Skip to content

Instantly share code, notes, and snippets.

@bbarrows
Created June 25, 2012 17:56
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save bbarrows/2990205 to your computer and use it in GitHub Desktop.
Save bbarrows/2990205 to your computer and use it in GitHub Desktop.
LiveThread
import threading, logging, traceback
import sys, time
class LiveThread(threading.Thread):
def __init__(self, function_to_quit_the_process, thread_wait_in_seconds=1,
clean_up_wait_in_seconds=1, group=None, target=None, name="LiveThread",
args=(), kwargs={}, verbose=None, clean_up=None):
threading.Thread.__init__(self, group=group, target=target, name=name,
args=args, kwargs=kwargs, verbose=verbose)
self._function_to_quit_the_process = function_to_quit_the_process
self._should_stop_event = threading.Event()
self._done_running_event = threading.Event()
self._thread_wait_in_seconds = thread_wait_in_seconds
self._clean_up_wait_in_seconds = clean_up_wait_in_seconds
self._thread_name = name
self._target = target
self._args = args
self._kwargs = kwargs
self._clean_up_method = clean_up
self._has_ran = False
self._has_joined = False
def clean_up(self):
if self._clean_up_method:
self._clean_up_method()
def work_to_do(self):
if self._target:
self._target(*self._args, **self._kwargs)
def run(self):
if self._has_ran:
raise RuntimeError("This thread has already been started once.")
self._has_ran = True
try:
while not self._should_stop_event.isSet():
self._should_stop_event.wait(self._thread_wait_in_seconds)
self.work_to_do()
#self._threading_object.run()
self._done_running_event.set()
#this except could just raise or be removed too..
except (KeyboardInterrupt, SystemExit):
traceback_s = ''.join(traceback.format_exception(*sys.exc_info()))
logging.error("The thread: %s has cause a system exit our keyboard exit." % self._thread_name)
logging.error(traceback_s)
self._function_to_quit_the_process()
except:
traceback_s = ''.join(traceback.format_exception(*sys.exc_info()))
logging.error("The thread: %s has died with the following exception:" % self._thread_name)
logging.error(traceback_s)
self._function_to_quit_the_process()
finally:
# Avoid a refcycle if the thread is running a function with
# an argument that has a member that points to the thread.
# This finally is from the threading.Thread.run method
del self._target, self._args, self._kwargs
def join(self, timeout=None):
if not self._has_joined:
self._should_stop_event.set()
self._done_running_event.wait(self._clean_up_wait_in_seconds)
#Here I have given the run command 1 second to shutdown. I can
#do thread specific shutdown work if I want..
#self._threading_object.clean_up()
self.clean_up()
self._has_joined = True
threading.Thread.join(self, timeout)
#Something you want to thread has two required methods:
#run and clean_up
#run is the code looped over
#clean_up is ran after the run code has finished after the thread has been
# told to stop or until some time has elapsed
class SomeAutoUpgraderThing(LiveThread):
def __init__(self, count_til, count_by, function_to_quit_the_process):
LiveThread.__init__(self, function_to_quit_the_process, name="BradsThread")
self.count_til = count_til
self.count_by = count_by
self.current_count = 0
def clean_up(self):
print "Cleanin up my really important thread"
def work_to_do(self):
if self.current_count > self.count_til:
self.current_count = 0
else:
self.current_count = self.current_count + self.count_by
print "I am on %s" % str(self.current_count)
#This sleep emulates the thread doing some work
time.sleep(2)
def tear_down_proc():
print "Im dying gracefully!"
#2 is just a random time here. Just showing
#that we should specifiy a max amoutn of time
#we give the cleanup method to run
a_thing_to_thread.join(2)
another_thread.join(2)
sys.exit(0)
def my_exception_hook(exctype, value, tb):
try:
logging.error("This is the main threads exception handler:")
traceback_s = ''.join(traceback.format_exception(exctype, value, tb))
logging.error(traceback_s)
tear_down_proc()
except Exception, e:
logging.info("An error occurred in the custom traceback handling.")
logging.info(e)
sys.excepthook = my_exception_hook
a_thing_to_thread = SomeAutoUpgraderThing(100, 2, tear_down_proc)
a_thing_to_thread.start()
#The next 10 or so lines show the use of the target method of threading
#Versus extending a LiveThread
def this_is_another_important_thread():
print "Doooing lots of work!!"
time.sleep(1)
def another_clean_up_method():
print "Cleaning up the other thread!!!"
another_thread = LiveThread(tear_down_proc, clean_up=another_clean_up_method, target=this_is_another_important_thread, name="This is another thread")
another_thread.start()
time.sleep(3)
another_thread.join()
#This sleep emulates the process's other threads doign some work while the thread is off working
time.sleep(8)
tear_down_proc()
time.sleep(1)
print "Shoudlnt get here"
@tcherry
Copy link

tcherry commented Jun 25, 2012

I think the idea is right on. However, I am a little on the fence about abstracting the entire while loop. This seems to be a little too much abstraction:
pluses:

  1. It guarantees every thread has a stop event
  2. It guarantees things are logged and app quits if anything fails

minuses:
2) It forces you to use while loops. What about threads that use serve_forever or something other than while True.?
3) It changes the context of the inner-loop local loop variables (see: automation, udpcon). Not a big change, but would require some refactoring of certain loops to make sure no local variables are used.

@bbarrows
Copy link
Author

For point

  1. The case you are referring to is the HttpServer. If you look at the code for the HttpServer (SocketServer.py under the BaseServer class) they actually implementing almost exactly the same thing as LiveThread. They use a two part event system as well which is actually in a while loop.

So it would actually require the same changes as anywhere else.

With this specifically (since the serve_forever is NOT actually in a while True: but a while event.is_set()) you could just create new thread and call the BaseServer shutdown method on join however you would lose out on some of the benefits of LiveThread I mentioned before (logging, standardization).

So I would just have my work_to_do method call self._handle_request_noblock() and have the clean_up method call the servers shutdown method.

I actually have not yet seen a thread in our code base that does not fit this while loop format.. and if we do come up with one it can just not use this class. LiveThread is just a good way to standardize and not repeat ourself.

  1. I believe you are referring to the fact that because the while loop is now outside of the work_to_do function the local variable scope no longer lasts the entirety of the thread. To fix this, simple add self. to the front of all local variables you want to persist while the thread exists. I would say that this one minimal change is incredibly easy for what it buys us. We do not have to worry about someone implementing the looping event system incorrectly (or not at all) and it guarantees we can join all of our threads.

@bbarrows
Copy link
Author

Also I should be clear that LiveThread is just a way to help developers make they're threads act, as we as a team have deemed, correctly.

If there is a thread that does not fit the White running event paradigm that LiveThread enforces then that thread can just not use this class.

@bebrws
Copy link

bebrws commented Jul 1, 2019

I reported this as Abuse if some admin happens to read this comment.

I lost access to my account when I lost my phone, number, 2FA, etc.. and had to create a new account. Still fighting to get my gmail back but I don't think that will happen.

Anyway I was hoping to take down this Gist, I don't want it up anymore for some personal reasons but I can't take it down anymore without my account. Please consider doing so.

Thanks
Brad

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment