Remove validate_pycapsule

[Tpt]

Rationale for this change

The Bound<'_, PyCapsule>::pointer_checked method does the same validation and is already used across the codebase

What changes are included in this PR?

• remove validate_pycapsule
• remove some not useful .map_err(py_datafusion_err)?
• replace c_str! calls by c-literals (now supported by MSRV)

Are there any user-facing changes?

Some error messages have changfed

[kevinjqliu]

We lose the error messages by removing this function. Would this be potentially confusing to end users?

[Tpt]

Good point. The error messages from PyCapsule_GetPointer are ValueError("PyCapsule_GetPointer called with invalid PyCapsule object") and ValueError("PyCapsule_GetPointer called with incorrect name"). These error only happen if the "magic" methods return invalid capsules so they should only be encountered by people trying to implement them and not regular DataFusion users. Also, the stack trace should give a hint of which capsule is raising the error.

An option might be to wrap the error in a nicer message. WDYT?

[kevinjqliu]

Thanks for pointing out the error messages from PyCapsule_GetPointer! They do seem fairly similar 😄

I still think it could be helpful to expose validate_pycapsule as a public function for implementers. While PyCapsule_GetPointer errors will mostly be encountered by developers implementing the “magic” methods, having a helper with clearer validation and error messages could make debugging easier. Since https://github.com/apache/datafusion-python/pull/1414/files#diff-6dc4fc730855f068c83d61b03b141832727bb30bc71788f763761ef8cd2bde7eR154 introduces a new datafusion-python-util crate that includes this helper, exposing it there could provide implementers with a simple way to validate capsules before extracting pointers.

[Tpt]

Thanks for pointing out the error messages from PyCapsule_GetPointer! They do seem fairly similar 😄

I am also contributing to PyO3. It made me move deep in some CPython rabbit hole.

having a helper with clearer validation and error messages could make debugging easier

If we want nice helper what about abstracting as much as possible the PyCapsule details to get an API like:

#[pymethods]
impl MyTableProvider {
    pub fn __datafusion_table_provider__<'py>(
        &self,
        py: Python<'py>,
        session: PyDataFusionCapsule<Arc<dyn LogicalExtensionCodec>>,
    ) -> PyResult<PyDataFusionCapsule<Arc<dyn TableProvider>>> {
    }
}

If you want I can give it a try after #1414 is merged and in the meantime close this MR. My hope it that it could work by playing around with some traits.