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) 2017 Christian M. Zmasek
   5 # License::      GNU Lesser General Public License (LGPL)
   6 #
   7 # Last modified: 2017/02/07
   8
   9 require 'lib/evo/io/parser/msa_parser'
  10 require 'lib/evo/msa/msa'
  11
  12 #require 'iconv'
  13
  14 module Evoruby
  15
  16     class GeneralMsaParser < MsaParser
  17
  18         def initialize
  19         end
  20
  21         def parse( path )
  22             Util.check_file_for_readability( path )
  23             block                       = -1
  24             current_seq_index_per_block = -1
  25             current_name                = nil
  26             saw_ignorable = true
  27             is_first      = true
  28             msa = Msa.new
  29             ic = Iconv.new( 'UTF-8//IGNORE', 'UTF-8' )
  30             File.open( path ) do | file |
  31                 while line = file.gets
  32                     line = ic.iconv( line )
  33                     if can_ignore?( line )
  34                         saw_ignorable = true
  35                     elsif ( is_first && is_program_name_line?( line ) )
  36                     elsif( line =~ /^\S+\s+.+\s*$/ || line =~ /^\s+.+\s*$/ || line =~ /^\S+\s*$/ )
  37                         if ( saw_ignorable )
  38                             block += 1
  39                             current_seq_index_per_block = -1
  40                             saw_ignorable = false
  41                         end
  42                         current_seq_index_per_block += 1
  43                         if ( line =~ /^(\S+)\s+(.+?)\s*$/ )
  44                             name = $1
  45                             seq  = $2.gsub( /\s/, '.' )
  46                             a = msa.find_by_name( name, false, false )
  47                             if ( a.length < 1 )
  48                                 msa.add( name, seq )
  49                             elsif ( a.length == 1 )
  50                                 msa.get_sequence( a[ 0 ] ).append!( seq )
  51                             else
  52                                 error_msg = "Unexpected error at line: " + line
  53                                 raise IOError, error_msg
  54                             end
  55                             current_name = name
  56                         elsif ( line =~ /^\s+(.+?)\s*$/ )
  57                             seq = $1.gsub( /\s/, '.' )
  58                             a = msa.find_by_name( current_name, false, false )
  59                             if ( a.length != 1  )
  60                                 error_msg = "Unexpected error at line: " + line
  61                                 raise IOError, error_msg
  62                             else
  63                                 msa.get_sequence( a[ 0 ] ).append!( seq )
  64                             end
  65
  66                         elsif ( line =~ /^(\S+)\s*$/ )
  67                             seq = $1
  68                             if block == 0
  69                                 error_msg = "First block cannot contain unnamed sequences"
  70                                 raise IOError, error_msg
  71                             else
  72                                 msa.get_sequence( current_seq_index_per_block ).append!( seq )
  73                             end
  74                             current_name = nil
  75                         end
  76                     else
  77                         error_msg = "Unexpected line: " + line
  78                         raise IOError, error_msg
  79                     end
  80                     if ( is_first )
  81                         is_first = false
  82                     end
  83                 end
  84             end
  85             return msa
  86         end # def parse( path )
  87
  88         private
  89
  90         def can_ignore?( line )
  91             return ( line !~ /[A-Za-z\-?\*_\.]/ ||
  92                      line =~ /^\s+[*\.:]/ ||
  93                      line =~ /^\s*#/ ||
  94                      line =~ /^\s*%/ ||
  95                      line =~ /^\s*\/\// ||
  96                      line =~ /^\s*!!/  )
  97         end
  98
  99         def is_program_name_line?( line )
 100             return ( line =~ /^CLUSTAL\s/ ||
 101                      line =~ /^MUSCLE\s\(/ ||
 102                      line =~ /^PROBCONS\s/ )
 103         end
 104     end # class GeneralMsaParser
 105
 106 end # module Evoruby