Type Safety Cautions

1. Types Are Not Enforced at Runtime

The Scope

This affects everything in thread.

Why Not?

Although thread is intended to be used with strict type checking, we want to follow Python's design of "optional types". This allows thread to be compatible with different projects which may follow different architectural conventions.

What Does This Mean?

In other words, types are only enforced at the LSP/Linting level.

This means that you can annotate your code with types and still run it with while passing the wrong types. Your LSP/Linter will detect the type error and scream at you, thread will not validate types at runtime.

Simple example

Consider the following code;

threaded_function takes 1 integer as it's first argument and returns an integer. The parsed argument, "a", is not an integer. Thread will not raise a type error, however, a type error will be raised at runtime due to the invalid opperand between a string and an integer.

2. Args and Kwargs do not provide static type checking

The Scope

This only affects args and kwargs in thread.Thread() initialization.

import thread
 
def draw(x: int, y: int) -> bool: ...
worker = thread.Thread(target=draw, args=(thread,), kwargs={'y': ':<'}) # This will not be underlined by your LSP/linter
 
# Other methods from Thread are not affected
worker.result # Will be typed as bool

Why is this so?

thread.Thread() uses Python's typing library's ParamSpec to determine the arguments types of the target function. Due to ParamSpec.args and ParamSpec.kwargs requirement to be used as argument overloads, we cannot assign it to the args and kwargs arguments of thread.Thread().

P = ParamSpec('P')
T = TypeVar('T')
 
# What is required by ParamSpec.args and ParamSpec.kwargs
def __init__(
  self,
  target: Callable[P, T],
  *args: P.args,
  **kwargs: P.kwargs
) -> None: ...
 
# What we need
def __init__(
  self,
  target: Callable[P, T],
  args: P.args,
  kwargs: P.kwargs,
  *threadingArgs,
  **threadKwargs
) -> None: ...

Is this fixable?

Yes, if we change how we write thread.Thread()'s initialization. However, we want to make migrating to thread easy from the default threading library, and thus we will not change this for now and focus on adding new features.