While implementing [Feature #18774] It appeared to me that adding this timeout was quite obvious.
And @matz (Yukihiro Matsumoto) explicitly asked for a separate ticket to be opened for that feature if someone wanted it: https://github.com/ruby/dev-meeting-log/blob/master/DevMeeting-2022-05-19.md#feature-18774-add-queuepoptimeout-eregon
SizedQueue#push already has a
non_block argument like
pop, so mirroring the change makes sense, and the implementations share a lot of code so it's really isn't that much maintainance.
As for the use case, I'd like to use a
SizedQueue in our
StatsD library, the code would look a bit like this (simplified code):
@event_queue = SizedQueue.new(MAX_BUFFER_SIZE) @dispatcher_thread = Thread.new do while event = event_queue.pop send_event(event) end end def emit_event(event) unless @event_queue.push(event, timeout: 0.1) if dispatcher_thread.alive? StatsD.logger.warn("Dropped event") else StatsD.logger.warn("Dispatcher thread died, restarting it") # ... end end end
But more generally, any action that may block forever (or for very long) should offer a timeout even if it's only used to be able to log that something is not supposed to happen an then retry.
Updated by byroot (Jean Boussier) 6 months ago
- Status changed from Open to Closed