BUG: Improved error msg #32855
BUG: Improved error msg #32855
Conversation
jreback
added
Error Reporting
|
I made some changes. Please let me know if this is what you had mind. Thanks! |
| s = pd.Series(np.linspace(1, 2, 2), index=pd.MultiIndex.from_tuples(idx)) | ||
|
|
||
| with pytest.raises(KeyError, match=r".+? indexing with .+? failing for MultiIndex"): | ||
| df.at["a_row", "A"] |
There was a problem hiding this comment.
is there a similar problem for at.__setitem__?
There was a problem hiding this comment.
Multiindex with at.__setitem__ doesn't raise an exception and sets items at level 0:
In [1]: import pandas as pd
...: import numpy as np
...:
...: cols = [("a_col", chr(i + 65)) for i in range(2)]
...: idx = [("a_row", chr(i + 65)) for i in range(2)]
...: df = pd.DataFrame(
...: np.linspace(1, 4, 4).reshape(2, 2),
...: index=pd.MultiIndex.from_tuples(idx),
...: columns=pd.MultiIndex.from_tuples(cols),
...: )
...: display(df)
a_col
A B
a_row A 1.0 2.0
B 3.0 4.0
In [2]: df.at["a_row", "A"] = 888.0
...: display(df)
a_col
A B
a_row A 888.0 888.0
B 3.0 4.0
In [3]: df.at["a_row", "A"] = [888.0, 999.0]
...: display(df)
a_col
A B
a_row A 888.0 999.0
B 3.0 4.0
| if isinstance(self.obj.index, ABCMultiIndex): | ||
| raise KeyError( | ||
| f"Detected KeyError {err}, indexing with {key} " | ||
| "failing for MultiIndex" |
There was a problem hiding this comment.
doing this on master, the DataFrame case is giving a reasonable message, while the Series case is unhelpful. Maybe condition this on self.ndim == 1 and add a suggestion "pass a tuple"
There was a problem hiding this comment.
It the Series case goes down a different code path where an exception is raised in pandas/utils/_validators.py. Would adding an except for TypeError with a more informative error message work?
In [3]: ser.at["a_row", "A"]
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
<ipython-input-3-2aec4c8078ba> in <module>
----> 1 ser.at["a_row", "A"]
/workspaces/pandas-aykhan/pandas/core/indexing.py in __getitem__(self, key)
2090 return self.obj.loc[key]
2091
-> 2092 return super().__getitem__(key)
2093
2094 def __setitem__(self, key, value):
/workspaces/pandas-aykhan/pandas/core/indexing.py in __getitem__(self, key)
2035 try:
2036 # import pdb; pdb.set_trace()
-> 2037 return self.obj._get_value(*key, takeable=self._takeable)
2038 except KeyError as err:
2039 if isinstance(self.obj.index, ABCMultiIndex):
TypeError: _get_value() got multiple values for argument 'takeable'
| df.at["a_row", "A"] | ||
|
|
||
| with pytest.raises(TypeError, match=r".+? got multiple values for argument .+?"): | ||
| ser.at["a_row", "A"] |
There was a problem hiding this comment.
this looks like a bug. ser.at["a_row", "A"] should return 1.0, same as ser.loc["a_row", "A"]
There was a problem hiding this comment.
I tested this on my PR #32520, it returns 1.0.
There was a problem hiding this comment.
I can remove the series from the test. @rhshadrach is correct that the original context was DataFrames, and improving the error message should clarify why the KeyError is being raised.
|
closing as stale, if you'd like to update as above, pls ping / open a new PR. |
black pandasgit diff upstream/master -u -- "*.py" | flake8 --diffHopefully this is a more informative message. The original bug report mentions that the index key also fails for Series but didn't report the error type, so I took my best guess at the reporter's intent in the test.