Any plan to suppprt BackgroundTask from Starlette?

[oterrier]

Starlette allows to attach a list of background tasks to a response, that will run only once the response has been sent.
If the task is a not a coroutine is it executed on a specific executor to not block the event loop.

    if task.is_coroutine():
        future = asyncio.ensure_future(task())
    else:
        loop = asyncio.get_event_loop()
        future = await loop.run_in_executor(None, task.func)

Is there any possibility to add this great feature to Fastai ?

[euri10]

see #59 for quite a similar discussion Le jeu. 14 mars 2019 à 6:18 PM, oterrier <notifications@github.com> a écrit :

…

Starlette allows to attach a list of background tasks to a response, that will run only once the response has been sent. If the task is a not a coroutine is it executed on a specific executor to not block the event loop. if task.is_coroutine(): future = asyncio.ensure_future(task()) else: loop = asyncio.get_event_loop() future = await loop.run_in_executor(None, task.func) Is there any possibility to add this great feature to Fastai ? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#79>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABDZPoJBzMNoPbjTwAonRMJb9YJSe8i7ks5vWoRbgaJpZM4b0ykJ> .

[wshayes]

You can already do this since it's built on Starlette - make sure you return JSONResponse() otherwise it won't start the background tasks.

Example:

from starlette.background import BackgroundTasks
from starlette.responses import UJSONResponse, JSONResponse

@router.post("/nanopubs/flushtosearch", tags=["Nanopubs"])
def flushtosearch():
    """
    Flush all nanopubs to search endpoint
    """

    tasks = BackgroundTasks()
    tasks.add_task(services.search.elasticsearch_index_all_nanopubs)

    message = "Reset Index and Flush to search submitted"
    return JSONResponse(message, background=tasks)

[tiangolo]

Again, thanks for your help here guys @euri10 and @wshayes !

It is now integrated into FastAPI in a (probably) more intuitive way (in version 0.10.0).

The new docs are here: https://fastapi.tiangolo.com/tutorial/background-tasks/

In short:

from fastapi import BackgroundTasks, FastAPI

app = FastAPI()


def write_notification(email: str, message=""):
    with open("log.txt", mode="w") as email_file:
        content = f"notification for {email}: {message}"
        email_file.write(content)


@app.post("/send-notification/{email}")
async def send_notification(email: str, background_tasks: BackgroundTasks):
    background_tasks.add_task(write_notification, email, message="some notification")
    return {"message": "Notification sent in the background"}

[wshayes]

Wow - that's a lot more intuitive and will avoid the - "it's not working because you didn't return the JSONResponse()" issue from Starlette - thank you!

[tiangolo]

Awesome! I'm glad to hear you like the design/interface.

[madkote]

hi all, one similar question:

• let assume I would like to have a background task, which should NOT be executed after the response has been sent. but before?
• e.g. in REST, I receive the first chunk of data (multipart/form-data) and would like to start process it in background (e.g. thread - since it is blocking). After a final chunk, I would to await for the result from the executor and return the response to the client.

Is it possible with fastapi?

[tiangolo]

@madkote if your processing is in an async function (let's say, created with async def process_something(): ...), you can use an async function for your path operation too, and inside of it, use await process_something().

from starlette.concurrency import run_in_threadpool

@app.get("/items"/)
async def read_items():
    result = await process_something("argument 1", keyword_arg2="keyword argument")
    return result

If it's a normal function, you can use run_in_threadpool from Starlette, pass your standard function and await it, something like:

from starlette.concurrency import run_in_threadpool

@app.get("/items"/)
async def read_items():
    result = await run_in_threadpool(process_something, "argument 1", keyword_arg2="keyword argument")
    return result

If that doesn't solve your problem, please create a new issue for it so we can continue the discussion there.

---

As the original issue should be solved now with the support for BackgroundTasks, I'll close it now. But feel free to add more comments or create new issues.

[outofnames]

Hi!
Is it possible to use BackgroundTasks before HTTPException?
I have a scenario in which I need to send a email when an exception occurs, but BackgroundTasks doesn't seem to work when a HTTPException is raised

[madkote]

@outofnames I guess I would use celery in this scenario.

[outofnames]

Doesn't using celery only for sending emails seems overkill?

[madkote]

@outofnames well, otherwise you would need to extend HttpExceptino to be able to run background tasks, which might be difficult. Maybe, additionally, I would ask at starlette project directly.
If you find a solution, please share - I would very appreciate that!

[tiangolo]

@outofnames you can probably also use asyncio.ensure_future(awaitable_send_email()) inside of the exception handler, without needing to plug a background task.

https://docs.python.org/3/library/asyncio-future.html#asyncio.ensure_future

[outofnames]

@tiangolo thanks for the reply, I ended up adding more features to my project just to justify celery :D

[github-actions]

Assuming the original need was handled, this will be automatically closed now. But feel free to add more comments or create new issues or PRs.