From: Joel VanderWerf <vjoel@...>
Date: 2008-02-09T03:48:36+09:00
Subject: Re: Timeout::Error

Jeremy Thurgood wrote:
> Joel VanderWerf wrote:
>> Nobuyoshi Nakada wrote:
>>> Sorry, I can't remember it correctly now.  Perhaps, not to
>>> happen to be caught accidentally.
>>
>> That sounds like a good reason:
> 
> Not to me. I would expect catching StandardError (implicitly or 
> explicitly) would catch everything I have a reasonable chance of 
> recovering from, including timeouts. This is my standard strategy when I 
> am not entirely sure what exceptions may be thrown by third-party code. 
> There is code in the standard library that throws TimeoutErrors that 
> need to be caught explicitly:
> 
>   require 'open-uri'
>   begin
>     my_file = open("http://www.example.com/file")
>   rescue
>     puts "Sorry, I couldn't open the URL."
>   end
> 
> In this case, everything /except/ a timeout will be caught. This is 
> inconsistent and confusing behaviour to me, especially as it is not 
> properly documented anywhere that I have seen.

It not a question of _what_ is recoverable, but _where_ it should be 
recovered. A timeout should only be caught at a point where it's expected.

What if only the outer rescue, below, were allowed to catch the timeout?

begin
   Timeout.timeout do
     begin
     rescue # inner
     end
   end
rescue # outer
end

-- 
       vjoel : Joel VanderWerf : path berkeley edu : 510 665 3407