From: Robert Klemme <shortcutter@...>
Date: 2009-03-05T21:33:35+09:00
Subject: Re: Cut a string each 80 character

--------------090603070207090109050701
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

On 05.03.2009 12:32, lasitha wrote:
> On Thu, Mar 5, 2009 at 2:00 PM, Gennady Bystritsky
> <Gennady.Bystritsky@quest.com> wrote:
>>> -----Original Message-----
>>> From: Leo [mailto:minilith@gmail.com]
>>> Sent: Wednesday, March 04, 2009 9:53 PM
>>>
>>>> string.scan(%r{.{1,80}}).map { |_chunk|
>>>>   _chunk + '<br />'
>>>> }
>>> Isn't that pretty much the same as:
>>>
>>> string.gsub(%r{.{1,80}}, "\\0<br />")
>> [...]
>> The latter may be faster, though.
> 
> Doesn't appear to be.
> http://gist.github.com/74249
> 
> I threw in a primitive insert loop for comparison and was a little
> surprised to find it considerably slower than both the others.
> 
> ruby 1.9.1p0 (2009-01-30 revision 21907) [i386-darwin8.11.1]
>       user     system      total        real
> with_scan    0.750000   0.040000   0.790000 (  0.851104)
> with_gsub    0.840000   0.060000   0.900000 (  0.950050)
> with_insert  2.000000   0.050000   2.050000 (  2.210962)


There is an inconsistency in your test: the gsub version creates the 
second argument to gsub over and over again.  You should rather define a 
constant for that as well.  You could do that as well for the sanity 
check string.

With the attached changed script this is what I get:

[robert@ora01 ~]$ allruby bm.rb
ruby 1.8.5 (2006-08-25) [i386-linux]
Rehearsal -----------------------------------------------
with_scan     1.770000   0.040000   1.810000 (  1.821903)
with_gsub     1.720000   0.010000   1.730000 (  1.750455)
with_insert   9.230000   0.000000   9.230000 (  9.228760)
------------------------------------- total: 12.770000sec

                   user     system      total        real
with_scan     1.730000   0.010000   1.740000 (  1.741069)
with_gsub     1.730000   0.010000   1.740000 (  1.741975)
with_insert   9.130000   0.000000   9.130000 (  9.130356)
ruby 1.9.1p0 (2009-01-30 revision 21907) [i686-linux]
Rehearsal -----------------------------------------------
with_scan     1.040000   0.010000   1.050000 (  1.043332)
with_gsub     1.100000   0.010000   1.110000 (  1.120970)
with_insert   9.030000   0.000000   9.030000 (  9.077018)
------------------------------------- total: 11.190000sec

                   user     system      total        real
with_scan     1.030000   0.000000   1.030000 (  1.031019)
with_gsub     1.090000   0.010000   1.100000 (  1.105150)
with_insert   9.040000   0.000000   9.040000 (  9.049539)
[robert@ora01 ~]$

CentOS 5.2 in VM.

Kind regards

	robert

--------------090603070207090109050701
Content-Type: text/plain;
 name="bm.rb"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline;
 filename="bm.rb"

require 'benchmark'
 
S = ('0123456789' * (8 * 1_000)).freeze
R = '<br/>'.freeze
GS = "\\&#{R}".freeze
SANE = '<br/>0123456789'.freeze
REP = 500
DELTA = 80 + R.length
 
def with_scan s
  s = s.scan(/.{80}/).join(R)
  raise 'with_scan failed' unless s[-85..-71] == SANE
  s
end
 
def with_gsub s
  s = s.gsub(/.{80}/, GS).chomp!(R)
  raise 'with_gsub failed' unless s[-85..-71] == SANE
  s
end
 
def with_insert s
  s = s.dup
  i = 80
  while i < s.length
    s.insert i, R
    i += DELTA
  end
  raise 'with_insert failed' unless s[-85..-71] == SANE
  s
end
    
 
Benchmark.bmbm do |bm|
  bm.report('with_scan ') { REP.times { with_scan S }}
  bm.report('with_gsub ') { REP.times { with_gsub S }}
  bm.report('with_insert') { REP.times { with_insert S }}
end

--------------090603070207090109050701--