BUG: pivot_table always returns a DataFrame #13554
Conversation
pandas/tools/tests/test_pivot.py
Outdated
| 'col2': ['C', 'D', 'E'], | ||
| 'col3': [1, 3, 9]}) | ||
|
|
||
| pivoted_1 = df.pivot_table('col1', index=['col3', 'col2'], aggfunc=np.sum) |
There was a problem hiding this comment.
can you replicate all 3 tests form the issue:
pandas/tools/tests/test_pivot.py
Outdated
| index=Index([1, 3, 9], name='col3'), | ||
| columns=m) | ||
|
|
||
| tm.assert_frame_equal(result, expected) |
There was a problem hiding this comment.
I notice case 3 differs from the issue.
Should I rewrite to match with the issue?
The issue:
sum
col3
1 3
3 4
9 5
<class 'pandas.core.frame.DataFrame'>This PR:
sum
col1
col3
1 3
3 4
9 5
<class 'pandas.core.frame.DataFrame'>
jreback
changed the title
pivot_table always returns a DataFrame
|
can you rebase / update? |
|
can you rebase and add a release note for 0.20.0 |
|
Sorry to my late response, I rebased and add note for 0.20.0. |
|
can you rebase |
|
Rebased ! |
|
@yui-knk looks like some test failures: https://travis-ci.org/pandas-dev/pandas/jobs/186874138#L1939 Can you test those out locally and see if you reproduce? |
doc/source/whatsnew/v0.20.0.txt
Outdated
| @@ -321,3 +321,4 @@ Bug Fixes | |||
| - Require at least 0.23 version of cython to avoid problems with character encodings (:issue:`14699`) | |||
| - Bug in converting object elements of array-like objects to unsigned 64-bit integers (:issue:`4471`) | |||
| - Bug in ``pd.pivot_table()`` where no error was raised when values argument was not in the columns (:issue:`14938`) | |||
| - Bug in ``pivot_table`` returns ``Series`` in specific circumstance (:issue:`4386`) | |||
There was a problem hiding this comment.
add, when it should always return a DataFrame.
|
can you rebase. I think you have a failing test as well. |
|
can you rebase/update |
Before this commit, if * `values` is not list like * `columns` is `None` * `aggfunc` is not instance of `list` `pivot_table` returns a `Series`. This commit adds checking for `columns.nlevels` is greater than 1 to prevent from casting `table` to a `Series`. This will fix pandas-dev#4386.
|
I think you need to change this test to return DataFrame rather than Series. otherwise looks ok: https://travis-ci.org/pandas-dev/pandas/jobs/208224356 |
|
also, I'd like to have a sub-section in the whatsnew which shows the case where this used to return a Series, but now returns a DataFrame (you can use one of your tests examples). |
Before this commit, if * `values` is not list like * `columns` is `None` * `aggfunc` is not instance of `list` `pivot_table` returns a `Series`. This commit adds checking for `columns.nlevels` is greater than 1 to prevent from casting `table` to a `Series`. This will fix #4386. DOC: add docs for #13554
|
thanks @yui-knk rebase and merged. |
Before this commit, if * `values` is not list like * `columns` is `None` * `aggfunc` is not instance of `list` `pivot_table` returns a `Series`. This commit adds checking for `columns.nlevels` is greater than 1 to prevent from casting `table` to a `Series`. This will fix pandas-dev#4386. DOC: add docs for pandas-dev#13554
git diff upstream/master | flake8 --diffBefore this commit, if
valuesis not list likecolumnsisNoneaggfuncis not instance oflistpivot_tablereturns aSeries.This commit adds checking for
columns.nlevelsisgreater than 1 to prevent from casting
tabletoa
Series.This will fix #4386.