X-Git-Url: http://source.jalview.org/gitweb/?a=blobdiff_plain;f=forester%2Fruby%2Fevoruby%2Flib%2Fevo%2Ftool%2Fmulti_sequence_extractor.rb;h=40b1211ffca6858fe65d39692eeef35a7e0ac934;hb=205d83c44194c54ae2165623f6de89c6f4fdea11;hp=b499c7262fa555d38560d9656740d58e5bdad257;hpb=dbf5b588d65d1c62094dd5d339eca5056a5ade5f;p=jalview.git

diff --git a/forester/ruby/evoruby/lib/evo/tool/multi_sequence_extractor.rb b/forester/ruby/evoruby/lib/evo/tool/multi_sequence_extractor.rb
index b499c72..40b1211 100644
--- a/forester/ruby/evoruby/lib/evo/tool/multi_sequence_extractor.rb
+++ b/forester/ruby/evoruby/lib/evo/tool/multi_sequence_extractor.rb
@@ -23,12 +23,12 @@ module Evoruby
   class MultiSequenceExtractor
 
     PRG_NAME                           = "mse"
-    PRG_VERSION                        = "1.02"
+    PRG_VERSION                        = "1.03"
     PRG_DESC                           = "extraction of sequences by name from multiple multi-sequence ('fasta') files"
-    PRG_DATE                           = "2012.07.20"
-    COPYRIGHT                          = "2008-2012 Christian M Zmasek"
+    PRG_DATE                           = "131127"
+    COPYRIGHT                          = "2008-2013 Christian M Zmasek"
     CONTACT                            = "phylosoft@gmail.com"
-    WWW                                = "www.phylosoft.org"
+    WWW                                = "https://sites.google.com/site/cmzmasek/home/software/forester"
     HELP_OPTION_1                      = 'help'
     HELP_OPTION_2                      = 'h'
 
@@ -41,6 +41,10 @@ module Evoruby
     NORMALIZED_IDS_MAP_SUFFIX           = ".nim"
     PROTEINS_LIST_FILE_SEPARATOR        = "\t"
 
+    def initialize()
+      @file_to_msa = Hash.new
+      @seqs = 0
+    end
 
     def run()
 
@@ -245,6 +249,7 @@ module Evoruby
       puts basename
 
       File.open( input_file ) do | file |
+        species_counter = 1
         while line = file.gets
           line.strip!
           if !Util.is_string_empty?( line ) && !(line =~ /\s*#/ )
@@ -254,9 +259,9 @@ module Evoruby
               Util.fatal_error( PRG_NAME, "unexpected format: " + line )
             end
             species = values[ 0 ]
-            if species == "BRADI" || species == "ASPNG" || species == "SCLSC" || species == "PTEVA"  || species == "EIMTE"
-              next
-            end
+            #if species == "BRADI" || species == "ASPNG" || species == "SCLSC" || species == "PTEVA"  || species == "EIMTE"
+            #  next
+            #end
             seq_name = values[ 1 ]
             domain_ranges = nil
             if ( values.length > 3 )
@@ -286,35 +291,50 @@ module Evoruby
                 print_counts( per_species_counter, log, Constants::LINE_DELIMITER )
                 per_species_counter = 0
               end
-              puts " " + current_species + " [" + my_readlink + "]"
-              log << current_species + " [" + my_readlink + "]" + Constants::LINE_DELIMITER
+              puts " " + species_counter.to_s +  ":" + current_species + " [" + my_readlink + "]"
+              log << species_counter.to_s +  ": " + current_species + " [" + my_readlink + "]" + Constants::LINE_DELIMITER
+              species_counter += 1
             end
             puts "   " + seq_name
             log << "   " + seq_name + Constants::LINE_DELIMITER
             per_species_counter = per_species_counter + 1
             seq = nil
 
-            if current_msa.find_by_name_start( seq_name, true ).size > 0
-              begin
-                seq = current_msa.get_by_name_start( seq_name, true ).copy
-              rescue ArgumentError => e
-                Util.fatal_error( PRG_NAME, "error: " + e.to_s )
-              end
-            else
+            indices = current_msa.find_by_name_start( seq_name, true )
+            if indices.size == 1
+              seq =  current_msa.get_sequence( indices[ 0 ] )
+            elsif indices.size == 0
               # Not found, try finding by partial match.
               begin
                 seq = current_msa.get_by_name( seq_name, true, true )
               rescue ArgumentError => e
                 Util.fatal_error( PRG_NAME, "error: " + e.to_s )
               end
+            else
+              Util.fatal_error( PRG_NAME, "error: seq name \"" + seq_name + "\" not unique"  )
             end
 
+            # if current_msa.find_by_name_start( seq_name, true ).size > 0
+            #   begin
+            #     seq = current_msa.get_by_name_start( seq_name, true ).copy
+            #   rescue ArgumentError => e
+            #     Util.fatal_error( PRG_NAME, "error: " + e.to_s )
+            #   end
+            # else
+            #   # Not found, try finding by partial match.
+            #   begin
+            #     seq = current_msa.get_by_name( seq_name, true, true )
+            #   rescue ArgumentError => e
+            #     Util.fatal_error( PRG_NAME, "error: " + e.to_s )
+            #   end
+            # end
+
             normalized_id = per_species_counter.to_s( 16 ).upcase +
              "_" + current_species
 
             per_species_counter.to_i
 
-            ids_map_writer.write( normalized_id + ": " + seq.get_name + Constants::LINE_DELIMITER )
+            ids_map_writer.write( normalized_id + "\t" + seq.get_name + Constants::LINE_DELIMITER )
 
             orig_name = nil
             if seq != nil
@@ -450,8 +470,6 @@ module Evoruby
           Util.fatal_error( PRG_NAME, "error: " + e.to_s )
         end
       end
-
-
     end
 
 
@@ -523,6 +541,10 @@ module Evoruby
     end
 
     def read_fasta_file( input )
+      if @file_to_msa.has_key?( input )
+        return @file_to_msa[ input ]
+      end
+
       f = MsaFactory.new()
       msa = nil
       begin
@@ -530,6 +552,11 @@ module Evoruby
       rescue Exception => e
         Util.fatal_error( PRG_NAME, "error: " + e.to_s )
       end
+      if @seqs <= 100000000
+        @file_to_msa[ input ] = msa
+        @seqs += msa.get_number_of_seqs
+        puts "   total seqs in memory: " + @seqs.to_s
+      end
       msa
     end