From: Robert Klemme <shortcutter@...>
Date: 2011-11-18T01:40:03+09:00
Subject: Re: Webcrawler that become enormous

On Thu, Nov 17, 2011 at 5:14 PM, Lucas Panthe <panthe@libero.it> wrote:
> Hi,
> I've made a script in Ruby 1.8 that use the gems mechanize, nokogiri and
> open-uri and run under Linux.
>
> This script is a webcrawler that scan a bigger site that contains a big
> amount of data of international firm.
>
> I'm interesting in create a db with only some data and not the full
> data.
> Mine script run perfectly and grab all data in the correct order but
> after 6-8 hour that the script run the amount of memory that use is
> enormous (1gb).
>
> I save in a file the results of scraping and empty the buffer of data
> every 10 firm collect.

It seems either you do not free the memory (and thus have created a
leak yourself) or you suffer from the mentioned bug.

> I've follow this post ofr obtain this results because before the script
> used this amount of memory just after 4hour.
> http://stackoverflow.com/questions/181406/ruby-memory-management
>
> Someone can help me to reduce this problem and optimize this script?
> Exist an IDE that make an efficient debug for ruby?
> I think that there is something that I've missed.

First thing I'd do is to update Ruby version to a more recent 1.9.*
version.  That will be faster also and likely has a fix for the
leakage bug mentioned on the stackoverflow page.  If your problem
persists, you need to look into your code.

A simple test would be to write out statistics per class on a regular
basis, e.g.

cnt = Hash.new 0
ObjectSpace.each_object(BasicObject) {|o| cnt[o.class] += 1}
cnt.sort_by {|cl,c| cl.to_s}.each {|cl,c| printf "%10d %s\n", c, cl}

Then compare counts per class.  Of course, you can get a bit more
fancy and calculate deltas etc.  But then there are better tools
around, I guess.

Kind regards

robert

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