threadtools

Support for signals and better locks in native Python.

Inspiration

PyQt lets you "emit" signals that have function callbacks tied to them. Why shouldn't we have that in native Python?

Typical Usage

Signals

import time
from threading import Thread

from threadtools import Signal, process_events


class ThreadedProcess:
    """Mimics a long-running process that updates its progress."""

    def __init__(self):
        self.somethingHappened = Signal[str]()
        self.countChanged = Signal[int]()
        self.finished = Signal()  # no typing implies no arguments to `emit()`

    def run(self):
        for i in range(1, 6):
            time.sleep(1)
            self.countChanged.emit(i)
            if i == 3:
                self.somethingHappened.emit("Something happened!")
        self.finished.emit()


threaded_process = ThreadedProcess()
thread = Thread(target=threaded_process.run)
# connect signals
threaded_process.countChanged.connect(print)
threaded_process.somethingHappened.connect(print)
threaded_process.finished.connect(lambda: print("Done!"))
# run the thread
thread.start()
# you must call `process_events()` to receive signals from other threads
# `emit()` was called from a different thread than `connect()`, so the callbacks are queued
while thread.is_alive():
    process_events()

# prints:
# 1
# 2
# 3
# Something happened!
# 4
# 5
# Done!

DataLock

from threading import Thread

from threadtools import DataLock

# DataLocks are generic; they support any type
LOCKED_INTEGER = DataLock(0)
LOCKED_STRING = DataLock("Hello, World!")


class DataAccessor:
    """Accesses and mutates data that is behind a lock."""

    def __init__(self, int_value: int):
        self.int_value = int_value

    def run(self):
        # using a context manager locks the lock and returns the stored data
        with LOCKED_INTEGER as locked_int:
            # reading data
            print(locked_int)

            # writing data
            
            # BAD!!
            # this does not change the value inside the lock
            # (Python references don't work that way)
            locked_int = self.int_value
            
            # good
            # DataLocks are reentrant; this will not cause a deadlock
            LOCKED_INTEGER.set(self.int_value)

            # you can also get the data inside the lock using `get()`
            print(LOCKED_INTEGER.get())
        # the lock is unlocked here, at the end of the context


first_accessor = DataAccessor(1)
first_thread = Thread(target=first_accessor.run)

second_accessor = DataAccessor(2)
second_thread = Thread(target=second_accessor.run)

first_thread.start()
second_thread.start()

first_thread.join()
second_thread.join()

# if `first_thread` is started first, prints:
# 0
# 1
# 1
# 2

Thread Safety

Signals are thread-safe as long as they are connected correctly. See the connect() method for more details.