forester/ruby/evoruby/lib/evo/tool/fasta_extractor.rb

   1 #
   2 # = lib/evo/apps/fasta_extractor.rb - FastaExtractor class
   3 #
   4 # Copyright::    Copyright (C) 2017 Christian M. Zmasek
   5 # License::      GNU Lesser General Public License (LGPL)
   6
   7 require 'lib/evo/util/constants'
   8 require 'lib/evo/util/util'
   9 require 'lib/evo/util/command_line_arguments'
  10
  11 module Evoruby
  12   class FastaExtractor
  13
  14     PRG_NAME       = "fastx"
  15     PRG_VERSION    = "1.000"
  16     PRG_DESC       = "extraction of molecular sequences from a fasta file"
  17     PRG_DATE       = "170215"
  18     WWW            = "https://sites.google.com/site/cmzmasek/home/software/forester"
  19
  20     HELP_OPTION_1                      = 'help'
  21     HELP_OPTION_2                      = 'h'
  22     def run()
  23
  24       Util.print_program_information( PRG_NAME,
  25       PRG_VERSION,
  26       PRG_DESC ,
  27       PRG_DATE,
  28       WWW,
  29       STDOUT )
  30
  31       if ( ARGV == nil || ( ARGV.length < 1 )  )
  32         print_help
  33         exit( -1 )
  34       end
  35
  36       begin
  37         cla = CommandLineArguments.new( ARGV )
  38       rescue ArgumentError => e
  39         Util.fatal_error( PRG_NAME, "error: " + e.to_s )
  40       end
  41
  42       if ( cla.is_option_set?( HELP_OPTION_1 ) ||
  43       cla.is_option_set?( HELP_OPTION_2 ) )
  44         print_help
  45         exit( 0 )
  46       end
  47
  48       if ( cla.get_number_of_files != 3 )
  49         print_help
  50         exit( -1 )
  51       end
  52
  53       allowed_opts = Array.new
  54
  55       disallowed = cla.validate_allowed_options_as_str( allowed_opts )
  56       if ( disallowed.length > 0 )
  57         Util.fatal_error( PRG_NAME,
  58         "unknown option(s): " + disallowed,
  59         STDOUT )
  60       end
  61
  62       input_file  = cla.get_file_name( 0 )
  63       query       = cla.get_file_name( 1 )
  64       output_file = cla.get_file_name( 2 )
  65
  66       if  !File.exist?( input_file )
  67         Util.fatal_error( PRG_NAME, "error: input file [#{input_file}] does not exist" )
  68       end
  69       if File.exist?( output_file   )
  70         Util.fatal_error( PRG_NAME, "error: [#{output_file}] already exists" )
  71       end
  72
  73       results = extract_sequences( query, input_file, output_file )
  74
  75       Util.print_message( PRG_NAME, "matched: " + results )
  76       Util.print_message( PRG_NAME, "wrote:   " + output_file )
  77       Util.print_message( PRG_NAME, "OK" )
  78
  79     end
  80
  81     def extract_sequences( query, fasta_file, output_file )
  82       output = File.open( output_file, "a" )
  83       matching_state = false
  84       matches = 0
  85       total = 0
  86       File.open( fasta_file ) do | file |
  87         while line = file.gets
  88           if !Util.is_string_empty?( line )
  89             if line =~ /^\s*>/
  90               total += 1
  91               if total % 10000 == 0
  92                 STDOUT.write "\r#{matches}/#{total}"
  93                 STDOUT.flush
  94               end
  95               if line =~ /#{query}/
  96                 matching_state = true
  97                 matches += 1
  98                 output.print( line )
  99               else
 100                 matching_state = false
 101               end
 102             elsif matching_state
 103               output.print( line )
 104             end
 105           end
 106         end
 107       end
 108       output.close()
 109       matches.to_s + "/" + total.to_s
 110     end
 111
 112     def print_help()
 113       puts( "Usage:" )
 114       puts()
 115       puts( "  " + PRG_NAME + ".rb <input fasta file> <query> <output file>" )
 116       puts()
 117       puts( "Examples:" )
 118       puts
 119       puts( "  " + PRG_NAME + ".rb Pfam-A.fasta kinase kinases" )
 120       puts()
 121     end
 122
 123   end # class FastaExtractor
 124 end