BUG/TST: Series round with dtype object #62174
Conversation
pandas/core/series.py
Outdated
| return self._constructor( | ||
| new_values, index=self.index, copy=False | ||
| ).__finalize__(self, method="map") | ||
| except TypeError as e: |
There was a problem hiding this comment.
nitpick e->err. try to avoid single-letter variable names
pandas/core/series.py
Outdated
| raise TypeError("Expected numeric dtype, got object instead.") | ||
| try: | ||
| round_func = functools.partial(round, ndigits=decimals) | ||
| new_values = self._map_values(round_func) |
There was a problem hiding this comment.
only this line needs to go inside the try/except
In another PR you're going to make DataFrame do the same thing this does, right? If so, you'll need to change the Block behavior. |
| ser = Series([0.232], dtype="object") | ||
| expected = Series([0.2]) | ||
| tm.assert_series_equal(ser.round(1), expected) | ||
| ser2 = Series(["bar"], dtype="object") |
There was a problem hiding this comment.
can you make this a separate test
| # GH#61206, GH#62174 | ||
| ser = Series([0.232], dtype="object") | ||
| expected = Series([0.2]) | ||
| tm.assert_series_equal(ser.round(1), expected) |
There was a problem hiding this comment.
separate line for result = ser.round(1)
|
i suspect this will also fix/change the DataFrame.round behavior? can you test that and address it in the whatsnew note |
pandas/core/internals/blocks.py
Outdated
| round_func = functools.partial(round, ndigits=decimals) | ||
| mapper = functools.partial(algos.map_array, mapper=round_func) | ||
| try: | ||
| if self.values.ndim == 1: | ||
| values = algos.map_array(self.values, round_func) | ||
| else: | ||
| values = np.apply_along_axis(mapper, 0, self.values) |
There was a problem hiding this comment.
Definitely feels like there is a better way to account for 2d blocks but I couldn't think of a better way. Part of that was I couldn't find where in lib.map_infer() was defined (from the return statement in algos.map_array()) and thus didn't know why algos.map_array wasn't working for the 2d ndarray.
There was a problem hiding this comment.
algos.map_array(self.values.ravel(), round_func).reshape(self.values.shape)
There was a problem hiding this comment.
ohh okay makes sense, thank you
pandas/core/internals/blocks.py
Outdated
| @final | ||
| def __repr__(self) -> str: | ||
| # don't want to print out all of the items here | ||
| # don't want to out all of the items here |
There was a problem hiding this comment.
oops, thanks for catching that
2 participants
doc/source/whatsnew/vX.X.X.rstfile if fixing a bug or adding a new feature.I changed the behavior or series.round() to act like
series.map(round)if the dtype is object as discussed in #61682. I put this in a new PR as the original intent of the other PR was to change DataFrame behavior. I'm unsure if the way I approached it is a good way to make the change though because I saw incore/internals/blocks.pythatBlock'sround()method is intended to raise an error for dtype of object. So I wasn't sure I should approach this by updated theBlockclass instead. Any feedback is appreciated!