Closed stevemk14ebr closed 3 years ago
This crate requires: (1) a spawn
function that spawns a 'static future, and; (2) a block_on
function that blocks the current thread on a future. The second is only needed for the only fully-safe abstractions we could come up with.
Unfortunately, it seems tokio doesn't support a block_on
feature that can be called from an async context (as per the docs of Runtime::block_on
). I think this is a blocker for using tokio (as the safe abstractions would end up panicking). The unsafe interface should still work with tokio
though.
is this the equivalent? https://tokio-rs.github.io/tokio/doc/tokio/task/fn.block_in_place.html
Nice find! The function takes a FnOnce
and blocks on it, but what we currently need is a function that takes a Future
. I think we should be able to use this functionality for instance, by running a single-thread executor of the Future
we wish to drive within the FnOnce
passed to block_in_place
. However, we have to evaluate this a bit more carefully, and ensure complete safety.
Unfortunately, I can't provide a timeline for doing this yet. I could guide you on this, if you are interested in providing a PR.
@stevemk14ebr I tried a tokio implementation in this PR: https://github.com/rmanoka/async-scoped/pull/4 . See tests for example usage.
tokio's block_in_place
is such a great functionality; interestingly it prevents a particular deadlock that happens with deep recursive usage of Scope
. However, it can only be called from an async context, so slightly different from similar functionality of async_std.
Thanks for taking the time to work on that!
Are there plans for official support for tokio? I am curious what the technical blockers to this may be.