Add foreign table providers #921
Conversation
timsaucer
force-pushed
the
feature/ffi
branch
from
fcf9475 to
f5f983c
Compare
| pub fn register_table_provider( | ||
| &mut self, | ||
| name: &str, | ||
| provider: Bound<'_, PyAny>, |
There was a problem hiding this comment.
Just to understand, taking delta-rs as an example.
The provider would be an instance of a deltatable object?
There was a problem hiding this comment.
Here is what you would need to implement in delta-rs as a python method on RawDeltaTable
fn __datafusion_table_provider__<'py>(&self, py: Python<'py>) -> PyResult<Bound<'py, PyCapsule>> {
let name = CString::new("datafusion_table_provider").unwrap();
let provider = FFI_TableProvider::new(Arc::new(self._table.clone()));
PyCapsule::new_bound(py, provider, Some(name.clone()))
}
Does that answer the question?
| provider: Bound<'_, PyAny>, | ||
| py: Python, | ||
| ) -> PyResult<()> { | ||
| if provider.hasattr("__datafusion_table_provider__")? { |
There was a problem hiding this comment.
Should we raise when it doesn't have this attribute, it might give the wrong impression for users that they registered it but nothing happened
timsaucer
force-pushed
the
feature/ffi
branch
from
f5f983c to
7240d44
Compare
timsaucer
force-pushed
the
feature/ffi
branch
from
7240d44 to
01a15f1
Compare
|
Now that 43.0.0 has merged into main, I've rebased and I'll work next on adding example and unit tests. |
…un during the first pass of pytest when the module hasn't been built
| if provider.hasattr("__datafusion_table_provider__")? { | ||
| let capsule = provider.getattr("__datafusion_table_provider__")?.call0()?; | ||
| let capsule = capsule.downcast::<PyCapsule>()?; | ||
| // validate_pycapsule(capsule, "arrow_array_stream")?; |
There was a problem hiding this comment.
Should this still be commented out?
| """Remove a table from the session.""" | ||
| self.ctx.deregister_table(name) | ||
|
|
||
| def register_table_provider(self, name: str, provider: Any) -> None: |
There was a problem hiding this comment.
You could use a protocol typehint perhaps, so that you get warnings when that attribute is missing
|
Thank you for the reviews! I've added an issue to track those two small requests, but I'd like to get this in now |
3 participants
Which issue does this PR close?
Closes #823
Closes #941
Rationale for this change
This feature enables DataFusion python to interoperate with foreign table providers that implement TableProvider and expose it via PyCapsule
What changes are included in this PR?
Adds a feature to register a table provider in the session context.
Are there any user-facing changes?
Addition only, no existing APIs modified.