Tips on how to run async code

[simonw]

Steps to reproduce

>>> import py_mini_racer
>>> context = py_mini_racer.MiniRacer()
>>> result = context.eval("""
... async function pretendToDelay() {
...     return new Promise(resolve => {
...         setTimeout(() => resolve('Data loaded!'), 1000);
...     });
... }
... pretendToDelay();
... """)
>>> print(result, result.__dict__)
<py_mini_racer.py_mini_racer.JSObject object at 0x100ec6620> {'id': 965435}

That code to paste:

import py_mini_racer
context = py_mini_racer.MiniRacer()
result = context.eval("""
async function pretendToDelay() {
    return new Promise(resolve => {
        setTimeout(() => resolve('Data loaded!'), 1000);
    });
}
pretendToDelay();
""")
print(result, result.__dict__)

I tried using await pretendToDelay(); on the last line instead and got this:

py_mini_racer.py_mini_racer.JSParseException: SyntaxError: await is only valid in async functions and the top level bodies of modules

Expected behavior

I'm trying to figure out how to return the result from resolving that promise.

Assuming there's a pattern for doing this, it would be great if the test suite and documentation described it!

[bpcreech]

There isn't a pattern for this except for polling. test_async in tests/test_eval.py does this.

Also, you didn't run into this yet, but setTimeout doesn't exist, because it's an HTML DOM and Web Worker API not an ECMAScript API. That test above uses Atomics.waitAsync as a workaround for that.

I'm working on adding proper async support, modeled as async/await on the Python side too. This requires changing the Python<->DLL interaction to use callbacks. I have this working on a branch (or, well, 5 branches with iterations of this)... current status is it adds 40x overhead to all calls, so I'm working on that before landing it. :)

[simonw]

Neat! That test_async polling example is exactly what I needed. Sounds like it's not worth documenting right now though if a better solution is incoming soon.

Generally I'd find documentation on "JavaScript patterns that work well with PyMiniRacer" to be super useful.

[bpcreech]

Sounds good, feel free to start a "js_cookbook.md" if you want! For the immediate future I'm focusing instead on making async work, and before that, making the pre-commit GH Action actually work. :)

[bpcreech]

An update: #14 will resolve this in v0.10.0. But I plan to release a v0.9.0 first with some foundational work, as a separate release to help bisect issues in case anything goes wrong with all the C++ refactoring I'm doing in here. So this should be resolved in a few days (PyMiniRacer builds are super slow, what with emulated aarch64 on the free GH runners!).

[bpcreech]

Ok v0.10.0, just pushed to PyPI, should resolve this, by adding and documenting support for JS Promises, awaitable via blocking and asyncio APIs.

I documented this in the README... I didn't go to the length of starting a "cookbook" yet because the README is still of manageable length, I think.