forester/ruby/evoruby/lib/evo/io/parser/general_msa_parser.rb

   1 #
   2 # = lib/evo/io/parser/general_msa_parser - GeneralMsaParser class
   3 #
   4 # Copyright::  Copyright (C) 2006-2007 Christian M. Zmasek
   5 # License::    GNU Lesser General Public License (LGPL)
   6 #
   7 # $Id: general_msa_parser.rb,v 1.8 2009/10/08 22:44:54 cmzmasek Exp $
   8 #
   9 # last modified: 2009/10/08
  10
  11 require 'lib/evo/io/parser/msa_parser'
  12 require 'lib/evo/msa/msa'
  13
  14 require 'iconv'
  15
  16 module Evoruby
  17
  18     class GeneralMsaParser < MsaParser
  19
  20         def initialize
  21         end
  22
  23         def parse( path )
  24             Util.check_file_for_readability( path )
  25             block                       = -1
  26             current_seq_index_per_block = -1
  27             current_name                = nil
  28             saw_ignorable = true
  29             is_first      = true
  30             msa = Msa.new
  31             ic = Iconv.new( 'UTF-8//IGNORE', 'UTF-8' )
  32             File.open( path ) do | file |
  33                 while line = file.gets
  34                     line = ic.iconv( line )
  35                     if can_ignore?( line )
  36                         saw_ignorable = true
  37                     elsif ( is_first && is_program_name_line?( line ) )
  38                     elsif( line =~ /^\S+\s+.+\s*$/ || line =~ /^\s+.+\s*$/ || line =~ /^\S+\s*$/ )
  39                         if ( saw_ignorable )
  40                             block += 1
  41                             current_seq_index_per_block = -1
  42                             saw_ignorable = false
  43                         end
  44                         current_seq_index_per_block += 1
  45                         if ( line =~ /^(\S+)\s+(.+?)\s*$/ )
  46                             name = $1
  47                             seq  = $2.gsub( /\s/, '.' )
  48                             a = msa.find_by_name( name, false, false )
  49                             if ( a.length < 1 )
  50                                 msa.add( name, seq )
  51                             elsif ( a.length == 1 )
  52                                 msa.get_sequence( a[ 0 ] ).append!( seq )
  53                             else
  54                                 error_msg = "Unexpected error at line: " + line
  55                                 raise IOError, error_msg
  56                             end
  57                             current_name = name
  58                         elsif ( line =~ /^\s+(.+?)\s*$/ )
  59                             seq = $1.gsub( /\s/, '.' )
  60                             a = msa.find_by_name( current_name, false, false )
  61                             if ( a.length != 1  )
  62                                 error_msg = "Unexpected error at line: " + line
  63                                 raise IOError, error_msg
  64                             else
  65                                 msa.get_sequence( a[ 0 ] ).append!( seq )
  66                             end
  67
  68                         elsif ( line =~ /^(\S+)\s*$/ )
  69                             seq = $1
  70                             if block == 0
  71                                 error_msg = "First block cannot contain unnamed sequences"
  72                                 raise IOError, error_msg
  73                             else
  74                                 msa.get_sequence( current_seq_index_per_block ).append!( seq )
  75                             end
  76                             current_name = nil
  77                         end
  78                     else
  79                         error_msg = "Unexpected line: " + line
  80                         raise IOError, error_msg
  81                     end
  82                     if ( is_first )
  83                         is_first = false
  84                     end
  85                 end
  86             end
  87             return msa
  88         end # def parse( path )
  89
  90         private
  91
  92         def can_ignore?( line )
  93             return ( line !~ /[A-Za-z\-?\*_\.]/ ||
  94                      line =~ /^\s+[*\.:]/ ||
  95                      line =~ /^\s*#/ ||
  96                      line =~ /^\s*%/ ||
  97                      line =~ /^\s*\/\// ||
  98                      line =~ /^\s*!!/  )
  99         end
 100
 101         def is_program_name_line?( line )
 102             return ( line =~ /^CLUSTAL\s/ ||
 103                      line =~ /^MUSCLE\s\(/ ||
 104                      line =~ /^PROBCONS\s/ )
 105         end
 106     end # class GeneralMsaParser
 107
 108 end # module Evoruby