using timers to force an execution time

[Nick Craig-Wood]

superpollo <user at example.net> wrote:
>  Diez B. Roggisch wrote:
> > superpollo wrote:
> >>am i wrong?
> > 
> > 
> > No, you are right, for threads that is. You can try & trick around with the
> > trace-functionality of python, and some ctypes-based
> > system-thread-module-tricks that are, as mentioned before, black-magic & a
> > spinning gatling gun pointed to your very own lower extremities.
> > 
> 
>  OUCH.
> 
> > The only reliable way of terminating an asynchronous computation is to use
> > processes, since python2.6 that's rather convenient with the
> > multiprocessing-module. However, that imposes some limits to what you can
> > do.
> 
>  fair enough.

Or if you are on unix you can use signals...

It is probably just about acceptable to raise an exception in a signal
handler like this code does.

import signal, os, time

class TimeOut(Exception):
    """Thrown on alarm"""
    pass

def sig_alarm(signum, frame):
    raise TimeOut()

def time_out(t, fn, *args, **kwargs):
    """Calls fn with the args and kwargs returning its result or raising a TimeOut exception if it doesn't complete within t seconds"""

    # Turn alarm off and read old value
    old_alarm = signal.alarm(0)
    # Install new handler remembering old
    old_handler = signal.signal(signal.SIGALRM, sig_alarm)
    # Set the timer going
    signal.alarm(t)

    try:
        rc = fn(*args, **kwargs)
    finally:
        # Restore the old handler
        signal.signal(signal.SIGALRM, old_handler)
        signal.alarm(0)          

def test():
    for repeat in range(10):
        print time.time()
        time.sleep(0.66)

if __name__ == "__main__":
    try:
        time_out(3, test)
    except TimeOut:
        print "timed out"

-- 
Nick Craig-Wood <nick at craig-wood.com> -- http://www.craig-wood.com/nick