From: Charles Oliver Nutter <charles.nutter@...>
Date: 2009-05-16T07:05:58+09:00
Subject: [ruby-core:23467] Re: REXML Thread Safety

FWIW, most DOM implementations are not thread-safe.

Weston Sewell wrote:
> I've been using Ruby for several years but I just joined the list 
> because I ran into a problem that I'm not sure how to solve (sorry for 
> not lurking for the requisite eon, I've been working night and day on a 
> Ruby project for work).
> 
> It appears that REXML may not be completely thread-safe. Is it intended 
> to be? I've run into a problem that I was able to reproduce in a test 
> that I'll post at the bottom. Basically, it appears that accessing 
> either #to_s or #value on REXML::Attribute causes a change in the state 
> of the instance (setting @normalized and @unnormalized to a computed 
> value or 'nil'). And since these are not intended to be "destructive" 
> calls, neither are some of the Element methods that call them (i.e. 
> Element#inspect). When using threads of course we are not protecting 
> non-destructive calls with semaphors (and actually we are using XMPP4r 
> so we don't have complete conrol over what gets called and how).
> 
> Is there anyone who is familiar with this situation?
> 
> TIA
> Weston
> 
> Test Code  - exception occurs in either #to_s or #value (on my system 
> this always happens within 900 iterations)
> ------------------------------------------------------------------------------
> common_doc = REXML::Document.new( '<foo bar="somevalue" />' )
> 
> def parse(doc)
>     foo = doc.root
>     foo.inspect
>     foo.attributes['bar']
> end
> 
> def multiparse(doc)
>     1000000.times do |iteration|
>         puts "Iteration #{iteration}" if(iteration == 0 || (iteration % 
> 100 == 0))
>         parse(doc)
>     end
> end
> 
> threads = []
> threads << Thread.new { multiparse(common_doc) }
> threads << Thread.new { multiparse(common_doc) }
> threads << Thread.new { multiparse(common_doc) }
> threads << Thread.new { multiparse(common_doc) }
> threads << Thread.new { multiparse(common_doc) }
> threads << Thread.new { multiparse(common_doc) }
> 
> threads.each do |thread|
>     thread.join
> end
>