Skip to content

TokenChingy/pytasking

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

31 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Downloads

Pytasking

A simple library for Python 3.5+ that provides an easy interface for multitasking.

Table of Contents

Dependencies

  • 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).

Installation

Source

  • Include the directory pytasking in your project root directory.
  • If on Linux/MacOS; run python -m pip install -r requirements.txt.

PyPi

  • Run pip install pytasking.

Usage

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

API

class pytasking.Manager()

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.

add_task(task, *args, **kwargs)

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.

delete_task(t_id)

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.

get_task(t_id)

If you want to retrieve the underlying coroutine, you can use this method and provide the task id to get it.

get_tasks()

This will return all the task ids as a list, you can use this method in conjunction with get_task(t_id).

add_proc(proc, *args, **kwargs)

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

delete_proc(p_id)

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.

get_proc(p_id)

If you want to retrieve the underlying process, you can use this method and provide the process id to get it.

get_procs()

This will return all the process ids as a list, you can use this method in conjunction with get_process(p_id).

start()

This begins the Manager instance and starts all added tasks and processes.

Known Issues

Recursive spawning

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.

Pipe/Queue corruption

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.

Changes

1.2

  • Changing naming convention, moving toward 1.x convention.

Breaking Changes

  • All wrapped exceptions and data structures from the asyncio and multiprocessing modules have now been namespaced into pytasking. For example; pytasking.CancelledError is now pytasking.asyncio.CancelledError. This change is so that it is more explicit and natural.

1.1.0

  • Improved documentation.
  • Implemented additional helper methods for the Manager class – see the documentation for details.

1.0.0

  • This is the initial release of pytasking.

Stargazers Over Time

Stargazers over time