Update pylibcudf testing utilities
#15772
Conversation
brandon-b-miller
added
feature request
| @@ -24,28 +24,28 @@ def metadata_from_arrow_array( | |||
| return metadata | |||
|
|
|||
|
|
|||
| def assert_column_eq(plc_column: plc.Column, pa_array: pa.Array) -> None: | |||
| def assert_column_eq(expect: pa.Array, got: plc.Column) -> None: | |||
There was a problem hiding this comment.
Do we want to model this on
cudf/python/cudf/cudf/testing/_utils.py
Line 116 in 6d5f965
We may not always be comparing an “expected/actual” result. The assert_eq just uses left/right names. The order is perhaps better defined by the caller, and hint at that intended use case in a docstring?
Additionally, it’s not obvious that this function expects a PyArrow and a pylibcudf object. Should we normalize this to accept either argument as either type, more like assert_eq?
There was a problem hiding this comment.
I updated this to be more symmetric, although I think we're stuck with having to have at least one of the objects be a pyarrow array, even if we don't know which one, unless we want to make broader changes. This is due to needing the metadata from the pyarrow array to compare nested data.
brandon-b-miller
requested review from
msarahan,
vyasr and
ttnghia
and removed request for
a team
github-actions
bot
added
libcudf
brandon-b-miller
removed
libcudf
|
/merge |
Cleans up some testing utilities for pylibcudf as suggested in #15418 (comment).