8, t, u, simplify_step,
        score_sum_method, &rmsd, local_d0_search, Lnorm, score_d8, d0);

    double Lnorm_d0;
    if (a_opt>0)
    {
        //normalized by average length of structures A, B
        Lnorm_0=(xlen+ylen)*0.5;
        parameter_set4final(Lnorm_0, D0_MIN, Lnorm, d0, d0_search, mol_type);
        d0a=d0;
        d0_0=d0a;
        local_d0_search = d0_search;

        TM3 = TMscore8_search(r1, r2, xtm, ytm, xt, n_ali8, t0, u0,
            simplify_step, score_sum_method, &rmsd, local_d0_search, Lnorm,
            score_d8, d0);
        TM_0=TM3;
    }
    if (u_opt)
    {
        //normalized by user assigned length
        parameter_set4final(Lnorm_ass, D0_MIN, Lnorm,
            d0, d0_search, mol_type);
        d0u=d0;
        d0_0=d0u;
        Lnorm_0=Lnorm_ass;
        local_d0_search = d0_search;
        TM4 = TMscore8_search(r1, r2, xtm, ytm, xt, n_ali8, t0, u0,
            simplify_step, score_sum_method, &rmsd, local_d0_search, Lnorm,
            score_d8, d0);
        TM_0=TM4;
    }
    if (d_opt)
    {
        //scaled by user assigned d0
        parameter_set4scale(ylen, d0_scale, Lnorm, d0, d0_search);
        d0_out=d0_scale;
        d0_0=d0_scale;
        //Lnorm_0=ylen;
        Lnorm_d0=Lnorm_0;
        local_d0_search = d0_search;
        TM5 = TMscore8_search(r1, r2, xtm, ytm, xt, n_ali8, t0, u0,
            simplify_step, score_sum_method, &rmsd, local_d0_search, Lnorm,
            score_d8, d0);
        TM_0=TM5;
    }

    /* derive alignment from superposition */
    int ali_len=xlen+ylen; //maximum length of alignment
    seqxA.assign(ali_len,'-');
    seqM.assign( ali_len,' ');
    seqyA.assign(ali_len,'-');
    
    //do_rotation(xa, xt, xlen, t, u);
    do_rotation(xa, xt, xlen, t0, u0);

    int kk=0, i_old=0, j_old=0;
    d=0;
    for(int k=0; k<n_ali8; k++)
    {
        for(int i=i_old; i<m1[k]; i++)
        {
            //align x to gap
            seqxA[kk]=seqx[i];
            seqyA[kk]='-';
            seqM[kk]=' ';                    
            kk++;
        }

        for(int j=j_old; j<m2[k]; j++)
        {
            //align y to gap
            seqxA[kk]='-';
            seqyA[kk]=seqy[j];
            seqM[kk]=' ';
            kk++;
        }

        seqxA[kk]=seqx[m1[k]];
        seqyA[kk]=seqy[m2[k]];
        Liden+=(seqxA[kk]==seqyA[kk]);
        d=sqrt(dist(&xt[m1[k]][0], &ya[m2[k]][0]));
        if(d<d0_out) seqM[kk]=':';
        else         seqM[kk]='.';
        kk++;  
        i_old=m1[k]+1;
        j_old=m2[k]+1;
    }

    //tail
    for(int i=i_old; i<xlen; i++)
    {
        //align x to gap
        seqxA[kk]=seqx[i];
        seqyA[kk]='-';
        seqM[kk]=' ';
        kk++;
    }    
    for(int j=j_old; j<ylen; j++)
    {
        //align y to gap
        seqxA[kk]='-';
        seqyA[kk]=seqy[j];
        seqM[kk]=' ';
        kk++;
    }
    seqxA=seqxA.substr(0,kk);
    seqyA=seqyA.substr(0,kk);
    seqM =seqM.substr(0,kk);

    /* free memory */
    clean_up_after_approx_TM(invmap0, invmap, score, path, val,
        xtm, ytm, xt, r1, r2, xlen, minlen);
    delete [] m1;
    delete [] m2;
    return 0; // zero for no exception
}

void MMalign_dimer(double & total_score, 
    const vector<vector<vector<double> > >&xa_vec,
    const vector<vector<vector<double> > >&ya_vec,
    const vector<vector<char> >&seqx_vec, const vector<vector<char> >&seqy_vec,
    const vector<vector<char> >&secx_vec, const vector<vector<char> >&secy_vec,
    const vector<int> &mol_vec1, const vector<int> &mol_vec2,
    const vector<int> &xlen_vec, const vector<int> &ylen_vec,
    double **xa, double **ya, char *seqx, char *seqy, char *secx, char *secy,
    int len_aa, int len_na, int chain1_num, int chain2_num, double **TMave_mat,
    vector<vector<string> >&seqxA_mat, vector<vector<string> >&seqyA_mat,
    int *assign1_list, int *assign2_list, vector<string>&sequence,
    double d0_scale, bool fast_opt)
{
    int i,j;
    int xlen=0;
    int ylen=0;
    vector<int> xlen_dimer;
    vector<int> ylen_dimer;
    for (i=0;i<chain1_num;i++)
    {
        j=assign1_list[i];
        if (j<0) continue;
        xlen+=xlen_vec[i];
        ylen+=ylen_vec[j];
        xlen_dimer.push_back(xlen_vec[i]);
        ylen_dimer.push_back(ylen_vec[j]);
    }
    if (xlen<=3 || ylen<=3) return;

    bool **mask; // mask out inter-chain region
    NewArray(&mask, xlen+1, ylen+1);
    for (i=0;i<xlen+1;i++) for (j=0;j<ylen+1;j++) mask[i][j]=false;
    for (i=0;i<xlen_dimer[0]+1;i++) mask[i][0]=true;
    for (j=0;j<ylen_dimer[0]+1;j++) mask[0][j]=true;
    int c,prev_xlen,prev_ylen;
    prev_xlen=1;
    prev_ylen=1;
    for (c=0;c<xlen_dimer.size();c++)
    {
        for (i=prev_xlen;i<prev_xlen+xlen_dimer[c];i++)
            for (j=prev_ylen;j<prev_ylen+ylen_dimer[c];j++) mask[i][j]=true;
        prev_xlen+=xlen_dimer[c];
        prev_ylen+=ylen_dimer[c];
    }
    vector<int>().swap(xlen_dimer);
    vector<int>().swap(ylen_dimer);

    seqx = new char[xlen+1];
    secx = new char[xlen+1];
    NewArray(&xa, xlen, 3);
    seqy = new char[ylen+1];
    secy = new char[ylen+1];
    NewArray(&ya, ylen, 3);

    int mol_type=copy_chain_pair_data(xa_vec, ya_vec, seqx_vec, seqy_vec,
        secx_vec, secy_vec, mol_vec1, mol_vec2, xlen_vec, ylen_vec,
        xa, ya, seqx, seqy, secx, secy, chain1_num, chain2_num,
        seqxA_mat, seqyA_mat, assign1_list, assign2_list, sequence);

    /* declare variable specific to this pair of TMalign */
    double t0[3], u0[3][3];
    double TM1, TM2;
    double TM3, TM4, TM5;     // for a_opt, u_opt, d_opt
    double d0_0, TM_0;
    double d0A, d0B, d0u, d0a;
    double d0_out=5.0;
    string seqM, seqxA, seqyA;// for output alignment
    double rmsd0 = 0.0;
    int L_ali;                // Aligned length in standard_TMscore
    double Liden=0;
    double TM_ali, rmsd_ali;  // TMscore and rmsd in standard_TMscore
    int n_ali=0;
    int n_ali8=0;

    double Lnorm_ass=len_aa+len_na;

    TMalign_dimer_main(xa, ya, seqx, seqy, secx, secy,
        t0, u0, TM1, TM2, TM3, TM4, TM5,
        d0_0, TM_0, d0A, d0B, d0u, d0a, d0_out, seqM, seqxA, seqyA,
        rmsd0, L_ali, Liden, TM_ali, rmsd_ali, n_ali, n_ali8,
        xlen, ylen, mask, sequence, Lnorm_ass, d0_scale,
        1, false, true, false, fast_opt, mol_type, -1);

    /* clean up TM-align */
    delete [] seqx;
    delete [] seqy;
    delete [] secx;
    delete [] secy;
    DeleteArray(&xa,xlen);
    DeleteArray(&ya,ylen);
    DeleteArray(&mask,xlen+1);

    /* re-compute chain level alignment */
    total_score=0;
    for (i=0;i<chain1_num;i++)
    {
        xlen=xlen_vec[i];
        if (xlen<3)
        {
            for (j=0;j<chain2_num;j++) TMave_mat[i][j]=-1;
            continue;
        }
        seqx = new char[xlen+1];
        secx = new char[xlen+1];
        NewArray(&xa, xlen, 3);
        copy_chain_data(xa_vec[i],seqx_vec[i],secx_vec[i],
            xlen,xa,seqx,secx);

        double **xt;
        NewArray(&xt, xlen, 3);
        do_rotation(xa, xt, xlen, t0, u0);

        for (j=0;j<chain2_num;j++)
        {
            if (mol_vec1[i]*mol_vec2[j]<0) //no protein-RNA alignment
            {
                TMave_mat[i][j]=-1;
                continue;
            }

            ylen=ylen_vec[j];
            if (ylen<3)
            {
                TMave_mat[i][j]=-1;
                continue;
            }
            seqy = new char[ylen+1];
            secy = new char[ylen+1];
            NewArray(&ya, ylen, 3);
            copy_chain_data(ya_vec[j],seqy_vec[j],secy_vec[j],
                ylen,ya,seqy,secy);

            /* declare variable specific to this pair of TMalign */
            d0_out=5.0;
            seqM.clear();
            seqxA.clear();
            seqyA.clear();
            rmsd0 = 0.0;
            Liden=0;
            int *invmap = new int[ylen+1];

            double Lnorm_ass=len_aa;
            if (mol_vec1[i]+mol_vec2[j]>0) Lnorm_ass=len_na;

            /* entry function for structure alignment */
            se_main(xt, ya, seqx, seqy, TM1, TM2, TM3, TM4, TM5,
                d0_0, TM_0, d0A, d0B, d0u, d0a, d0_out, seqM, seqxA, seqyA,
                rmsd0, L_ali, Liden, TM_ali, rmsd_ali, n_ali, n_ali8,
                xlen, ylen, sequence, Lnorm_ass, d0_scale,
                0, false, 2, false, mol_vec1[i]+mol_vec2[j], 1, invmap);

            /* print result */
            seqxA_mat[i][j]=seqxA;
            seqyA_mat[i][j]=seqyA;

            TMave_mat[i][j]=TM4*Lnorm_ass;
            if (assign1_list[i]==j)
            {
                if (TM4<=0) assign1_list[i]=assign2_list[j]=-1;
                else        total_score+=TMave_mat[i][j];
            }

            /* clean up */
            seqM.clear();
            seqxA.clear();
            seqyA.clear();

            delete[]seqy;
            delete[]secy;
            DeleteArray(&ya,ylen);
            delete[]invmap;
        }
        delete[]seqx;
        delete[]secx;
        DeleteArray(&xa,xlen);
        DeleteArray(&xt,xlen);
    }
    return;
}






int main(int argc, char *argv[])
{
    if (argc < 2) print_help();


    clock_t t1, t2;
    t1 = clock();

    /**********************/
    /*    get argument    */
    /**********************/
    string xname       = "";
    string yname       = "";
    string fname_super = ""; // file name for superposed structure
    string fname_lign  = ""; // file name for user alignment
    string fname_matrix= ""; // file name for output matrix
    vector<string> sequence; // get value from alignment file
    double d0_scale    =0;

    bool h_opt = false; // print full help message
    bool v_opt = false; // print version
    bool m_opt = false; // flag for -m, output rotation matrix
    bool o_opt = false; // flag for -o, output superposed structure
    int  a_opt = 0;     // flag for -a, do not normalized by average length
    bool d_opt = false; // flag for -d, user specified d0

    bool   full_opt  = false;// do not show chain level alignment
    double TMcut     =-1;
    int    infmt1_opt=-1;    // PDB or PDBx/mmCIF format for chain_1
    int    infmt2_opt=-1;    // PDB or PDBx/mmCIF format for chain_2
    int    ter_opt   =1;     // ENDMDL or END
    int    split_opt =2;     // split by chain
    int    outfmt_opt=0;     // set -outfmt to full output
    bool   fast_opt  =false; // flags for -fast, fTM-align algorithm
    int    mirror_opt=0;     // do not align mirror
    int    het_opt   =0;     // do not read HETATM residues
    string atom_opt  ="auto";// use C alpha atom for protein and C3' for RNA
    string mol_opt   ="auto";// auto-detect the molecule type as protein/RNA
    string suffix_opt="";    // set -suffix to empty
    string dir1_opt  ="";    // set -dir1 to empty
    string dir2_opt  ="";    // set -dir2 to empty
    vector<string> chain1_list; // only when -dir1 is set
    vector<string> chain2_list; // only when -dir2 is set

    for(int i = 1; i < argc; i++)
    {
        if ( !strcmp(argv[i],"-o") && i < (argc-1) )
        {
            fname_super = argv[i + 1];     o_opt = true; i++;
        }
        else if ( !strcmp(argv[i],"-a") && i < (argc-1) )
        {
            if (!strcmp(argv[i + 1], "T"))      a_opt=true;
            else if (!strcmp(argv[i + 1], "F")) a_opt=false;
            else 
            {
                a_opt=atoi(argv[i + 1]);
                if (a_opt!=-2 && a_opt!=-1 && a_opt!=1)
                    PrintErrorAndQuit("-a must be -2, -1, 1, T or F");
            }
            i++;
        }
        else if ( !strcmp(argv[i],"-full") && i < (argc-1) )
        {
            if (!strcmp(argv[i + 1], "T"))      full_opt=true;
            else if (!strcmp(argv[i + 1], "F")) full_opt=false;
            else PrintErrorAndQuit("-full must be T or F");
            i++;
        }
        else if ( !strcmp(argv[i],"-d") && i < (argc-1) )
        {
            d0_scale = atof(argv[i + 1]); d_opt = true; i++;
        }
        else if ( !strcmp(argv[i],"-v") )
        {
            v_opt = true;
        }
        else if ( !strcmp(argv[i],"-h") )
        {
            h_opt = true;
        }
        else if (!strcmp(argv[i], "-m") && i < (argc-1) )
        {
            fname_matrix = argv[i + 1];    m_opt = true; i++;
        }// get filename for rotation matrix
        else if (!strcmp(argv[i], "-fast"))
        {
            fast_opt = true;
        }
        else if ( !strcmp(argv[i],"-infmt1") && i < (argc-1) )
        {
            infmt1_opt=atoi(argv[i + 1]); i++;
        }
        else if ( !strcmp(argv[i],"-infmt2") && i < (argc-1) )
        {
            infmt2_opt=atoi(argv[i + 1]); i++;
        }
        else if ( !strcmp(argv[i],"-ter") && i < (argc-1) )
        {
            ter_opt=atoi(argv[i + 1]); i++;
        }
        else if ( !strcmp(argv[i],"-split") && i < (argc-1) )
        {
            split_opt=atoi(argv[i + 1]); i++;
        }
        else if ( !strcmp(argv[i],"-atom") && i < (argc-1) )
        {
            atom_opt=argv[i + 1]; i++;
        }
        else if ( !strcmp(argv[i],"-mol") && i < (argc-1) )
        {
            mol_opt=argv[i + 1]; i++;
        }
        else if ( !strcmp(argv[i],"-dir1") && i < (argc-1) )
        {
            dir1_opt=argv[i + 1]; i++;
        }
        else if ( !strcmp(argv[i],"-dir2") && i < (argc-1) )
        {
            dir2_opt=argv[i + 1]; i++;
        }
        else if ( !strcmp(argv[i],"-suffix") && i < (argc-1) )
        {
            suffix_opt=argv[i + 1]; i++;
        }
        else if ( !strcmp(argv[i],"-outfmt") && i < (argc-1) )
        {
            outfmt_opt=atoi(argv[i + 1]); i++;
        }
        else if ( !strcmp(argv[i],"-TMcut") && i < (argc-1) )
        {
            TMcut=atof(argv[i + 1]); i++;
        }
        else if ( !strcmp(argv[i],"-het") && i < (argc-1) )
        {
            het_opt=atoi(argv[i + 1]); i++;
        }
        else if (xname.size() == 0) xname=argv[i];
        else if (yname.size() == 0) yname=argv[i];
        else PrintErrorAndQuit(string("ERROR! Undefined option ")+argv[i]);
    }

    if(yname.size()==0)
    {
        if (h_opt) print_help(h_opt);
        if (v_opt)
        {
            print_version();
            exit(EXIT_FAILURE);
        }
        if (xname.size()==0)
            PrintErrorAndQuit("Please provide input structures");
        PrintErrorAndQuit("Please provide the second input structure");
    }

    if (suffix_opt.size() && dir1_opt.size()+dir2_opt.size()==0)
        PrintErrorAndQuit("-suffix is only valid if -dir1 or -dir2 is set");
    if ((dir1_opt.size() || dir2_opt.size()) && (m_opt || o_opt))
        PrintErrorAndQuit("-m or -o cannot be set with -dir1 or -dir2");
    if (atom_opt.size()!=4)
        PrintErrorAndQuit("ERROR! Atom name must have 4 characters, including space.");
    if (mol_opt!="auto" && mol_opt!="protein" && mol_opt!="RNA")
        PrintErrorAndQuit("ERROR! Molecule type must be either RNA or protein.");
    else if (mol_opt=="protein" && atom_opt=="auto")
        atom_opt=" CA ";
    else if (mol_opt=="RNA" && atom_opt=="auto")
        atom_opt=" C3'";

    if (d_opt && d0_scale<=0)
        PrintErrorAndQuit("Wrong value for option -d!  It should be >0");
    if (outfmt_opt>=2 && (a_opt || d_opt))
        PrintErrorAndQuit("-outfmt 2 cannot be used with -a, -d");
    if (ter_opt!=0 && ter_opt!=1)
        PrintErrorAndQuit("-ter should be 1 or 0");
    if (split_opt!=1 && split_opt!=2)
        PrintErrorAndQuit("-split should be 1 or 2");
    else if (split_opt==1 && ter_opt!=0)
        PrintErrorAndQuit("-split 1 should be used with -ter 0");

    if (m_opt && fname_matrix == "") // Output rotation matrix: matrix.txt
        PrintErrorAndQuit("ERROR! Please provide a file name for option -m!");

    /* parse file list */
    if (dir1_opt.size()==0) chain1_list.push_back(xname);
    else file2chainlist(chain1_list, xname, dir1_opt, suffix_opt);

    if (dir2_opt.size()==0) chain2_list.push_back(yname);
    else file2chainlist(chain2_list, yname, dir2_opt, suffix_opt);

    if (outfmt_opt==2)
        cout<<"#PDBchain1\tPDBchain2\tTM1\tTM2\t"
            <<"RMSD\tID1\tID2\tIDali\tL1\tL2\tLali"<<endl;

    /* declare previously global variables */
    vector<vector<vector<double> > > xa_vec; // structure of complex1
    vector<vector<vector<double> > > ya_vec; // structure of complex2
    vector<vector<char> >seqx_vec; // sequence of complex1
    vector<vector<char> >seqy_vec; // sequence of complex2
    vector<vector<char> >secx_vec; // secondary structure of complex1
    vector<vector<char> >secy_vec; // secondary structure of complex2
    vector<int> mol_vec1;          // molecule type of complex1, RNA if >0
    vector<int> mol_vec2;          // molecule type of complex2, RNA if >0
    vector<string> chainID_list1;  // list of chainID1
    vector<string> chainID_list2;  // list of chainID2
    vector<int> xlen_vec;          // length of complex1
    vector<int> ylen_vec;          // length of complex2
    int    i,j;                    // chain index
    int    xlen, ylen;             // chain length
    double **xa, **ya;             // structure of single chain
    char   *seqx, *seqy;           // for the protein sequence 
    char   *secx, *secy;           // for the secondary structure 
    int    xlen_aa,ylen_aa;        // total length of protein
    int    xlen_na,ylen_na;        // total length of RNA/DNA
    vector<string> resi_vec1;  // residue index for chain1
    vector<string> resi_vec2;  // residue index for chain2

    /* parse complex */
    parse_chain_list(chain1_list, xa_vec, seqx_vec, secx_vec, mol_vec1,
        xlen_vec, chainID_list1, ter_opt, split_opt, mol_opt, infmt1_opt,
        atom_opt, mirror_opt, het_opt, xlen_aa, xlen_na, o_opt, resi_vec1);
    if (xa_vec.size()==0) PrintErrorAndQuit("ERROR! 0 chain in complex 1");
    parse_chain_list(chain2_list, ya_vec, seqy_vec, secy_vec, mol_vec2,
        ylen_vec, chainID_list2, ter_opt, split_opt, mol_opt, infmt2_opt,
        atom_opt, 0, het_opt, ylen_aa, ylen_na, o_opt, resi_vec2);
    if (ya_vec.size()==0) PrintErrorAndQuit("ERROR! 0 chain in complex 2");
    int len_aa=getmin(xlen_aa,ylen_aa);
    int len_na=getmin(xlen_na,ylen_na);
    if (a_opt)
    {
        len_aa=(xlen_aa+ylen_aa)/2;
        len_na=(xlen_na+ylen_na)/2;
    }

    /* perform monomer alignment if there is only one chain */
    if (xa_vec.size()==1 && ya_vec.size()==1)
    {
        xlen = xlen_vec[0];
        ylen = ylen_vec[0];
        seqx = new char[xlen+1];
        seqy = new char[ylen+1];
        secx = new char[xlen+1];
        secy = new char[ylen+1];
        NewArray(&xa, xlen, 3);
        NewArray(&ya, ylen, 3);
        copy_chain_data(xa_vec[0],seqx_vec[0],secx_vec[0], xlen,xa,seqx,secx);
        copy_chain_data(ya_vec[0],seqy_vec[0],secy_vec[0], ylen,ya,seqy,secy);
        
        /* declare variable specific to this pair of TMalign */
        double t0[3], u0[3][3];
        double TM1, TM2;
        double TM3, TM4, TM5;     // for a_opt, u_opt, d_opt
        double d0_0, TM_0;
        double d0A, d0B, d0u, d0a;
        double d0_out=5.0;
        string seqM, seqxA, seqyA;// for output alignment
        double rmsd0 = 0.0;
        int L_ali;                // Aligned length in standard_TMscore
        double Liden=0;
        double TM_ali, rmsd_ali;  // TMscore and rmsd in standard_TMscore
        int n_ali=0;
        int n_ali8=0;

        /* entry function for structure alignment */
        TMalign_main(xa, ya, seqx, seqy, secx, secy,
            t0, u0, TM1, TM2, TM3, TM4, TM5,
            d0_0, TM_0, d0A, d0B, d0u, d0a, d0_out,
            seqM, seqxA, seqyA,
            rmsd0, L_ali, Liden, TM_ali, rmsd_ali, n_ali, n_ali8,
            xlen, ylen, sequence, 0, d0_scale,
            0, a_opt, false, d_opt, fast_opt,
            mol_vec1[0]+mol_vec2[0],TMcut);

        /* print result */
        output_results(
            xname.substr(dir1_opt.size()),
            yname.substr(dir2_opt.size()),
            chainID_list1[0], chainID_list2[0],
            xlen, ylen, t0, u0, TM1, TM2, TM3, TM4, TM5, rmsd0, d0_out,
            seqM.c_str(), seqxA.c_str(), seqyA.c_str(), Liden,
            n_ali8, L_ali, TM_ali, rmsd_ali, TM_0, d0_0, d0A, d0B,
            0, d0_scale, d0a, d0u, (m_opt?fname_matrix:"").c_str(),
            outfmt_opt, ter_opt, true, split_opt, o_opt, fname_super,
            0, a_opt, false, d_opt, mirror_opt, resi_vec1, resi_vec2);

        /* clean up */
        seqM.clear();
        seqxA.clear();
        seqyA.clear();
        delete[]seqx;
        delete[]seqy;
        delete[]secx;
        delete[]secy;
        DeleteArray(&xa,xlen);
        DeleteArray(&ya,ylen);
        chain1_list.clear();
        chain2_list.clear();
        sequence.clear();

        vector<vector<vector<double> > >().swap(xa_vec); // structure of complex1
        vector<vector<vector<double> > >().swap(ya_vec); // structure of complex2
        vector<vector<char> >().swap(seqx_vec); // sequence of complex1
        vector<vector<char> >().swap(seqy_vec); // sequence of complex2
        vector<vector<char> >().swap(secx_vec); // secondary structure of complex1
        vector<vector<char> >().swap(secy_vec); // secondary structure of complex2
        mol_vec1.clear();       // molecule type of complex1, RNA if >0
        mol_vec2.clear();       // molecule type of complex2, RNA if >0
        chainID_list1.clear();  // list of chainID1
        chainID_list2.clear();  // list of chainID2
        xlen_vec.clear();       // length of complex1
        ylen_vec.clear();       // length of complex2

        t2 = clock();
        float diff = ((float)t2 - (float)t1)/CLOCKS_PER_SEC;
        printf("#Total CPU time is %5.2f seconds\n", diff);
        return 0;
    }

    /* declare TM-score tables */
    int chain1_num=xa_vec.size();
    int chain2_num=ya_vec.size();
    vector<string> tmp_str_vec(chain2_num,"");
    double **TMave_mat;
    double **ut_mat; // rotation matrices for all-against-all alignment
    int ui,uj,ut_idx;
    NewArray(&TMave_mat,chain1_num,chain2_num);
    NewArray(&ut_mat,chain1_num*chain2_num,4*3);
    vector<vector<string> >seqxA_mat(chain1_num,tmp_str_vec);
    vector<vector<string> > seqM_mat(chain1_num,tmp_str_vec);
    vector<vector<string> >seqyA_mat(chain1_num,tmp_str_vec);

    double maxTMmono=-1;
    int maxTMmono_i,maxTMmono_j;

    /* get all-against-all alignment */
    if (len_aa+len_na>500) fast_opt=true;
    for (i=0;i<chain1_num;i++)
    {
        xlen=xlen_vec[i];
        if (xlen<3)
        {
            for (j=0;j<chain2_num;j++) TMave_mat[i][j]=-1;
            continue;
        }
        seqx = new char[xlen+1];
        secx = new char[xlen+1];
        NewArray(&xa, xlen, 3);
        copy_chain_data(xa_vec[i],seqx_vec[i],secx_vec[i],
            xlen,xa,seqx,secx);

        for (j=0;j<chain2_num;j++)
        {
            ut_idx=i*chain2_num+j;
            for (ui=0;ui<4;ui++)
                for (uj=0;uj<3;uj++) ut_mat[ut_idx][ui*3+uj]=0;
            ut_mat[ut_idx][0]=1;
            ut_mat[ut_idx][4]=1;
            ut_mat[ut_idx][8]=1;

            if (mol_vec1[i]*mol_vec2[j]<0) //no protein-RNA alignment
            {
                TMave_mat[i][j]=-1;
                continue;
            }

            ylen=ylen_vec[j];
            if (ylen<3)
            {
                TMave_mat[i][j]=-1;
                continue;
            }
            seqy = new char[ylen+1];
            secy = new char[ylen+1];
            NewArray(&ya, ylen, 3);
            copy_chain_data(ya_vec[j],seqy_vec[j],secy_vec[j],
                ylen,ya,seqy,secy);

            /* declare variable specific to this pair of TMalign */
            double t0[3], u0[3][3];
            double TM1, TM2;
            double TM3, TM4, TM5;     // for a_opt, u_opt, d_opt
            double d0_0, TM_0;
            double d0A, d0B, d0u, d0a;
            double d0_out=5.0;
            string seqM, seqxA, seqyA;// for output alignment
            double rmsd0 = 0.0;
            int L_ali;                // Aligned length in standard_TMscore
            double Liden=0;
            double TM_ali, rmsd_ali;  // TMscore and rmsd in standard_TMscore
            int n_ali=0;
            int n_ali8=0;

            int Lnorm_tmp=len_aa;
            if (mol_vec1[i]+mol_vec2[j]>0) Lnorm_tmp=len_na;

            /* entry function for structure alignment */
            TMalign_main(xa, ya, seqx, seqy, secx, secy,
                t0, u0, TM1, TM2, TM3, TM4, TM5,
                d0_0, TM_0, d0A, d0B, d0u, d0a, d0_out,
                seqM, seqxA, seqyA,
                rmsd0, L_ali, Liden, TM_ali, rmsd_ali, n_ali, n_ali8,
                xlen, ylen, sequence, Lnorm_tmp, d0_scale,
                0, false, true, false, fast_opt,
                mol_vec1[i]+mol_vec2[j],TMcut);

            /* store result */
            for (ui=0;ui<3;ui++)
                for (uj=0;uj<3;uj++) ut_mat[ut_idx][ui*3+uj]=u0[ui][uj];
            for (uj=0;uj<3;uj++) ut_mat[ut_idx][9+uj]=t0[uj];
            seqxA_mat[i][j]=seqxA;
            seqyA_mat[i][j]=seqyA;
            TMave_mat[i][j]=TM4*Lnorm_tmp;
            if (TMave_mat[i][j]>maxTMmono)
            {
                maxTMmono=TMave_mat[i][j];
                maxTMmono_i=i;
                maxTMmono_j=j;
            }

            /* clean up */
            seqM.clear();
            seqxA.clear();
            seqyA.clear();

            delete[]seqy;
            delete[]secy;
            DeleteArray(&ya,ylen);
        }

        delete[]seqx;
        delete[]secx;
        DeleteArray(&xa,xlen);
    }

    /* calculate initial chain-chain assignment */
    int *assign1_list; // value is index of assigned chain2
    int *assign2_list; // value is index of assigned chain1
    assign1_list=new int[chain1_num];
    assign2_list=new int[chain2_num];
    double total_score=enhanced_greedy_search(TMave_mat, assign1_list,
        assign2_list, chain1_num, chain2_num);
    if (total_score<=0) PrintErrorAndQuit("ERROR! No assignable chain");

    /* refine alignment for large oligomers */
    int aln_chain_num=count_assign_pair(assign1_list,chain1_num);
    bool is_oligomer=(aln_chain_num>=3);
    if (aln_chain_num==2) // dimer alignment
    {
        int na_chain_num1,na_chain_num2,aa_chain_num1,aa_chain_num2;
        count_na_aa_chain_num(na_chain_num1,aa_chain_num1,mol_vec1);
        count_na_aa_chain_num(na_chain_num2,aa_chain_num2,mol_vec2);

        /* align protein-RNA hybrid dimer to another hybrid dimer */
        if (na_chain_num1==1 && na_chain_num2==1 && 
            aa_chain_num1==1 && aa_chain_num2==1) is_oligomer=false;
        /* align pure protein dimer or pure RNA dimer */
        else if ((getmin(na_chain_num1,na_chain_num2)==0 && 
                    aa_chain_num1==2 && aa_chain_num2==2) ||
                 (getmin(aa_chain_num1,aa_chain_num2)==0 && 
                    na_chain_num1==2 && na_chain_num2==2))
        {
            adjust_dimer_assignment(xa_vec,ya_vec,xlen_vec,ylen_vec,mol_vec1,
                mol_vec2,assign1_list,assign2_list,seqxA_mat,seqyA_mat);
            is_oligomer=false; // cannot refiner further
        }
        else is_oligomer=true; /* align oligomers to dimer */
    }

    if (aln_chain_num>=3 || is_oligomer) // oligomer alignment
    {
        /* extract centroid coordinates */
        double **xcentroids;
        double **ycentroids;
        NewArray(&xcentroids, chain1_num, 3);
        NewArray(&ycentroids, chain2_num, 3);
        double d0MM=getmin(
            calculate_centroids(xa_vec, chain1_num, xcentroids),
            calculate_centroids(ya_vec, chain2_num, ycentroids));

        /* refine enhanced greedy search with centroid superposition */
        //double het_deg=check_heterooligomer(TMave_mat, chain1_num, chain2_num);
        homo_refined_greedy_search(TMave_mat, assign1_list,
            assign2_list, chain1_num, chain2_num, xcentroids,
            ycentroids, d0MM, len_aa+len_na, ut_mat);
        hetero_refined_greedy_search(TMave_mat, assign1_list,
            assign2_list, chain1_num, chain2_num, xcentroids,
            ycentroids, d0MM, len_aa+len_na);
        
        /* clean up */
        DeleteArray(&xcentroids, chain1_num);
        DeleteArray(&ycentroids, chain2_num);
    }

    /* store initial assignment */
    int init_pair_num=count_assign_pair(assign1_list,chain1_num);
    int *assign1_init, *assign2_init;
    assign1_init=new int[chain1_num];
    assign2_init=new int[chain2_num];
    double **TMave_init;
    NewArray(&TMave_init,chain1_num,chain2_num);
    vector<vector<string> >seqxA_init(chain1_num,tmp_str_vec);
    vector<vector<string> >seqyA_init(chain1_num,tmp_str_vec);
    vector<string> sequence_init;
    copy_chain_assign_data(chain1_num, chain2_num, sequence_init,
        seqxA_mat,  seqyA_mat,  assign1_list, assign2_list, TMave_mat,
        seqxA_init, seqyA_init, assign1_init, assign2_init, TMave_init);

    /* perform iterative alignment */
    double max_total_score=0; // ignore old total_score because previous
                              // score was from monomeric chain superpositions
    int max_iter=5-(int)((len_aa+len_na)/200);
    if (max_iter<2) max_iter=2;
    MMalign_iter(max_total_score, max_iter, xa_vec, ya_vec, seqx_vec, seqy_vec,
        secx_vec, secy_vec, mol_vec1, mol_vec2, xlen_vec, ylen_vec,
        xa, ya, seqx, seqy, secx, secy, len_aa, len_na, chain1_num, chain2_num,
        TMave_mat, seqxA_mat, seqyA_mat, assign1_list, assign2_list, sequence,
        d0_scale, fast_opt);

    /* sometime MMalign_iter is even worse than monomer alignment */
    if (max_total_score<maxTMmono)
    {
        copy_chain_assign_data(chain1_num, chain2_num, sequence,
            seqxA_init, seqyA_init, assign1_init, assign2_init, TMave_init,
            seqxA_mat, seqyA_mat, assign1_list, assign2_list, TMave_mat);
        for (i=0;i<chain1_num;i++)
        {
            if (i!=maxTMmono_i) assign1_list[i]=-1;
            else assign1_list[i]=maxTMmono_j;
        }
        for (j=0;j<chain2_num;j++)
        {
            if (j!=maxTMmono_j) assign2_list[j]=-1;
            else assign2_list[j]=maxTMmono_i;
        }
        sequence[0]=seqxA_mat[maxTMmono_i][maxTMmono_j];
        sequence[1]=seqyA_mat[maxTMmono_i][maxTMmono_j];
        max_total_score=maxTMmono;
        MMalign_iter(max_total_score, max_iter, xa_vec, ya_vec, seqx_vec, seqy_vec,
            secx_vec, secy_vec, mol_vec1, mol_vec2, xlen_vec, ylen_vec,
            xa, ya, seqx, seqy, secx, secy, len_aa, len_na, chain1_num, chain2_num,
            TMave_mat, seqxA_mat, seqyA_mat, assign1_list, assign2_list, sequence,
            d0_scale, fast_opt);
    }

    /* perform cross chain alignment
     * in some cases, this leads to dramatic improvement, esp for homodimer */
    int iter_pair_num=count_assign_pair(assign1_list,chain1_num);
    if (iter_pair_num>=init_pair_num) copy_chain_assign_data(
        chain1_num, chain2_num, sequence_init,
        seqxA_mat, seqyA_mat, assign1_list, assign2_list, TMave_mat,
        seqxA_init, seqyA_init, assign1_init,  assign2_init,  TMave_init);
    double max_total_score_cross=max_total_score;

    //if (init_pair_num!=2 && is_oligomer==false) MMalign_cross(
        //max_total_score_cross, max_iter, xa_vec, ya_vec, seqx_vec, seqy_vec,
        //secx_vec, secy_vec, mol_vec1, mol_vec2, xlen_vec, ylen_vec,
        //xa, ya, seqx, seqy, secx, secy, len_aa, len_na, chain1_num, chain2_num,
        //TMave_init, seqxA_init, seqyA_init, assign1_init, assign2_init, sequence_init,
        //d0_scale, true);
    //else 
    MMalign_dimer(max_total_score_cross, xa_vec, ya_vec, seqx_vec, seqy_vec,
        secx_vec, secy_vec, mol_vec1, mol_vec2, xlen_vec, ylen_vec,
        xa, ya, seqx, seqy, secx, secy, len_aa, len_na, chain1_num, chain2_num,
        TMave_init, seqxA_init, seqyA_init, assign1_init, assign2_init, sequence_init,
        d0_scale, fast_opt);
    if (max_total_score_cross>max_total_score) 
    {
        max_total_score=max_total_score_cross;
        copy_chain_assign_data(chain1_num, chain2_num, sequence,
            seqxA_init, seqyA_init, assign1_init, assign2_init, TMave_init,
            seqxA_mat,  seqyA_mat,  assign1_list, assign2_list, TMave_mat);
    } 

    /* final alignment */
    if (outfmt_opt==0) print_version();
    MMalign_final(xname.substr(dir1_opt.size()), yname.substr(dir2_opt.size()),
        chainID_list1, chainID_list2,
        fname_super, fname_lign, fname_matrix,
        xa_vec, ya_vec, seqx_vec, seqy_vec,
        secx_vec, secy_vec, mol_vec1, mol_vec2, xlen_vec, ylen_vec,
        xa, ya, seqx, seqy, secx, secy, len_aa, len_na,
        chain1_num, chain2_num, TMave_mat,
        seqxA_mat, seqM_mat, seqyA_mat, assign1_list, assign2_list, sequence,
        d0_scale, m_opt, o_opt, outfmt_opt, ter_opt, split_opt,
        a_opt, d_opt, fast_opt, full_opt, mirror_opt, resi_vec1, resi_vec2);

    /* clean up everything */
    delete [] assign1_list;
    delete [] assign2_list;
    DeleteArray(&TMave_mat,chain1_num);
    DeleteArray(&ut_mat,   chain1_num*chain2_num);
    vector<vector<string> >().swap(seqxA_mat);
    vector<vector<string> >().swap(seqM_mat);
    vector<vector<string> >().swap(seqyA_mat);
    vector<string>().swap(tmp_str_vec);

    delete [] assign1_init;
    delete [] assign2_init;
    DeleteArray(&TMave_init,chain1_num);
    vector<vector<string> >().swap(seqxA_init);
    vector<vector<string> >().swap(seqyA_init);

    vector<vector<vector<double> > >().swap(xa_vec); // structure of complex1
    vector<vector<vector<double> > >().swap(ya_vec); // structure of complex2
    vector<vector<char> >().swap(seqx_vec); // sequence of complex1
    vector<vector<char> >().swap(seqy_vec); // sequence of complex2
    vector<vector<char> >().swap(secx_vec); // secondary structure of complex1
    vector<vector<char> >().swap(secy_vec); // secondary structure of complex2
    mol_vec1.clear();       // molecule type of complex1, RNA if >0
    mol_vec2.clear();       // molecule type of complex2, RNA if >0
    vector<string>().swap(chainID_list1);  // list of chainID1
    vector<string>().swap(chainID_list2);  // list of chainID2
    xlen_vec.clear();       // length of complex1
    ylen_vec.clear();       // length of complex2
    vector<string>().swap(chain1_list);
    vector<string>().swap(chain2_list);
    vector<string>().swap(sequence);

    t2 = clock();
    float diff = ((float)t2 - (float)t1)/CLOCKS_PER_SEC;
    printf("#Total CPU time is %5.2f seconds\n", diff);
    return 0;
}