From: why the lucky stiff <ruby-talk@...>
Date: 2002-12-11T14:58:28+09:00
Subject: Re: replacing chars in string

Shashank Date (sdate@kc.rr.com) wrote:
> I am trying to globally replace characters in a string by "chaining" the
> String#gsub method like so:
> 
> str =  "count=(a+b); if(a<=b) d[i]=10;f(a,b)"
> puts str.gsub(/,/,' , ').gsub(/\(/,' ( ').gsub(/\)/,' )
> ').gsub(/([^<>])=/,'\1 = ').gsub(/;/,";\n")
> 
> Is there a better (faster) way to do this ?

For your need, the chained gsub is the fasted I can think of.  Here's my
comparison against two solutions involving the block form of gsub.

I'm using ruby 1.7.3 (2002-11-27) [i586-linux].

  require 'benchmark'
  include Benchmark
  
  n = 1000
  str =  "count=(a+b); if(a<=b) d[i]=10;f(a,b)"
  test = nil
  bm do |x|
      x.report( "chained gsubs:" ) do
          n.times do
              test = str.gsub(/,/,' , ').gsub(/\(/,' ( ').gsub(/\)/,' ) ').gsub(/([^<>])=/,'\1 = ').gsub(/;/,";\n")  
          end
          # puts
          # p test
      end
      x.report( "grouped gsub:" ) do
          n.times do
              test = str.gsub( /(?:(,|\(|\)|[^<>]=)|(;))/ ) do 
                  if $1
                      "#{ $1[0..-2] } #{ $1[-1,1] } "
                  else
                      "#{ $2 }\n"
                  end
              end
          end
          # puts
          # p test
      end
      x.report( "gsub and case:" ) do
          n.times do
              test = str.gsub( /[,();]|[^<>]=/ ) do |m|
                  case m
                      when /=$/
                          "#{ $` } = "
                      when /;/
                          ";\n"
                      else
                          " #{ m } "
                  end
              end
          end
          # puts
          # p test
      end
  end

Benchmark results:

                      user     system      total        real
  chained gsubs:  0.830000   0.000000   0.830000 (  0.831508)
  grouped gsub:   2.020000   0.000000   2.020000 (  2.022475)
  gsub and case:  1.270000   0.000000   1.270000 (  1.277238)

_why