feat: add from_arrow (which uses the PyCapsule Interface) #1181
Conversation
MarcoGorelli
force-pushed
the
pycapsule-import
branch
from
fdcd234
to
f5da0cb
Compare
|
Thanks @MarcoGorelli, this looks great! |
narwhals/functions.py
Outdated
| @@ -406,6 +411,100 @@ def _from_dict_impl( | |||
| return from_native(native_frame, eager_only=True) | |||
|
|
|||
|
|
|||
| def from_pycapsule( | |||
| supports_py_capsule: SupportsPyCapsule, *, native_namespace: ModuleType | |||
There was a problem hiding this comment.
I was wondering if we could rename this to somehting like py_capsule_obj or arrow_py_capsule_obj or similar
At first glance I thought this was a bool (True if something supports pycapsule) 😅
There was a problem hiding this comment.
🤔 maybe this should just be native_frame?
narwhals/functions.py
Outdated
| class SupportsPyCapsule(Protocol): | ||
| def __arrow_c_stream__(self) -> Any: ... |
There was a problem hiding this comment.
This isn't the correct type hint: https://arrow.apache.org/docs/format/CDataInterface/PyCapsuleInterface.html#protocol-typehints
There was a problem hiding this comment.
I also personally stick with the upstream naming suggestion and call this ArrowStreamExportable https://github.com/kylebarron/arro3/blob/45be4a12dd62cee025c5d0ecf8c8c081e13643ea/arro3-core/python/arro3/core/types.py#L52-L75
narwhals/functions.py
Outdated
| @@ -406,6 +411,100 @@ def _from_dict_impl( | |||
| return from_native(native_frame, eager_only=True) | |||
|
|
|||
|
|
|||
| def from_pycapsule( | |||
There was a problem hiding this comment.
Personally I'd call this from_arrow and under the hood use the PyCapsule Interface. I think the PyCapsule itself is an implementation detail, and not something the user should need to know about.
There was a problem hiding this comment.
based on pola-rs/polars#12530 (comment), i think this risks being confused with Polars' from_arrow, which does something different
i'd be more inclined to use from_pycapsule then
There was a problem hiding this comment.
Do you only have top-level functions? Do you export a DataFrame-like class? Why not put it on that class?
But there is a difference between yours and the polars function because your function seems to only convert to a DataFrame while the polars function tries to decide whether it should return a Series or DataFrame.
from_pycapsule still doesn't really resolve the ambiguity, but you could change it to from_arrow_dataframe or from_arrow_df or from_arrow_table.
There was a problem hiding this comment.
thanks! yeah maybe class constructors are the way to go for this one
MarcoGorelli
changed the title
|
thanks @EdAbati and @kylebarron for reviews! 🙏 any further comments? |
|
gonna ship this so we can make a release, if there's issues we can always address as a follow-up (it's a new feature so it doesn't affect existing users) - thanks all for comments! |
closes #1158
What type of PR is this? (check all applicable)
Related issues
Checklist
If you have comments or can explain your changes, please do so below.