psmq is publish subscribe message queue. It's a set of programs and libraries to implement publish/subscribe way of inter process communication on top of POSIX message queue. This program targets mainly small embedded systems, which do not have access to some more sophisticated IPC (like unix domain socket). It could also prove useful when you want to realy heavly on message priority, since every message has set a priority and broker processes messages from highest ones first. It's main focues is memory and runtime safety, speed is treated as second class citizen - but of course is not ignored completely, it's just memory and safety comes first. Always.
Good, question. Why would you want to add more code, to your project, when you can do all these things using ordinary queues - after all psmq is implemented on top of mqueue. Well, if you only communicate processes in 1:1 way, you won't find this library useful - it's easy enough to create 2 queues and send/receive from them. No. Publish subscribe model shines best when you need to communicate N:N processes. That is for example, you have process1 that handles hardware and reads battery level via CAN. Now couple of processes might be interested in battery level, so you simply publish that information on "/battery/level" topic and any other processes might subscribe to it and read it. Some examples who could be interested in such messages:
- graphical user interface - for nice displaying battery status
- engine control unit - to shut down engine to prevent battery damage
- dignostic module for monitoring battery live and performance.
Note that these processes could also receive multiple battery information (if there are more than 1 battery). So instead of creating bunch of queues yourself and sending multiple messages to different processes, remembering who wants what, you can shift responsibility to receiving end, since application knows best what it needs, so it can just subscribe to messages, and your client will just send frame to broker and forget about it. Of course, you could do all of this in one process, or multiple threads that share such information easily, but it's poor design.
Library overview with usage explanation can be found in psmq_overview(7) manual page.
For detailed information about using broker and library, check out man pages.
Example of receiving message is in psmq_receive(3) manual page, and example of publishing message is in psmq_publish(3) manual page.
You can also study psmq-pub and psmq-sub source code.
All code follows ANSI C and POSIX standard (version 200112). External dependencies are:
Broker and psmq-sub:
- =embedlog-0.4 (psmq-0.1.0 only)
Library and psmq-pub program are dependency free.
Library, broker and companion programs have been tested on following systems/machines.
operating system tests
- arm-cortex-m3-nuttx-10.1 (manual)
Compiling and installing
Instruction on how to compile, install and integrate are in psmq_building manual page.
For building on Unix it's classic "./configure && make install"
Michał Łyszczek firstname.lastname@example.org
Library is licensed under BSD 2-clause license. See LICENSE file for details
- mtest unit test framework psmq uses
- git repository to browse code online
- continous integration for project
- polarhome nearly free shell accounts for virtually any unix there is.
- pvs studio static code analyzer with free licenses for open source projects