Stream response body in ASGITransport #3059
Conversation
tests/test_asgi.py
Outdated
| @pytest.mark.anyio | ||
| async def test_asgi_stream_returns_before_waiting_for_body(): | ||
| start_response_body = anyio.Event() | ||
|
|
||
| async def send_response_body_after_event(scope, receive, send): | ||
| status = 200 | ||
| headers = [(b"content-type", b"text/plain")] | ||
| await send( | ||
| {"type": "http.response.start", "status": status, "headers": headers} | ||
| ) | ||
| await start_response_body.wait() | ||
| await send({"type": "http.response.body", "body": b"body", "more_body": False}) | ||
|
|
||
| async with httpx.AsyncClient(app=send_response_body_after_event) as client: | ||
| async with client.stream("GET", "http://www.example.org/") as response: | ||
| assert response.status_code == 200 | ||
| start_response_body.set() | ||
| await response.aread() | ||
| assert response.text == "body" |
There was a problem hiding this comment.
This test makes sense to me yep.
How about the case where we're streaming the response body, and ensuring that we're able to receive it incrementally? Are we able to test that also?
await start_response_body.wait()
await send({"type": "http.response.body", "body": b"1", "more_body": True})
await keep_going.wait()
await send({"type": "http.response.body", "body": b"2", "more_body": True})
await nearly_there.wait()
await send({"type": "http.response.body", "body": b"3", "more_body": False})There was a problem hiding this comment.
Yes, we should be able to test that, I will update the PR accordingly (I will also take in account the pending deprecation of the app argument in AsyncClient in the added tests)
jhominal
force-pushed
the
asgi-transport-streaming
branch
2 times, most recently
from
460832c
to
e249563
Compare
jhominal
force-pushed
the
asgi-transport-streaming
branch
from
e249563
to
672f459
Compare
|
The PR was updated, with a new |
|
After working with the version of I have an idea in order to fix that, here is how I think it could work:
I would like your opinion on whether the idea mentioned above should be added to the current PR, or be the object of a new PR. |
Either would be okay with me. |
Summary
As part of my job, we needed a variant of
ASGITransportthat supports streaming (as in #2186), and this is my PR to implement that.Something that I am particularly proud of is that this PR was written without having to spawn a new task, with the consequence that it avoids issues related to task groups and context variables.
Checklist
Fixes #2186