ENH: Add future.python_scalars #63016
Open
+323
−129
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Member
Author
Member
Member
Author
Plan to run a full set of ASVs next week, some microbenchmarks from pandas.core.dtypes.cast import maybe_unbox_numpy_scalar
with pd.option_context("python_scalars", True):
%timeit maybe_unbox_numpy_scalar(np.int64(2))
# 828 ns ± 9.91 ns per loop (mean ± std. dev. of 7 runs, 1,000,000 loops each)
%timeit maybe_unbox_numpy_scalar(2)
# 161 ns ± 0.414 ns per loop (mean ± std. dev. of 7 runs, 10,000,000 loops each)
ser = pd.Series([1, 2, 3] * 10_000)
with pd.option_context("python_scalars", True):
%timeit ser.sum()
# 9.42 μs ± 423 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)
with pd.option_context("python_scalars", False):
%timeit ser.sum()
# 8.28 μs ± 137 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each) |
1 task
Dr-Irv
reviewed
Nov 10, 2025
Member
Author
|
Full ASVs are below, only showing where there was a 10% of more regression. In the full list, only the following two actually hit the function Full listI was curious why only min/max showed up as being regressions in series_methods.NanOps |
Member
Author
|
@jbrockmendel - you good with the ASVs here? |
Member
|
No complaints here. |
6 tasks
| else: | ||
| result = result.reshape(1) | ||
| if using_python_scalars(): | ||
| result = np.array([result]) |
Member
There was a problem hiding this comment.
why doing this instead of maybe_unbox_numpy_scalar?
Member
Author
There was a problem hiding this comment.
The result here prior to L1543 is already a Python scalar when future.python_scalars=True due to calling the reduction function. In this block, keepdims=True so we need to convert it to a NumPy array.
| if isinstance(result, np.longdouble): | ||
| result = float(result) | ||
| else: | ||
| result = value.item() |
Member
There was a problem hiding this comment.
I know this will mess up on a timedelta64:
obj = np.timedelta64(1, "ns")
assert isinstance(obj, np.generic)
>>> obj.item()
1
I don't know if there are other cases where obj.item() messes up, but I'm wary of it. Heads up.
Member
Author
There was a problem hiding this comment.
Thanks - will add a test for all dtypes. Here is the full list of scalars and their corresponding item type without datetime/timedelta. Only other problematic one is complex256.
? bool <class 'bool'>
b int8 <class 'int'>
h int16 <class 'int'>
i int32 <class 'int'>
l int64 <class 'int'>
q int64 <class 'int'>
n int64 <class 'int'>
p int64 <class 'int'>
B uint8 <class 'int'>
H uint16 <class 'int'>
I uint32 <class 'int'>
L uint64 <class 'int'>
Q uint64 <class 'int'>
N uint64 <class 'int'>
P uint64 <class 'int'>
e float16 <class 'float'>
f float32 <class 'float'>
d float64 <class 'float'>
g float128 <class 'numpy.longdouble'>
F complex64 <class 'complex'>
D complex128 <class 'complex'>
G complex256 <class 'numpy.clongdouble'>
S |S1 <class 'bytes'>
U <U1 <class 'str'>
V |V0 <class 'bytes'>
For datetime/timedelta, the arrow dtypes already return pd.Timedelta and pd.Timestamp. I think we can align the NumPy dtypes to do the same.
| result = getattr(df.C, op)() | ||
| assert isinstance(result, np.float64) | ||
| if using_python_scalars: | ||
| assert isinstance(result, float) |
Member
There was a problem hiding this comment.
i think float64 subclasses float, so this won't exclude float64
| if using_python_scalars: | ||
| assert isinstance(result, int) | ||
| else: | ||
| assert result.dtype == "uint64" |
Member
There was a problem hiding this comment.
just check type rather than dtype?
| assert 0 == s.skew() | ||
| assert isinstance(s.skew(), np.float64) # GH53482 | ||
| if using_python_scalars: | ||
| assert isinstance(s.skew(), float) |
pandas/tests/series/test_ufunc.py
Outdated
| tm.assert_series_equal(result, expected) | ||
| else: | ||
| expected = values[1] | ||
| if using_python_scalars and values.dtype.kind in ["i", "f"]: |
Member
There was a problem hiding this comment.
NBD but checking for kind in "if" is very slightly faster than checking for kind in ["i", "f"].
| if isinstance(all_data, Series): | ||
| assert not (std_x < 0).any() | ||
| else: | ||
| assert not (std_x < 0).any().any() |
Member
There was a problem hiding this comment.
i guess np.bool_(True).any() returns itself? thats kind of convenient. could go in the list joris asked for of potential downsides
Member
Author
There was a problem hiding this comment.
For this particular case, I'd say axis=None should be preferred. But perhaps there are others where that is useful.
| if isinstance(all_data, Series): | ||
| assert not (var_x < 0).any() | ||
| else: | ||
| assert not (var_x < 0).any().any() |
Member
There was a problem hiding this comment.
if this pattern is going to show up a lot, could make a helper in pd._testing?
Member
Author
There was a problem hiding this comment.
Changed to axis=None, no branching.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
doc/source/whatsnew/vX.X.X.rstfile if fixing a bug or adding a new feature.Adds an experimental option to return Python scalars instead of NumPy scalars across the API. This is not yet fully implemented everywhere, e.g.
Series.__getitem__, but I'm hoping reductions are a substantial chunk.This is complicated by #62988 where it was found that many of our doctests are not running. We run those doctests using NumPy>=2, and if we were to get those doctests to pass as-is, we would need to change the NumPy reprs from e.g.
2tonp.int64(2). If we then change reductions et al to returning Python scalars, we'd then change all the reprs back from e.g.np.int64(2)to2. So instead I think we can:future.python_scalarstoTruein 4.0, deprecate the future option.