From: Mike Steiner <mikejaysteiner@...>
Date: 2007-05-25T00:14:06+09:00
Subject: Re: optimizing for speed - Array#each

------=_Part_80053_3014564.1180019646342
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

One of the things I'm doing is a lot of File.join's, and I have to do a gsub
( /\// , "\\" ) after each one (because I'm running under Windows).

I can't use sets because I have duplicate records. There is no unique record
key, so for each record (which I use as a key in a hash), I have an array
listing the files it appears in (which can include the same file more than
once).

Thanks for the help!

Mike Steiner

On 5/24/07, Robert Klemme <shortcutter@googlemail.com> wrote:
>
> On 24.05.2007 00:15, Stefan Rusterholz wrote:
> > Mike Steiner wrote:
> >> I ran the profiler on my program and it said that 51% of the time is
> >> spent
> >> in Array#each. Is there any way to speed this up (like using
> >> Array#each_index perhaps)?
>
> Note that the profiler output is often irritating because time for each
> also includes all the time spend in the block - and while #each is
> invoked only once the block is invoked n times.
>
> >> And is there a quicker way to do a .gsub if I'm using just strings and
> >> not a regexp?
>
> Quicker than what?  Are you using the block form of gsub?  Are you using
> gsub! or gsub?  Can you show some code?
>
> >> By the way, the program compares 2 databases and determines which
> >> records
> >> have been inserted and deleted, and each database has about 15,000
> >> records
> >> to compare.
>
> 15,000 does not sound like much.  What is the runtime of the program
> without profile?
>
> And another idea: could be that you are using inefficient methods to
> test for existence.  Finding something in an Array is always O(n) unless
> you have additional constraints; i.e. array content is sorted in which
> case you can use binary search which is O(log(n)).  It seems you want to
> be doing set operations - so why don't you try to use Set for which
> member tests are far more efficient (O(1) because it's using hashing
> internally).
>
> Kind regards
>
>         robert
>
>

------=_Part_80053_3014564.1180019646342--