BUG: Index does not inherit existing Index or DatatetimeIndex object … #11695
Conversation
|
Great!
I wonder if you could just call |
|
I'll rebase on master and squash my commits. I should update my PR shortly. |
jreback
added
Bug
Indexing
| @@ -105,3 +105,5 @@ Performance Improvements | |||
|
|
|||
| Bug Fixes | |||
| ~~~~~~~~~ | |||
|
|
|||
| - Bug in ``Index`` prevent inheriting, from the passed Index or DatetimeIndex object, the existing name when a new one is not provided (:issue:`11193`) | |||
There was a problem hiding this comment.
just say creation, don't use inheriting, which has a different meaning
| - Bug in ``.loc`` against ``CategoricalIndex`` may result in normal ``Index`` (:issue:`11586`) | ||
| - Bug groupby on tz-aware data where selection not returning ``Timestamp`` (:issue:`11616`) | ||
| - Bug in timezone info lost when broadcasting scalar datetime to ``DataFrame`` (:issue:`11682`) | ||
| - Bug in ``Index`` prevents creation of a new Index object which has the same name of the passed Index or DatetimeIndex object, when a new name is not provided (:issue:`11193`) |
There was a problem hiding this comment.
use double backticks around Index and DatetimeIndex
|
minor correction, pls rebase as I just merged #11497 which might impact this. ping when green. |
|
I've just rebased @jreback I'll add a second comment when the Travis CI build is completed. |
| idx = self.create_index() | ||
| idx.name = "foo" | ||
| result = pd.Index(idx) | ||
| self.assertTrue("'foo'" in str(result)) |
There was a problem hiding this comment.
just construct the expected index and compare
expected = self.create_index()
expected.name = 'foo'
tm.assert_index_equal(result, expected)
also check with a new passed name
There was a problem hiding this comment.
I can work on this. Please ignore the last PR.
There was a problem hiding this comment.
@jreback What about the case when the object passed to Index is a MultiIndex? I think this is a corner case and should be avoided in principle. As a workaround, I would write the unit test as follows, even though is not very elegant:
expected = self.create_index()
if not isinstance(expected, MultiIndex):
expected.name = 'foo'
result = pd.Index(expected)
tm.assert_index_equal(result, expected)
result = pd.Index(expected, name='bar')
expected.name = 'bar'
tm.assert_index_equal(result, expected)
Should Index raise an error when we try to pass a MultiIndex object?
|
@jreback let me know if this needs further changes. |
| @@ -192,6 +193,21 @@ Bug Fixes | |||
|
|
|||
|
|
|||
| - Bug in ``.loc`` result with duplicated key may have ``Index`` with incorrect dtype (:issue:`11497`) | |||
|
|
|||
|
|
|||
There was a problem hiding this comment.
need a doc-note, and you are changing things which are not relevant. e.g. pls rebase on master first.
|
@IamGianluca can you update |
…ame, when a new name is not provided fix missing lines in whatsnew cosmetic change in whatsnew refactor unit test, failing for MultiIndex short sentence in whatsnew doc fix unit test remove commented code fix typo add doc string add unit test for multiindex case
|
I've updated the unit test @jreback |
|
merged via fc40dcc thanks! |
This PR is meant to fix issue #11193
The issue prevent, when creating a new
Indexobject based on a passedIndexorDatetimeIndexobject, to inherit the existingIndexname if a new name is not provided.I've also created two unit test which test the new behaviour on both cases (passing an
IndexorDatetimeIndexobject) and added a comment in the whatsnew documentation.