Nicer handling of incorrect usage of DataFrameGroupBy #8695
Description
I was trying to debug a DataFrameGroupBy operation and naively tried to .compute() it. This raised a KeyError that the compute column was missing.
import dask
df = dask.datasets.timeseries()
df.groupby("name").compute()
---------------------------------------------------------------------------
KeyError Traceback (most recent call last)
...
-> 2003 raise AttributeError(e) from e
AttributeError: 'Column not found: compute'In hindsight, this is clearly not the right thing to do. On the other hand, as a naive Pandas-familiar user, I was trying to "compute" the Pandas GroupBy object so that I could inspect it. I think it would be nice to have a more useful error in this case. However, since dask.dataframe.DataFrameGroupBy re-raises the KeyError as an AttributeError, I'm not sure what the appropriate resolution should be 😄
Full traceback 👇
---------------------------------------------------------------------------
KeyError Traceback (most recent call last)
File ~/GitHub/dask/dask/dataframe/groupby.py:2001, in DataFrameGroupBy.__getattr__(self, key)
2000 try:
-> 2001 return self[key]
2002 except KeyError as e:
File ~/GitHub/dask/dask/dataframe/groupby.py:1987, in DataFrameGroupBy.__getitem__(self, key)
1986 # error is raised from pandas
-> 1987 g._meta = g._meta[key]
1988 return g
File ~/mambaforge/envs/test-environment/lib/python3.9/site-packages/pandas/core/groupby/generic.py:1538, in DataFrameGroupBy.__getitem__(self, key)
1532 warnings.warn(
1533 "Indexing with multiple keys (implicitly converted to a tuple "
1534 "of keys) will be deprecated, use a list instead.",
1535 FutureWarning,
1536 stacklevel=2,
1537 )
-> 1538 return super().__getitem__(key)
File ~/mambaforge/envs/test-environment/lib/python3.9/site-packages/pandas/core/base.py:232, in SelectionMixin.__getitem__(self, key)
231 if key not in self.obj:
--> 232 raise KeyError(f"Column not found: {key}")
233 subset = self.obj[key]
KeyError: 'Column not found: compute'
The above exception was the direct cause of the following exception:
AttributeError Traceback (most recent call last)
Input In [3], in <module>
----> 1 df.groupby("name").compute()
File ~/GitHub/dask/dask/dataframe/groupby.py:2003, in DataFrameGroupBy.__getattr__(self, key)
2001 return self[key]
2002 except KeyError as e:
-> 2003 raise AttributeError(e) from e
AttributeError: 'Column not found: compute'