Steffen Müller > ZMQ-Declare-0.03 > ZMQ::Declare



Annotate this POD


Open  0
View/Report Bugs
Module Version: 0.03   Source  


ZMQ::Declare - Declarative 0MQ Infrastructure


  use ZMQ::Declare;
  # Read the network "topology" (who talks to whom and how) from a shared
  # file or alternatively, provide an equivalent nested Perl data structure.
  my $spec = ZMQ::Declare::ZDCF->new(tree => 'mynetwork.zdcf');
  # Pick the device in your network that this code path is to implement
  my $broker = $spec->application("events")->device("event_broker");
  # Set up your main loop
  $broker->implementation( sub {
    my ($runtime) = @_;
    my $input_sock = $runtime->get_socket_by_name("event_listener");
    my $output_sock = $runtime->get_socket_by_name("work_distributor");
    while (1) {
      ... recv, send, recv, send ...
  # Kick it off. This will create the actual 0MQ objects, make
  # connections, configure them, potentially fork off many processes,
  # and then hand control to your main loop with everything set up!
  # If this was not the broker but the implementation for the event processors:
  #$worker->run(nforks => 20);

Actual, runnable examples can be found in the examples/ subdirectory of the ZMQ::Declare distribution.


This is experimental software. Interfaces and implementation are subject to change. If you are interested in using this in production, please get in touch to gauge the current state of stability.

One guaranteed user-visible change will be that the underlying libzmq wrapper will be switched from to (with ZMQ::LibZMQ2 or 3 as backend) when becomes stable.

0MQ is a light-weight messaging library built on TCP.

The Perl module ZMQ::Declare aims to provide a declarative and/or configuration-driven way of establishing a network of distributed processes that collaborate to perform a certain task. The individual processes ("applications" in ZMQ::Declare) can each have one or more threads ("devices" in 0MQ speak) which talk to one another using 0MQ. For example, such a setup could be an entire event processing stack that has many different clients producing events, a broker, many event processing workers, and a result aggregator.

Normally using the common Perl binding, ZeroMQ, requires you to explicitly write out the code to create 0MQ context and sockets, and to write the connect/bind logic for each socket. Since the use of 0MQ commonly implies that multiple disjunct piece of code talk to one another, it's easy to either scatter this logic in many places or re-invent application-specific network configurations. (Which side of the connection is supposed to bind() and which is supposed to connect() again?) For what it's worth, I've always felt that the networked components that I've written were simply flying in close formation instead of being obvious parts of a single stack.

ZMQ::Declare is an attempt to concentrate the information about your network of 0MQ sockets and connections in one place, to create and connect all sockets for you, and to allow you to focus on the actual implementation of the various devices that talk to one another using 0MQ. It turns out that I am not the only one who thought this would come in useful: defines a standard device configuration data structure (it does not specify encoding format) called ZDCF (ZeroMQ Device Configuration File).

Despite the name ZDCF, there's no technical need for this information to live in a file. ZMQ::Declare implements ZDCF file reading/decoding (parsing) as well as some degree of validation. This is implemented in the ZMQ::Declare::ZDCF class which represents a single such configuration. The default decoder/encoder assumes JSON input/output, but is pluggable.

The envisioned typical use of ZMQ::Declare is that you write a single ZDCF specification file or data structure that defines various applications and devices in your network and how they interact with one another. This approach means that as long as you have a library to handle ZDCF files, you can write your devices in a multitude of programming languages and mix and match to your heart's content. For example, you might choose to implement your tight-loop message broker in C for performance, but prefer to write the parallelizable worker components in Perl for ease of development.

For details on the ZDCF format, please refer to ZMQ::Declare::ZDCF. For a domain specific language for defining ZDCF structures in pure Perl, see ZMQ::Declare::DSL.


ZMQ::Declare::ZDCF, ZMQ::Declare::Application, ZMQ::Declare::Device, ZMQ::Declare::Device::Runtime, ZMQ::Declare::DSL



Steffen Mueller <>


Copyright (C) 2011,2012,2014 by Steffen Mueller

This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself, either Perl version 5.8.1 or, at your option, any later version of Perl 5 you may have available.

syntax highlighting: