From: Dossy <dossy@...>
Date: 2002-05-10T04:57:16+09:00
Subject: Re: Thread#join doesn't accept a timeout?

On 2002.05.10, Nathaniel Talbott <nathaniel@talbott.ws> wrote:
> 
> Are the threads critical, or could you use the timeout extension?
> 
>    timeout_length = 3 # seconds
>    rd, wr = IO.pipe
>    begin
>      timeout(timeout_length) do
>        # do some stuff
>        wr.write "done"
>        wr.close
>      end
>    rescue TimeoutError
>    end
>    rd.close

I did try using the timeout extension, but this just doesn't
seem to work:

  timeout(@timeout) { send(@method_name) }

I'm busy right now, so wrapping my mind around why it isn't
working will have to wait until later.  It's probably something
simple, but I'm having a stupid moment right now.

> > Also, is there any interest in my Test::Unit::TimedTestCase?
> 
> Funny you should ask... I just ran in to a need to do some performance
> tests, here's what I came up with:
> 
> 	module Test
> 		module Unit
> 			module Assertions
> 				def assert_takes(milliseconds)
> 					start_time = Time.now
> 					yield
> 					end_time = Time.now
> 					actual_time = end_time.to_f -
> start_time.to_f
> 					assert(actual_time <
> milliseconds, "The operation took #{actual_time - milliseconds} ms
> longer than #{milliseconds} ms")
> 				end
> 			end
> 		end
> 	end
>
> Perhaps it's solving a slightly different problem, but it was nice to
> write my test without having to deal with a decorator. Of course, this
> doesn't provide a timeout like yours does, although I don't think it
> would be hard to add. I was/am considering adding something like this to
> Test::Unit, so any comments/suggestions/pros/cons/etc. would be greatly
> appreciated.

I thought about writing a new assertion as well, but didn't know
if that was "okay" or not.  I'm too used to open sourced projects
where getting changes to the "core" is a long, uphill battle, so
I'm used to having to write extensions that don't touch the
core.

Of course, I just realized yet another reason why Ruby rules ...
I can extend the core without having to actually touch any of
its files -- aah!  I'm going to write an assert_xpath assertion
to make my REXML-using test cases easier to write!  God, this is
great!

I guess we could implement #assert_takes_nowait that has a
timeout and all of my code can be thrown away.  :-)  What do
you think?

Nathaniel, I've posted my code (and unit tests) to this URL:

  http://panoptic.com/ruby

Check it out.  It could probably use some work and will be
much improved once the Thread#join patch makes it into stable
and it can be used instead of my pipes/select work-around.

I'll let others help refactor the code ... I'm sure something
more elegant than what I threw together will pop out that I
just can't see because I'm standing in the forest looking
at trees ...

-- Dossy

-- 
Dossy Shiobara                       mail: dossy@panoptic.com 
Panoptic Computer Network             web: http://www.panoptic.com/ 
  "He realized the fastest way to change is to laugh at your own
    folly -- then you can let go and quickly move on." (p. 70)