Rework PyAny::is_instance_of for performance

[davidhewitt]

I was talking to @samuelcolvin about the fastest way to identify object types (relevant e.g. for pythonize and also pydantic-core) and noticed that PyAny::is_instance_of is quite unoptimised because it expands to an ffi call to PyObject_IsInstance.

This PR proposes PyAny::is_instance_of::<T>(obj) is changed to be equivalent to T::is_type_of(obj), plus add a sprinkling of inlining. We often have implementations such as PyDict_Check which can pretty much be optimised away to just checking a bit on the type object.

The accompanying benchmark to run through a bunch of object types is approx 40% faster after making this change.

For completeness, I've also added PyAny::is_exact_instance and PyAny::is_exact_instance_of, to pair with T::is_exact_type_of. (This could be split into a separate PR if preferred.)

[davidhewitt]

Given the proximity to releasing 0.18 and that there's a couple of design choices / refinements to add here, I'm going to park this until after 0.18 is out.

[samuelcolvin]

Idea looks great, thanks so much. But I agree, no massive hurry.

[davidhewitt]

Rebased, changed PyAny::is_instance_of to return bool instead of PyResult<bool> and split out is_exact_instance[_of] into #3161.

The benchmark added in bench_any is about 30% faster for these changes compared to main. It's nice, though I'm not entirely sure it's worth the churn.

[adamreichold]

The benchmark added in bench_any is about 30% faster for these changes compared to main. It's nice, though I'm not entirely sure it's worth the churn.

Considering the source of the churn is an improvement IMHO (dropping the Result layer) and that idiomatic Python API will often check types before committing to an implementation due the high cost of error/exception handling, I would say this is worth it.

Hot paths would probably still benefit from look-up tables as discussed in #3104 but this could give a good boost across the ecosystem where the LUT approach would be inappropriately heavy-weight.

[davidhewitt]

Ok, that's a good argument in favour. Let's ship it!

bors r=adamreichold

[bors]

Build succeeded!

The publicly hosted instance of bors-ng is deprecated and will go away soon.

If you want to self-host your own instance, instructions are here.
For more help, visit the forum.

If you want to switch to GitHub's built-in merge queue, visit their help page.

• conclusion