Add tests to ensure sort preserved by groupby, add docs #10931
Conversation
nickeubank
force-pushed
the
test_groupby_sort_preservation
branch
5 times, most recently
from
ed6eb1a to
8bcd4a5
Compare
| g.get_group('A') | ||
|
|
||
| g.get_group('B') | ||
|
|
There was a problem hiding this comment.
hmm, this is True, but I think the average user will find this a bit confusing. Can you expand/reword a bit. (its an important point), but not sure how to describe it.
There was a problem hiding this comment.
Ok -- dropped the discussion of what exactly is happening behind the scenes, tried to clear up the explanation above, and fleshed out the examples a little more. That clearer?
There was a problem hiding this comment.
Is there any reason a user would even expect that the already created groupby object would change after sorting the original frame?
Is there an example of confusion? As adding this explanation of something that I would have never thought of seems to make it only more complex than it should be IMO
There was a problem hiding this comment.
I agree with Joris. This behavior is entirely consistent with the standard memory model in Python. It does not need explaining.
nickeubank
force-pushed
the
test_groupby_sort_preservation
branch
2 times, most recently
from
c77e701 to
f579d6e
Compare
|
@jreback updated! |
|
|
||
| .. note:: | ||
|
|
||
| Users should be careful about re-sorting their data after execution of ``groupby()``, however. ``groupby()`` ensures that the order of observations within each group will always be the same as they were in the original data at the time that ``groupby()`` was executed. If a user creates a ``groupby()`` object *and then re-sorts the original data*, the order of observations within each group created by ``groupby()`` will reflect the order in which observations were sorted when ``groupby()`` was executed, **not** the order of observations following the re-sorting. For example: |
There was a problem hiding this comment.
Can you hard wrap here?
|
The basic part on sort=True/False and sort preserved within groups looks good! However, I don't really seem to grasp the need to explain the sorting after groupby: "Users should be careful about re-sorting their data after execution of |
nickeubank
force-pushed
the
test_groupby_sort_preservation
branch
2 times, most recently
from
b060c0e to
4817cb3
Compare
|
@jorisvandenbossche @shoyer @jreback OK, warning dropped! |
| Sort group keys. Get better performance by turning this off | ||
| Sort group keys. Get better performance by turning this off. | ||
| Note this does not influence the order of observations within each group. | ||
| groupby preserves the sorted order of the target object |
There was a problem hiding this comment.
sorted order -> order ? (as the order within a group is not necessarily sorted?)
There was a problem hiding this comment.
what about "order in which observations appear"?
There was a problem hiding this comment.
or just order of observations?
There was a problem hiding this comment.
or 'rows'?
groupby preserves the order of the rows within each group
nickeubank
force-pushed
the
test_groupby_sort_preservation
branch
from
4817cb3 to
3b49fad
Compare
|
@jorisvandenbossche great, added. |
Add tests to ensure sort preserved by groupby, add docs
|
@nickeubank thanks! |
xref #9651
closes #8588
Adds test to ensure the sort of a target object is preserved within
groupby()groups, modifies docs to make it clear sort is preserved within groups.