From: Robert Klemme <shortcutter@...>
Date: 2006-09-08T15:35:41+09:00
Subject: Re: Compiling Regexp only once

--------------060104090601070606060308
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

Rick DeNatale wrote:
> On 9/4/06, Robert Klemme <shortcutter@googlemail.com> wrote:
> 
>> Note though that it's usually faster to have a regexp in line.  So in
>> your case you might have a method that does the line parsing (or
>> multiple line parsing) and that's where you can put the inline regexp
>> for max efficiency.
> 
> Robert,
> 
> I'm not sure what you mean by inline, do you mean a regexp literal like
> /[ABC]/ vs. RegEx.new('/[ABC]/')  if so that's not clear.

No, I meant that

if /foo/ =~ something

is faster than

FOO = /foo/
...
if FOO =~ something

Maybe the term "inline" was confusing.  It refers to the fact that the 
RX is defined at the place where it matches.

$ ruby rxbm.rb
Rehearsal ----------------------------------------------------
inline match       0.781000   0.000000   0.781000 (  0.781000)
inline non match   1.500000   0.000000   1.500000 (  1.500000)
const match        0.969000   0.000000   0.969000 (  0.985000)
const non match    1.719000   0.000000   1.719000 (  1.718000)
------------------------------------------- total: 4.969000sec

                        user     system      total        real
inline match       0.797000   0.000000   0.797000 (  0.797000)
inline non match   1.484000   0.000000   1.484000 (  1.485000)
const match        1.000000   0.000000   1.000000 (  1.000000)
const non match    1.719000   0.000000   1.719000 (  1.719000)


(BM attached)

Kind regards

	robert

--------------060104090601070606060308
Content-Type: text/plain;
 name="rxbm.rb"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline;
 filename="rxbm.rb"


require 'benchmark'
include Benchmark

REP = 1_000_000
RX = /foo/
MATCH = ("foo" * 100).freeze
NON_MATCH = ("fo" * 100).freeze

bmbm do |b|
  b.report "inline match" do
    REP.times do
      /foo/ =~ MATCH
    end
  end

  b.report "inline non match" do
    REP.times do
      /foo/ =~ NON_MATCH
    end
  end

  b.report "const match" do
    REP.times do
      RX =~ MATCH
    end
  end

  b.report "const non match" do
    REP.times do
      RX =~ NON_MATCH
    end
  end

end


--------------060104090601070606060308--