Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Deprecate BlockingTrioPortal in favor of from_thread.run and from_thread.run_sync #1122

Merged
merged 12 commits into from
Jul 21, 2019

Conversation

epellis
Copy link
Contributor

@epellis epellis commented Jun 23, 2019

This addressses issue #810 by implementing trio.from_thread.run and trio.from_thread.run_sync and deprecating BlockingTrioPortal. There is probably some more stuff we need to add before this PR is fully complete, like integrating #1098 and the run_sync_soon API, but this is a good start.

Support re-entering the same Trio.run() loop by stashing the current
Trio Token in Thread Local Storage right before run_sync_in_thread
spawns a worker thread. When trio.from_thread.run_sync or
trio.from_thread.run are called in this thread, they can access the
token and use it to re-enter the Trio.run() loop.

This commit deprecates BlockingTrioPortal. For the majority of
use cases, using the thread local Trio Token should be sufficient. If
for any reason, another Trio Token needs to be used, it can be
passed as a kwarg to from_thread.run and from_thread.run_sync.

Here is a snippet from how the new API works:

import trio

def thread_fn():
    start = trio.from_thread.run_sync(trio.current_time)
    print("In Trio-land, the time is now:", start)
    trio.from_thread.run(trio.sleep, 1)
    end = trio.from_thread.run_sync(trio.current_time)
    print("And now it's:", end)

async def main():
    await trio.run_sync_in_thread(thread_fn)

trio.run(main)

Here is how the same code works in "old" Trio:

import trio

def thread_fn(token):
    portal = trio.BlockingTrioPortal(token)
    start = portal.run_sync(trio.current_time)
    print("In Trio-land, the time is now:", start)
    portal.run(trio.sleep, 1)
    end = portal.run_sync(trio.current_time)
    print("And now it's:", end)

async def main():
    token = trio.hazmat.current_trio_token()
    await trio.run_sync_in_thread(thread_fn, token)

trio.run(main)

This addressses issue #810 by implementing `trio.from_thread.run` and
`trio.from_thread.run_sync` and deprecating `BlockingTrioPortal`.

Support re-entering the same `Trio.run()` loop by stashing the current
trio token in Thread Local Storage right before `run_sync_in_thread`
spawns a worker thread. When `trio.from_thread.run_sync` or
`trio.from_thread.run` are called in this thread, they can access the
token and use it to re-enter the `Trio.run()` loop.

This commit deprecates `BlockingTrioPortal`. For the majority of
use cases, using the thread local Trio Token should be sufficient. If
for any reason, another special Trio Token needs to be used, it can be
passed as a kwarg to `from_thread.run` and `from_thread.run_sync`.

Here is a snippet from how the new API works:
```python3
import trio

def thread_fn():
    start = trio.from_thread.run_sync(trio.current_time)
    print("In Trio-land, the time is now:", start)
    trio.from_thread.run(trio.sleep, 1)
    end = trio.from_thread.run_sync(trio.current_time)
    print("And now it's:", end)

async def main():
    await trio.run_sync_in_thread(thread_fn)

trio.run(main)
```

Here is how the same code works in "old" Trio:
```python3
import trio

def thread_fn(token):
    portal = trio.BlockingTrioPortal(token)
    start = portal.run_sync(trio.current_time)
    print("In Trio-land, the time is now:", start)
    portal.run(trio.sleep, 1)
    end = portal.run_sync(trio.current_time)
    print("And now it's:", end)

async def main():
    token = trio.hazmat.current_trio_token()
    await trio.run_sync_in_thread(thread_fn, token)

trio.run(main)
```
@epellis epellis requested a review from njsmith June 23, 2019 22:23
Copy link
Member

@njsmith njsmith left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is a great start! Making the trio_token an optional argument is probably better than the complicated thing that I was suggesting.

Some boring comments:

  • We need to update the docs too. Generally when we deprecate something, we remove it from the docs immediately. (Anyone still using the old deprecated thing can look at the old docs if they really need them, and new users don't need to know about the old deprecated stuff.) So basically, you'll want to search for BlockingTrioPortal, and everywhere it shows up convert the docs to talk about trio.from_thread instead.

  • We'll need a newsfragment to tell people about the change. In this case, we already have a 810.removal.rst, so I guess we'll want to update that... we can figure out the details here at the end once the rest is sorted out, but I wanted to let you know it was coming :-).

Regarding the code itself:

  • We have a bunch of tests that use BlockingTrioPortal incidentally, or that are testing edge cases of BlockingTrioPortal. We should switch these to test trio.from_thread instead! Basically just a search-and-replace job I guess.

  • Once we've done that, codecov will complain that we've got lots of untested code in BlockingTrioPortal. To keep duplication down, we might as well delete all that code and switch it to using trio.from_thread instead... like, BlockingTrioPortal.__init__ will still want to capture a trio token, and then BlockingTrioPortal.run will just be a call to from_thread.run(..., trio_token=self._token), and ditto for run_sync.

Let us know if you have any questions!

trio/_threads.py Outdated
``run_sync_in_thread`` also injects the current ``TrioToken`` into the
spawned thread's local storage so that these threads can re-enter the Trio
loop by calling either :func: ``trio.from_thread.run`` or
:func: ``trio.from_thread.run_sync`` for async or synchronous functions,
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ReStrustructured Text format is a little weird: double-backticks, like ``trio.from_thread.run``, mean raw literal code – it's the equivalent of markdown's single-backticks. If you want to use :func: to link to a named function, then you have to use single-backticks: :func:`trio.from_thread.run`.

But, since #1091 was merged recently, we don't even need to write :func: most of the time – if you just plain single-backticks, like `trio.from_thread.run`, then sphinx will now automatically find the object and link to it – you don't have to use :func: to specify the type.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for the hint! Yeah, I have never used RST, only Markdown, and also never Sphinx either so the docs are a bit of a learning curve for me.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It looks like these double backticks still need to be replaced with single backticks.

trio/_threads.py Outdated Show resolved Hide resolved
trio/tests/test_threads.py Outdated Show resolved Hide resolved
trio/tests/test_threads.py Outdated Show resolved Hide resolved
trio/tests/test_threads.py Outdated Show resolved Hide resolved
trio/tests/test_threads.py Outdated Show resolved Hide resolved
epellis added 3 commits June 30, 2019 18:39
Rework thread local storage to use the canonical threading.local()
accessor, update from_thread.* unit tests to better reflect use cases,
updated no token error msg to give a specific reason for failure.
Updated documentation and replaced examples containing
`BlockingTrioPortal` to now reference `trio.from_thread.run` and
`trio.from_thread.run_sync`. Removed documentation on
`BlockingTrioPortal`.
Migrate all uses of `BlockingTrioPortal` in `test_thread.py` to use
`trio.from_thread.*` API instead.
@codecov
Copy link

codecov bot commented Jul 3, 2019

Codecov Report

Merging #1122 into master will decrease coverage by <.01%.
The diff coverage is 98.79%.

@@            Coverage Diff             @@
##           master    #1122      +/-   ##
==========================================
- Coverage   99.51%   99.51%   -0.01%     
==========================================
  Files         104      105       +1     
  Lines       12631    12711      +80     
  Branches      968      969       +1     
==========================================
+ Hits        12570    12649      +79     
- Misses         40       41       +1     
  Partials       21       21
Impacted Files Coverage Δ
trio/from_thread.py 100% <100%> (ø)
trio/tests/test_threads.py 100% <100%> (ø) ⬆️
trio/__init__.py 97.5% <100%> (+0.13%) ⬆️
trio/_threads.py 99.12% <96.42%> (-0.88%) ⬇️

@codecov
Copy link

codecov bot commented Jul 3, 2019

Codecov Report

Merging #1122 into master will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master    #1122      +/-   ##
==========================================
+ Coverage   99.54%   99.55%   +<.01%     
==========================================
  Files         104      105       +1     
  Lines       12646    12715      +69     
  Branches      967      969       +2     
==========================================
+ Hits        12589    12658      +69     
  Misses         36       36              
  Partials       21       21
Impacted Files Coverage Δ
trio/_core/_exceptions.py 100% <ø> (ø) ⬆️
trio/from_thread.py 100% <100%> (ø)
trio/_threads.py 100% <100%> (ø) ⬆️
trio/tests/test_threads.py 100% <100%> (ø) ⬆️
trio/__init__.py 100% <100%> (ø) ⬆️

epellis added 6 commits July 2, 2019 21:23
Applied `fixup_module_metadata` to `trio.from_thread` and corrected most
Sphinx build warnings.
Fixed issue where `RunFinishedError` was secretly refrencing deprecated
`BlockingTrioPortal`
Found the remaining RST linting errors thanks to @Fuyukai and the trio
gitter.im. Thanks!
Add back in some unit tests for the legacy `BlockingTrioPortal` to
ensure that the new `trio.from_thread` can handle being called from
`portal.run()` and `portal.run_sync()`.
Fixed naming conflict of `test_do_in_trio_thread_from_trio_thread`.
@epellis
Copy link
Contributor Author

epellis commented Jul 4, 2019

Okay, hello again! I have fixed up all of the issues mentioned in the last round of the PR and have also made sure all of the documentation and code coverage is at par with or better than where it was before.

@epellis epellis requested review from smurfix and njsmith July 4, 2019 04:10
@epellis
Copy link
Contributor Author

epellis commented Jul 10, 2019

@njsmith mind giving any feedback on this? I think I made all of the requested changes but want to confirm before I merge this.

Copy link
Member

@njsmith njsmith left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sorry for the slow follow up here.

This is looking good overall, so now you get all the fiddly little comments to polish up the details :-)

docs/source/reference-core/from-thread-example.py Outdated Show resolved Hide resolved
trio/_threads.py Outdated
``run_sync_in_thread`` also injects the current ``TrioToken`` into the
spawned thread's local storage so that these threads can re-enter the Trio
loop by calling either :func: ``trio.from_thread.run`` or
:func: ``trio.from_thread.run_sync`` for async or synchronous functions,
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It looks like these double backticks still need to be replaced with single backticks.

trio/_threads.py Show resolved Hide resolved
trio/_threads.py Outdated Show resolved Hide resolved
@epellis
Copy link
Contributor Author

epellis commented Jul 20, 2019

@njsmith Got some more time to polish this up, can you let me know if the latest round of changes are up to spec? Thanks as always!

@epellis epellis requested a review from njsmith July 20, 2019 02:02
@njsmith
Copy link
Member

njsmith commented Jul 21, 2019

@epellis Looks great, thanks! BTW, I think github is listing your last few commits as coming from your work account – doesn't matter to me, but FYI in case you wanted to keep those separate.

Any thoughts on what you want to work on next?

@njsmith njsmith merged commit 906456b into python-trio:master Jul 21, 2019
@epellis
Copy link
Contributor Author

epellis commented Jul 23, 2019

@njsmith Thanks for letting me know, I thought I signed out of that account on this laptop but I guess my git cache says otherwise 🤷‍♂. I'd love to continue working with the scheduler and threading/async interop so I'll take stock of how the rest of #1085 and friends are doing and try to comment there in the next week or so with chunks I think I could manage. Thanks again for your feedback, this PR was a really enjoyable process! :)

njsmith added a commit to njsmith/trio that referenced this pull request Aug 1, 2019
When it was deprecated in python-triogh-1122, it accidentally got moved into
trio.hazmat.

Fixes: python-triogh-1167
njsmith added a commit to njsmith/trio that referenced this pull request Aug 1, 2019
When it was deprecated in python-triogh-1122, it accidentally got moved into
trio.hazmat.

Fixes: python-triogh-1167
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants