This PR builds on the previous work on callbacks (#23), and promises (#32).
In uniffi parlance:
JS callbacks are "foreign" traits,
Asynchronous JS promises map to Rust futures, so:
Asynchronous JS callbacks are "foreign" futures.
As part of this work,
the fixtures/futures test fixture, and translates the tests
additional test infrastructure to measure how long a call takes.
the callinvoker is now passed to the toJs Bridging methods.
Some extra things to think about:
errors are not being passed reliably back to the calling function. This is most likely due to "complex" errors that haven't yet been implemented. For that reason, I have left this work for a different PR. I have implemented the tests, but x'd them out.
performance does not seem to be very fast for futures– the alwaysReady function can take around 100ms! I am fairly sure this is down to the test-runner naive implementations of the event loop and call invoker. I would like to measure this in an app as soon as possible.
the concept of task cancellation is missing in Javascript, so this is difficult to test.
jhugman
commented
1 month ago
According to The Big O of Code Reviews, this is a O(n) change.
This PR builds on the previous work on callbacks (#23), and promises (#32).
In uniffi parlance:
As part of this work,
fixtures/futurestest fixture, and translates the teststoJsBridging methods.Some extra things to think about:
x'd them out.alwaysReadyfunction can take around 100ms! I am fairly sure this is down to the test-runner naive implementations of the event loop and call invoker. I would like to measure this in an app as soon as possible.