SOAPdenovo2のscafStatisticsの統計情報を集計してtsv形式で出力するスクリプト

soapdenovo2_stats.rb

#/usr/bin/env ruby

require "optparse"

def v(s)
  if s == "NaN"
    return s
  elsif s.include?("%")
    return s.gsub("%","").to_f
  else
    return s.to_i
  end
end

def parse_scafstatistics(file)
  stats = []
  header = []
  open(file) { |f|
    header << "Filename"
    stats << File.basename(file)
    f.each_line do |line|
      if !line.include?("<--") && line != "\n"
        l = line.chomp.split("\t")
        if l[0] == "GC_Content"
          header << l[0]
          stats << v(l[1])
        elsif l.length == 3
          if l[2].include?("%")
            header << l[0].strip
            header << l[0].strip + " (%)"
            stats << v(l[1])
            stats << v(l[2])
          else
            header << l[0]
            header << "Contigs >0 in " + l[0]
            stats << v(l[1])
            stats << v(l[2])
          end
        elsif l[0] == "Average_number_of_contigs_per_scaffold"
          header << l[0]
          stats << l[1].to_f
        else
          header << l[0]
          stats << v(l[1])
        end
      end
    end
  }
  return header, stats
end

def print_tsv(header, stats, print_all, print_scaffold)
  h = []
  s = []
  if print_scaffold
    if print_all
      h = header[0..57]
      s = stats.map{|a| a[0..57]}
    else
      h = [header[0..8],header[45]]
      s = stats.map{|a| [a[0..8],a[45]]}
    end
  else
    if print_all
      h = header[58..header.length]
      s = stats.map{|a| a[58..a.length]}
    else
      h = [header[58..65].flatten,header[98]]
      s = stats.map{|a| [a[58..65].flatten,a[98]]}
    end
  end
  puts h.join("\t")
  s.each do |f|
    puts f.join("\t")
  end
end


if __FILE__ == $PROGRAM_NAME
  header = []
  stats = []

  sort_column = nil
  print_all = false
  print_scaffold = true

  ARGV.options do |opt|
    opt.on( "-a","--all") { print_all = true }
    opt.on( "-s VAL","--sort") { |a| sort_column = a }
    opt.on( "-c", "--contig") { print_scaffold = false }
    opt.on( "-h","--help") { puts opt ;exit }
    opt.parse!
  end

  Dir.glob(ARGV).each do |f|
    h, s = parse_scafstatistics(f)
    header = h
    stats << s
  end

  if sort_column
    i = header.index(sort_column)
    if i == nil
      puts "ERROR : unknown column name '#{sort_column}' "
      exit 1
    end
    stats = stats.sort{|a,b| b[i] <=> a[i] }
  end

  print_tsv(header, stats, print_all, print_scaffold)
end

http://yagays.github.com/blog/2012/10/11/soapdenovo2-stats/