PR#3372 also updates starlette, however it seems to take a long time to merge as it includes many other changes.

I hope we can merge this smaller branch quicker to resolve#3589 .

I copied the test fix from#3372 to prevent conflicts later

If its easier for any reason - the same can be achieved by updating to 0.15.0

@graingert - Removed aiofiles and tested it still works forFileResponse andStaticFiles

what do you think about the contextlib2 backport changes? Starlette depends on it on python3.6 so FastAPI doesn't need the async_generator or async_exit_stack deps anymore

@graingert - Not to familiar with these, aren't these more related to the trio/anyio change? can we remove the backports without changes from#3372 ?

@alonme not directly related to anyio/trio no

@graingert - Thanks, backported these changes from the other PR.

Can anyone enable tests?

@tiangolo is the only person with merge and test run access unfortunately

+1
Please merge#3589

resolves#3674

PR#3372 also updates starlette, however it seems to take a long time to merge as it includes many other changes.

I'm not sure which changes in#3372 would delay the PR

@graingert, me neither - it just seemed to me that it wasn't getting merge quickly at the time.
This pr was much smaller than#3372 , however after your rightful comments they aren't that different anymore.
This is still simpler, and i hoped it would be possible to merge it quickly as it fixes a bug which i think is pretty bad.

I was considering adding wait_all to starlette atstarlette.concurrency:wait_all as it's the most complicated bit of the changes imho

@tiangolo - Can we proceed with one of the PR's?

It would be good to get a release with a new Starlette that supports py3.6-py3.10 before the December EOL of Python 3.6

@tiangolo - can we please go forward with this? or at least get a comment?
There are at least 2 issues that this will fix, and another project that waits for features from starlette 0.16

thanks

any update on this?

I have to say this is not a good look for the project. Not only this PR but many others are stalled with no comments from the maintainer. Actually it seems like the project is unmaintained.

@stephan-hesselmann-by you are correct unfortunately, this project is effectively abandoned

@stephan-hesselmann-by you are correct unfortunately, this project is effectively abandoned

💯 I'm considering dropping down to starlette directly esp considering all the shims and overrides I'm putting into place to get everything how I want it (custom error responses, finer grained transaction control, singleton dependencies,#1359, etc...). API doc is nice I suppose?

The project is not abandoned at all.

In fact, one of the main things wanted in the existing PRs and issues was a way to improve how models are defined for databases and Pydantic, and support for the latest versions of SQLAlchemy, so I spent months buildinghttps://github.com/tiangolo/sqlmodel, which ismadefor FastAPI. 🚀

I have a lot of pending PRs to review and a lot of things in the backlog, and I'm covering them all gradually and with the minimum disruption possible. This also applies to all my Docker images, that for example, now support Python 3.9, all of them.

I'm carefully reviewing each PR and issue myself, and I have been changing my whole work-life in the last months, including dealing with visas and German bureaucracy to be able to dedicate much more time to all my open source projects.

Also, the comments claiming that it is abandoned were made 2 hours ago, when the last commit improving theHTTPS guide was done 3 hours ago. 🤦 😞

The improved HTTPS guide is part of an effort into improving the information about deployments and how (and when) to use which Docker images, etc. So that will come soon too. 🎉

Thank you for the response@tiangolo! I'm very glad this project is alive 😄

First off, I think it is amazing the amount and quality of work you are able to do as one person. That said, I do think it might be best for the health of the projects to build teams by giving repeat contributors that have shown to produce high quality work more privileges, that way they can get PRs into a good state (by providing feedback to PR authors) before you even have to look at them, even if you are still the one giving the final ✅ . This can also be codified, e.g. w/CODEOWNERS files.

The project is not abandoned at all.

In fact, one of the main things wanted in the existing PRs and issues was a way to improve how models are defined for databases and Pydantic, and support for the latest versions of SQLAlchemy, so I spent months buildinghttps://github.com/tiangolo/sqlmodel, which ismadefor FastAPI. rocket

I have a lot of pending PRs to review and a lot of things in the backlog, and I'm covering them all gradually and with the minimum disruption possible. This also applies to all my Docker images, that for example, now support Python 3.9, all of them.

I'm carefully reviewing each PR and issue myself, and I have been changing my whole work-life in the last months, including dealing with visas and German bureaucracy to be able to dedicate much more time to all my open source projects.

Also, the comments claiming that it is abandoned were made 2 hours ago, when the last commit improving theHTTPS guide was done 3 hours ago. facepalm disappointed

The improved HTTPS guide is part of an effort into improving the information about deployments and how (and when) to use which Docker images, etc. So that will come soon too. tada

Awesome to hear from you@tiangolo . I am using fastapi for almsot a year now. And the popularity keep rising.
I am sure FastAPI would be benefit from Community Driven development , since the project is growing faster for one person to manage, have you consider starting a foundation / organization and appointing good contributors to help you review code / merge?

That would make FastAPI grow bigger and FastAPI is really the web framework python deserves.

@alonme I think you need to rebase, mainly because a similar PR was merged and contains a few updates

Does this PR add anything compared to the merged one?

@brnosouza - thanks, will get to that a bit later
@agronholm - from what i have seen, the update is only to starlette 0.15.0, and not 0.16.0

so my update also accounts for the aiofiles documentation@tiangolo described here#2899 (comment)

rebased

ok, now i see@tiangolo created his own PR to update to 0.16.0

So this only includes docs fix

Codecov Report

Merging#3594 (6232e44) intomaster (8a353ab) willnot change coverage.
The diff coverage isn/a.

@@            Coverage Diff            @@##            master     #3594   +/-   ##=========================================  Coverage   100.00%   100.00%           =========================================  Files          531       531             Lines        13629     13629           =========================================  Hits         13629     13629

Continue to review full report at Codecov.

Legend -Click here to learn more
Δ = absolute <relative> (impact),ø = not affected,? = missing data
Powered byCodecov. Last update8a353ab...6232e44. Read thecomment docs.

📝 Docs preview for commit6232e44 at:https://6279ba36fab0da14c006f463--fastapi.netlify.app

Thanks for the discussion everyone, this was solved a while ago in another PR and has been available for a while in recent releases.

Given that, I reduced the scope of this particular PR to remove the docs/translations references toaiofiles. Thanks for the interest and discussion you all! ☕