Converting existing module/objects to threads

martdi martin.dion at
Thu Oct 19 04:40:48 CEST 2006

jdlists at wrote:
> I have inheirted some existing code, that i will explain in a moment,
> have needed to extend and ultimately should be able to run in threads.
> I've done a bunch of work with python but very little with threads and
> am looking for some pointers on how to implement, and if the lower
> level modules/objects need to be rewritten to use threading.local for
> all local variables.
> I have a module that communicates with a hardware device, which reads
> data off of sensors, that can only talk with one controller at a time.
> The controller (my module) needs to (in its simplest form) init,
> configure the device, request data, and write out xml, sleep, repeat.
> The new request is that the device needs to be queried until a
> condition is true, and then start requesting data.  So an instance of a
> controller needs to be deadicated to a hardware device forever, or
> until the program ends....which ever comes first.
> This currently works in a non-threaded version, but only for one device
> at a time, there is a need to create a single windows(yeach) service
> that talks to many of these devices at once.  I don't need worker
> threads that handle seperate portions of the entire job, i need a
> single application to spawn multiple processes to run through the
> entire communication from configure to report, sleep until the next
> interval time and run again.  The communication could last from 1
> minute to 10 minutes before it ends.
>   Here is the code layout in pseudocode.
> module.Object - controller.Main - handles all socket communications
> class subcontroller(controller.Main):
>   def __init__(self,id,configurationFile):
>    controller.Main.__init__(self)
>    // instantiate variables and local objects that handle
> configuration, logic and data output
>   def configure(self,configurationFile):
>     //read configurationFile and configure device
>   def process(self):
>     while 1:
>       //based on configuration file, query the device until condition
> is true and then write xml, sleep until time to repeat and run again.
> within controller there are 5 objects and subcontroller is a sinlge
> object that loads other objects from the inherited controller.System
> I'm trying to figure out how difficult it is going to be to convert
> this to a threaded application.  The original controller.Main is built
> to talk to devices in series, never in parallel. so no objects are
> considered to be thread safe, but no instance of any of the objects
> should need to share resources with any other instance of teh same
> object.  they would all have unique configuration files and talk to
> devices on unique ip/ports.
> on a unix system, forking,while potentially not optimal, would be a
> fine solution, unfortunantely this needs to run on windows.
> I know i have left out many details, but hopefully this is enough to at
> least enable some kind soles to lend an opinnion or two.
> many thanks
> jd

Taking a look at asyncore could be worthwhile, but if you want to
implement it with threads, you may be able to do it this way:

In your main file, from where you start the program, let's call it

Load Required configuration
spawn threads (1 for each controller)
define a queue object from module queue.queue used for communication
with threads
enter an infinite loop that checks for the conditions
    once conditions are met, notify the proper thread

class ControllerThread(threading.Thread):
    def __init__(self):
        define a queue here, to process messages from the main
        call threading.Thread.__init__(self)

    define method to load config for thread objects
    (you might want to pass an argument to init to load your configs
from a file)
    (you might also want to pass the queue of the main program to the
thread to send it messages)
    define methods to post messages to the queue like read, send to the
machine, stop, ...

    define the run method that is what will be called when you start
your thread.
        this method should enter an infinite loop that will check if
something has to be done (check in the queue).

hope this might help you
good luck

More information about the Python-list mailing list