TST: parametrize sparse array arithmetic tests #27271
Conversation
| from pandas.core.sparse.api import SparseDtype | ||
| import pandas.util.testing as tm | ||
|
|
||
|
|
||
| @pytest.fixture(params=["integer", "block"]) | ||
| def kind(request): | ||
| return request.param |
There was a problem hiding this comment.
can you add a description here as a doc-string (just a 1-liner in quotes is good)
| ] | ||
| ) | ||
| def op(request): | ||
| return request.param |
There was a problem hiding this comment.
does not all_arithmetic_operators do this?
There was a problem hiding this comment.
that passes the strings, which leads to different NotImplemented semantics
There was a problem hiding this comment.
then make a fixture which converts the existing strings to this if need be
this is duplicating things unnecessarily
There was a problem hiding this comment.
just moved the fixture to the top conftest, named all_arithmetic_functions. we can't re-use the string one because that one doesnt include divmod and rdivmod.
| self._assert((b_dense / a).to_dense(), b_dense * 1.0 / a_dense) | ||
| if op in [operator.floordiv, ops.rfloordiv]: | ||
| # FIXME: GH#13843 | ||
| if self._base == pd.Series and a.dtype.subtype == np.dtype("int64"): |
There was a problem hiding this comment.
can you exfail the ones that are not defined / working
jreback
added
Sparse
|
Codecov fail looks like false positive |
|
any deal-breakers here? I'd like to get this in before trying to troubleshoot the (hopefully) last division-by-zero branch |
|
thanks @jbrockmendel nice! |
Motivated by a tedious troubleshooting process in a different branch.
I'm pretty sure we can take this even further, saving it for another pass since the diff here is already pretty big.