ENH: Support ExtensionArray in Groupby #20502
Conversation
```python
In [1]: import pandas as pd
In [2]: from cyberpandas import IPArray
In [3]: df = pd.DataFrame({"A": IPArray([0, 0, 1, 2, 2]), "B": [1, 5, 1, 1, 3]})
In [4]: df
Out[4]:
A B
0 0.0.0.0 1
1 0.0.0.0 5
2 0.0.0.1 1
3 0.0.0.2 1
4 0.0.0.2 3
In [5]: df.groupby("A").B.mean()
Out[5]:
A
0.0.0.1 1
0.0.0.2 2
Name: B, dtype: int64
```
TomAugspurger
added
Groupby
ExtensionArray
|
What I have so far is relatively straightforward (surprisingly). But I'm probably missing things. Are there edge cases or other operations we should test? |
Codecov Report
@@ Coverage Diff @@
## master #20502 +/- ##
==========================================
+ Coverage 91.82% 91.84% +0.02%
==========================================
Files 152 152
Lines 49249 49249
==========================================
+ Hits 45225 45235 +10
+ Misses 4024 4014 -10
Continue to review full report at Codecov.
|
|
If you use |
|
Not easily. By the time we're wrapping up the output, we've long since converted the input to an Index. That said, once we have ExtensionIndexes, it should be a one-line change: Line 3037 in cdfce2b |
|
lgtm. and another reason to make EA a base class for |
Note that right now
Out[5].indexjust just anIndexwith object dtype. In the future, we could tie an Index type to an ExtensionArray type, and ensure that the extension type propagates through.