Multiprocess threaded ctags generation with ripper-tags

ctags.rb

#!/usr/bin/env ruby
require 'etc'
require 'shell'

def ruby_files
  @ruby_files ||= %x{git ls-files | grep 'rb$'}.lines.map!(&:chomp)
end

dir = %x{git rev-parse --git-dir}.chomp

processors = Etc.nprocessors

puts "Number of processors: #{processors}"

total_processes = [processors.size, ruby_files.size].min
files_per_process = (ruby_files.size.to_f / total_processes).ceil.to_i
files_split = ruby_files.each_slice(files_per_process).to_a

threads = total_processes.times.map do |i|
  Thread.new do
    puts "Process #{i}"
    tag_file = "#{dir}/#{i}.tags"

    sh = Shell.new
    sh.debug = false
    sh.transact do
      echo(files_split[i].join("\n")) |
        system("ripper-tags --exclude \"vendor\" --tag-relative -L - -f#{tag_file}")
    end
  end
end

threads.each(&:join)

# Merge and sort
abs_dir = File.absolute_path(dir)
src_tag_files = total_processes.times.map { |i| "#{abs_dir}/#{i}.tags" }
dst_tag_file = "#{abs_dir}/tags"
command = "cat #{src_tag_files.join(' ')} | sort -u > #{dst_tag_file}"

sh = Shell.new
sh.transact do
  sh.system(command)
end

# Cleanup
src_tag_files.each { |f| File.unlink(f) }