From: Robert Klemme <shortcutter@...>
Date: 2010-09-01T23:27:17+09:00
Subject: Re: Speed issues iterating over chars

On Wed, Sep 1, 2010 at 1:35 PM, Martin Hansen <mail@maasha.dk> wrote:
> Thanks Robert,
>
> The gsub solution seems to be reasonably efficient.
>
>>  seq.gsub! /./ do |m|
>>   scores[$`.length].ord - BASE_SOLEXA < cutoff ? m.downcase! : m
>> end
>
> But my original proposed naive loop is twice as fast:
>
>> scores.each_char do |score|
>>   seq[i] = seq[i].downcase if score.ord - BASE_SOLEXA <= cutoff
>>   i += 1
>> end
>
> I dont really know how gsub and tr compares to the Perl equivalents
> speed wise - in Perl tr is precompiling a lookup table that is evil fast
> and the regex engine is also primed at compile time and runs extremely
> fast.

Regexp is precompiled but I suspect that tr works at runtime only.
For definitive answer you'll have to look at the source.

> I suspect that you need some C extension to go faster than this, but I
> don't really want to spend the time on that. I was exploring Inline C
> but that appears very fragile - I cannot even get the example from the
> cookbook up and running under Ruby 191/192.
>
> Looking at your last proposal I see three iterations where two are
> running narrow if loops. I have not testet it, but it looks suspicious.

Well, but the caching should avoid that too many loops are executed.
I do not know however, how often you reuse values.  If you need this
in several processes you could save the current state in a file via
Marshal which is quite fast.

Kind regards

robert

-- 
remember.guy do |as, often| as.you_can - without end
http://blog.rubybestpractices.com/