[ArrayManager] TST: run (+fix/skip) pandas/tests/frame/indexing tests #40323
Conversation
jorisvandenbossche
added
Indexing
jorisvandenbossche
changed the title
| def test_xs_droplevel_false_view(self): | ||
| # GH#37832 | ||
| df = DataFrame([[1, 2, 3]], columns=Index(["a", "b", "c"])) | ||
| result = df.xs("a", axis=1, drop_level=False) | ||
| df.values[0, 0] = 2 |
There was a problem hiding this comment.
this version feels much clearer as to whats going on
There was a problem hiding this comment.
Do you have a suggestion for how to then do it? (we can't rely on .values to mutate)
There was a problem hiding this comment.
i guess what we really care about here is np.shares_memory(result._values, df.iloc[:, 0])
There was a problem hiding this comment.
Added the np.shares_memory check. And turned out that also with ArrayManager this actually gives a view. But it's the iloc modifying the parent that doesn't result in modifying the child result, when using ArrayManager, but also when using a mixed DataFrame. So added that as an explicit test case as well.
| @@ -382,6 +388,9 @@ def test_setitem_frame_duplicate_columns(self): | |||
| columns=cols, | |||
| dtype="object", | |||
| ) | |||
| if using_array_manager: | |||
| expected["B"] = expected["B"].astype("int64") | |||
There was a problem hiding this comment.
comment as to why the difference?
There was a problem hiding this comment.
Yes, added a comment (for "C" it's expected, since it's a __setitem__, for "B" it's actually a TODO to have .loc[: "col"] not overwrite)
| @@ -296,6 +298,10 @@ def test_setitem_dt64tz(self, timezone_frame): | |||
| tm.assert_series_equal(df["D"], Series(idx, name="D")) | |||
| del df["D"] | |||
|
|
|||
| if using_array_manager: | |||
| # TODO(ArrayManager) rewrite test for ArrayManager | |||
| return | |||
There was a problem hiding this comment.
The first part of the test is still useful to run, so I would rather not skip it
There was a problem hiding this comment.
if you skip/xfail at this point, then the part up to here will still run
There was a problem hiding this comment.
In the end was easy to rewrite using .arrays so it works for both
pandas/core/internals/managers.py
Outdated
| @@ -1617,6 +1617,15 @@ def get_slice(self, slobj: slice, axis: int = 0) -> SingleBlockManager: | |||
| def index(self) -> Index: | |||
| return self.axes[0] | |||
|
|
|||
| @property | |||
| def array(self) -> ArrayLike: | |||
There was a problem hiding this comment.
could put this on SingleDataManager and return self.arrays[0]?
There was a problem hiding this comment.
Yes, moved it to the base class (note this will require some type ignoring, because .arrays is not exactly the same for both classes)
|
@jbrockmendel ready to merge here? |
|
LGTM cc @jreback |
|
thanks @jorisvandenbossche |
xref #39146
Extracting part of the test changes of #39578. This PR is doing
pandas/tests/frame/indexing. Thepandas/tests/indexingtests are handled in #40323.