A simple library for Python 3.5+ that provides an easy interface for multitasking.
- Pytasking
- Python 3.5+
There are no external module dependencies outside of the standard library however, if you'd like to take advantage of uvloop
, you can install that and the pytasking
library will use it automatically (Only available on Linux/MacOS).
- Include the directory
pytasking
in your project root directory. - If on Linux/MacOS; run
python -m pip install -r requirements.txt
.
- Run
pip install pytasking
.
A basic python example:
#!/usr/bin/env python
import pytasking
import time
def hello(hello_queue):
while True:
hello_queue.put_nowait("Hello World!")
pytasking.sleep(1.5, sync=True)
async def ping():
while True:
try:
print("Ping!")
await pytasking.sleep(1.0)
print("Pong!")
except pytasking.asyncio.CancelledError:
print("Pang!")
break
async def main(task_manager):
hellos = 0
hello_queue = pytasking.multiprocessing.Queue()
hello_proc = task_manager.add_proc(hello, hello_queue)
while True:
try:
if hellos == 5:
task_manager.delete_proc(hello_proc)
if not hello_queue.empty():
try:
print(hello_queue.get_nowait())
hellos += 1
except:
pass
ping_task = task_manager.add_task(ping)
await pytasking.sleep(0.5)
task_manager.delete_task(ping_task)
except pytasking.asyncio.CancelledError:
break
if __name__ == "__main__":
task_manager = pytasking.Manager()
task_manager.add_task(main, task_manager)
try:
task_manager.start()
except KeyboardInterrupt:
pass
except:
raise
Instances of the Manager
class provide an asynchronous event loop to the program. Currently pytasking only supports 1 asynchronous event loop at any given time.
Asynchronous tasks and parallel processes are spawned and managed by the Manager
instance.
Create an asynchronous task from a function definition. Pass arguments and keyword arguments as you would normally. This function returns an id from the has of the task. You can use the id to retrieve and delete the task. Make sure you define your function with the following template:
async def asynchronous_task_definition(): # Define any arguments or keyword arguments as you normally would.
# Do whatever you need to do here as you normally would.
# If you want this task to run indefinitely, do this:
while True:
try:
# Do something forever.
await pytasking.sleep(1.0)
except pytasking.asyncio.CancelledError: # This one is important.
# Normally you catch the cancel event and do something with it, but in this case, use it to break the loop and allow the task to close the task.
break
except:
raise
Tasks will start immediately and you may add a task anytime.
Given a task id, you can call to delete a task. This method calls the cancel()
method of the coroutine, it will give the coroutine the chance to cleanup and even deny the request if caught and handled in the pytasking.CancelledError
.
If you want to retrieve the underlying coroutine, you can use this method and provide the task id to get it.
This will return all the task ids as a list, you can use this method in conjunction with get_task(t_id)
.
Create a parallel process from a function definition. Pass arguments and keyword arguments as you would normally. This function returns an id from the has of the process. You can use the id to retrieve and delete the process. Do note, by default the process runs sequentially. Try to follow this template:
def parallel_process(): # Define any arguments or keyword arguments as you normally would.
# Do whatever you need to do here as you normally would.
# If you want this task to run indefinitely, do this:
while True:
try:
# Do something forever.
pytasking.sleep(1.0, sync=True)
except:
raise
Given a process id, you can call to delete a process. This method calls terminate()
and join()
to attempt to cleanly close the process. Closing the process while it is accessing a Pipe or Queue, may corrupt the resource.
If you want to retrieve the underlying process, you can use this method and provide the process id to get it.
This will return all the process ids as a list, you can use this method in conjunction with get_process(p_id)
.
This begins the Manager
instance and starts all added tasks and processes.
There maybe situations where you cannot spawn a task in a task, process in a process, task in a process, or a process in a task. I'll need to investigate further.
If you decide to delete a process be wary, if the process was in the middle of accessing a Queue or Pipe, that Queue or Pipe will be liable to corruption and will not be usable again.
- Changing naming convention, moving toward 1.x convention.
- All wrapped exceptions and data structures from the
asyncio
andmultiprocessing
modules have now been namespaced into pytasking. For example;pytasking.CancelledError
is nowpytasking.asyncio.CancelledError
. This change is so that it is more explicit and natural.
- Improved documentation.
- Implemented additional helper methods for the Manager class – see the documentation for details.
- This is the initial release of pytasking.