From: spox <spox@...>
Date: 2009-09-11T00:33:33+09:00
Subject: Re: Timer implementation

On Thursday 10 September 2009 04:15:08 am Robert Larsen wrote:
> Hi all
>
> I am porting a timer class that I wrote in Java many years ago but I cannot
> see how to implement it in Ruby.
>
> What I want is a single thread, that sleeps forever or until the next timer
> block should be called or until a new timer block i added. The thread
> should be woken when a new timer block is added because it might be the
> next one to be called.
>
> I have implemented it like this:
>
> #!/usr/bin/ruby
>
> require 'thread'
> require 'bsearch'
>
> class Timer
>     class TimerObject
>         attr_reader :invoke_time, :block
>
>         def initialize(invoke_time, block)
>             @invoke_time, @block = invoke_time, block
>         end
>     end
>
>     def initialize
>         @timer_list = []
>         @mutex = Mutex.new
>         @condition = ConditionVariable.new
>         @running = false
>         start
>     end
>
>     def stop
>         running = false
>         @mutex.synchronize { @condition.signal }
>     end
>
>     def start
>         @running = true
>         Thread.new do
>             @mutex.synchronize do
>                 an_hour = 60 * 60 * 1000
>                 while running
>                     now_in_milliseconds = (Time.now.to_f * 1000).to_i
>                     milliseconds_to_sleep = @timer_list.empty? ? an_hour :
> @timer_list.first.invoke_time - now_in_milliseconds @condition.wait(@mutex,
> now_in_milliseconds) while @timer_list.empty?
>
>                     now_in_milliseconds = (Time.now.to_f * 1000).to_i
>                     while @time_list.empty? == false and
> @time_list.first.invoke_time < now_in_milliseconds
> @time_list.delete_at(0).block.call
>                     end
>                 end
>             end
>         end
>     end
>
>     def postpone (timeout, &block)
>         now = (Time.now.to_f * 1000).to_i
>         time = now + timeout
>
>         timerobject = TimerObject.new time, block
>
>         @mutex.synchronize do
>             idx = @timer_list.bsearch_lower_boundary { |x| x.invoke_time
> <=> time } @timer_list.insert(idx, timerobject) unless idx.nil?
>             @timer_list << timerobject if idx.nil?
>             @condition.signal
>         end
>     end
> end
>
>
> BUT...the 'start' method does not work since 'ConditionVariable' does not
> have a 'wait' method that takes a maximum wait time in milliseconds like we
> have in Java.
>
> Any ideas as to how this can be done ?
> Or is there a better way ?
>
> I really don't want a thread per timer.
>
> Best regards,
> Robert

Hi Robert. I have a simple timer that I built awhile ago. It uses a single 
thread for timed events, and a thread pool for executing blocks. If you are 
looking to build your own, the source can probably help give you some ideas.

Project:
http://rubyforge.org/projects/actiontimer

Source:
http://github.com/spox/actiontimer

- spox