BUG: import of maybe_convert_indices in pandas.core.index.py, #10610 #10872
Conversation
…checking in the case of a mixed-integer index. Added a test to verify in test_index
…checking in the case of a mixed-integer index. Added a test to verify in test_index
…e_index # Conflicts: # pandas/tests/test_index.py
…into mixed_type_index
…into mixed_type_index n test_index
| @@ -971,11 +971,13 @@ def _convert_list_indexer(self, keyarr, kind=None): | |||
|
|
|||
| if self.inferred_type == 'mixed-integer': | |||
| indexer = self.get_indexer(keyarr) | |||
| if not np.all(np.in1d(indexer, self.values)): | |||
There was a problem hiding this comment.
why is this line necessary?
There was a problem hiding this comment.
Without the line, when passed a list containing index values, the function would output the data associated with the nearest integer index value. For other cases where an invalid index is passed to a DataFrame, an IndexError would be raised, so I made sure to explicitly check for that condition.
There was a problem hiding this comment.
your test doesn't hit that line though it goes thru
can u make a test that is explicitly caught there?
jreback
added
Bug
Indexing
jreback
changed the title
|
I removed the index checking if-statement, and added a line that would change the -1 indices to equal the length of the Index being evaluated, so that the IndexError would be raised in maybe_convert_indices. |
| # values outside the range of indices so as to trigger an IndexError in maybe_convert_indices | ||
| indexer[indexer < 0] = len(self) | ||
| from pandas.core.indexing import maybe_convert_indices | ||
| return maybe_convert_indices(indexer, len(self)) |
There was a problem hiding this comment.
this is already done inside maybe_convert_indices why do you think its needed here?
There was a problem hiding this comment.
It is necessary because negative indices passed into maybe_convert_indices are summed by the length of the array before the bounds checking is done. Because get_indexer assigns labels that are not found in the index with a value of -1, the last row or column is assigned to those positions instead of either NaNs or an exception being raised. With the line I added, the bounds checking in maybe_convert_indices will now throw an IndexError for out of bounds indices.
|
ok, this is fine. pls squash to a single commit. ping when green. |
| @@ -1729,6 +1729,11 @@ def test_equals_op_multiindex(self): | |||
| df.index == index_a | |||
| tm.assert_numpy_array_equal(index_a == mi3, np.array([False, False, False])) | |||
|
|
|||
| def test_multitype_list_index_access(self): | |||
| df = pd.DataFrame(np.random.random((10, 5)), columns=["a"] + [20, 21, 22, 23]) | |||
There was a problem hiding this comment.
add the issue number as a comment here
|
merged via 12df3ea thanks! |
closes #10610
I fixed the import statement and added a test to check for proper behavior when accessing a mixed-integer index with a list of values.