A subprocess replacement with tee support for both synchronous and asynchronous contexts.
Supports Python 3.8+.
Just import the run
function and use it as you would use subprocess.run
.
from tee_subprocess import run
process = run(["python", "--version"], tee=True, text=True, capture_output=True)
# ==> Python 3.11.2
print(process.stdout)
# ==> Python 3.11.2
Changing stdout
and stderr
changes the location to which the tee
occurs. You can supply any of the defined options in subprocess
or asyncio.subprocess
(STDOUT
, DEVNULL
, etc), as well as a writable text or binary file object; if providing a text file object, you must specify text = True
.
Internally, tee_subprocess
utilizes asyncio
to concurrently output and capture the subprocess logs. If an event loop is already running, run
will return an awaitable coroutine. Otherwise, it will call asyncio.run
for you. Practically, this means you can just treat run
as a coroutine if you're in an async context.
async def main():
process = await run(["python", "--version"], tee=True, text=True, capture_output=True)
# ==> Python 3.11.2
print(process.stdout)
# ==> Python 3.11.2
asyncio.run(main())
I do my best to provide a logical static function typing for any permitted invocation style. Mypy should complain about missing or invalid overloads if you attempt to use a combination of arguments with undefined behavior (like supplying text=True
while also providing a BytesIO
as stdout
, or supplying a PathLike
argument while shell=True
).
The one fairly large exception to this is await run(...)
vs run(...)
. For now, run
returns a union between a complete process and a coroutine due to the runtime-check for an asyncio context. As a result, you'll have to cast
the run(...)
call to either an awaitable or a CompletedProcess
depending on your specific use. The API may change in the future to avoid this problem.
subprocess-tee, the motivation for this library, has the same objective but fails to accommodate asynchronous applications and non-shell invocations. This library supports asynchronous contexts as well as direct, non-shell, program execution ("list-style" vs. "shell-style").
MIT License Copyright (c) 2023-2024 Elias Gabriel