From: kamphausen@... (SKa)
Date: 2001-08-24T23:52:26+09:00
Subject: [ruby-talk:20290] Performance of Substrings

Dear Ruby Community,

I got a problem at hand where I enhanced the Time class to understand
an ISO like time notation (w/o the dashes): 19910917 for the 17th of
September in 1991. Additionally you might provide a string which also
contains the hours, minutes and seconds: 19910917002055 (55 seconds
after 2:00 o'clock AM).

This is how I implemented it

File: isotime.rb
-----------------------------------
class Time
  def Time.iso(str)
    Time.gm(*iso_unformat(str))
  end
end

def iso_unformat(t)
  if t.size == 8
    [ t[0,4], t[4,2], t[6,2] ]
  elsif t.size == 14
    [ t[0,4], t[4,2], t[6,2], t[8,2], t[10,2], t[12,2] ]  
  end
end
-----------------------------------
end_of_file: isotime.rb
(I hope there are no typos in this: I can't actually paste the code)

The problem is that using the profiler (require 'profile') I found
that this
routine uses a LOT of time. In my case it makes up 30% of the whole
run, but of course this depends on what is done. It seems as if the
substring operator [] would need a lot of time, and since it is used
fairly often here this really shows in overall performance.

My question is if there is any reasonably faster way to achieve the
same goal. I'd like to use Time objects because I do a lot of
comparisons and intervall detections afterwards.


Thanks and Regards
Stefan