feat: FloatingArray, Series.__new__ + astype and also for Index
#1493
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
cmp0xff
commented
Nov 18, 2025
Comment on lines
-2367
to
-2410
| ASTYPE_FLOAT_ARGS: list[tuple[FloatDtypeArg, type]] = [ | ||
| # python float | ||
| (float, np.floating), | ||
| ("float", np.floating), | ||
| # pandas Float32 | ||
| (pd.Float32Dtype(), np.float32), | ||
| ("Float32", np.float32), | ||
| # pandas Float64 | ||
| (pd.Float64Dtype(), np.float64), | ||
| ("Float64", np.float64), | ||
| # numpy float16 | ||
| (np.half, np.half), | ||
| ("half", np.half), | ||
| ("e", np.half), | ||
| ("float16", np.float16), | ||
| ("f2", np.float16), | ||
| # numpy float32 | ||
| (np.single, np.single), | ||
| ("single", np.single), | ||
| ("f", np.single), | ||
| ("float32", np.float32), | ||
| ("f4", np.float32), | ||
| # numpy float64 | ||
| (np.double, np.double), | ||
| ("double", np.double), | ||
| ("d", np.double), | ||
| ("float64", np.float64), | ||
| ("f8", np.float64), | ||
| # numpy float128 | ||
| (np.longdouble, np.longdouble), | ||
| ("longdouble", np.longdouble), | ||
| ("g", np.longdouble), | ||
| ("f16", np.longdouble), | ||
| # ("float96", np.longdouble), # NOTE: unsupported | ||
| ("float128", np.longdouble), # NOTE: UNIX ONLY | ||
| # pyarrow float32 | ||
| ("float32[pyarrow]", float), | ||
| ("float[pyarrow]", float), | ||
| # pyarrow float64 | ||
| ("float64[pyarrow]", float), | ||
| ("double[pyarrow]", float), | ||
| ] |
Contributor
Author
There was a problem hiding this comment.
Moved to tests/__init__.py
| assert_type(s.astype("uint64[pyarrow]"), "pd.Series[int]") | ||
|
|
||
|
|
||
| @pytest.mark.parametrize("cast_arg, target_type", ASTYPE_FLOAT_ARGS, ids=repr) |
Contributor
Author
There was a problem hiding this comment.
Moved to test_series_float.py
cmp0xff
commented
Nov 18, 2025
Comment on lines
+102
to
+103
| # if TYPE_CHECKING_INVALID_USAGE: | ||
| # # numpy float16 | ||
| # pd.Index([1.0], dtype=np.half) | ||
| # pd.Index([1.0], dtype="half") | ||
| # pd.Index([1.0], dtype="float16") | ||
| # pd.Index([1.0], dtype="e") | ||
| # pd.Index([1.0], dtype="f2") |
Contributor
Author
There was a problem hiding this comment.
At the moment, our __new__ overloads in Index cover everything including these invalid ones. Maybe we will clean them up in the future and recognise them as invalid.
Comment on lines
+158
to
+162
| # if TYPE_CHECKING_INVALID_USAGE: | ||
| # # numpy float16 | ||
| # s.astype(np.half) | ||
| # s.astype("half") | ||
| # s.astype("float16") | ||
| # s.astype("e") | ||
| # s.astype("f2") |
Contributor
Author
There was a problem hiding this comment.
At the moment, our __new__ overloads in Index cover everything including these invalid ones. Maybe we will clean them up in the future and recognise them as invalid.
| # Refer to https://numpy.org/doc/stable/reference/arrays.datetime.html#datetime-units | ||
| TimedeltaDtypeArg: TypeAlias = Literal[ | ||
| ComplexDtypeArg: TypeAlias = BuiltinComplexDtypeArg | NumpyComplexDtypeArg | ||
| PandasAstypeTimedeltaDtypeArg: TypeAlias = Literal[ |
Contributor
Author
There was a problem hiding this comment.
import pandas as pd
pd.Series(["2025-01-01"], dtype="datetime64[D]") # TypeError: dtype=datetime64[D] is not supported. Supported resolutions are 's', 'ms', 'us', and 'ns'These only work upon astype.
Comment on lines
+365
to
+368
| | Literal["F", "c8", "complex64", "csingle"] | ||
| # https://numpy.org/doc/stable/reference/arrays.scalars.html#numpy.cdouble | ||
| | type[np.cdouble] | ||
| | Literal["D", "c16", "complex128", "cdouble", "cfloat", "complex_"] | ||
| | Literal["D", "c16", "complex128", "cdouble"] |
Contributor
Author
There was a problem hiding this comment.
The removed values do not work for me in pytests at runtime. I think they are NON_NUMPY20_ALIASES.
cmp0xff
commented
Nov 18, 2025
Comment on lines
+82
to
+83
| # Builtin bool type and its string alias | ||
| BuiltinBooleanDtypeArg: TypeAlias = type[bool] | Literal["bool"] | ||
| # Pandas nullable boolean type and its string alias | ||
| PandasBooleanDtypeArg: TypeAlias = pd.BooleanDtype | Literal["boolean"] |
Contributor
Author
There was a problem hiding this comment.
It is a bit annoying that all these has to be duplicated in tests/__init__.py.
cmp0xff
force-pushed
the
feature/floating-array
branch
from
b5454cb to
ee8615a
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
This is a PR simpler than #1469 which focuses more on adding
FloatingArray.I have made many more categories for the
DtypeArgs, because of the following reasons:Only
Pandasdtypes are relevant for making aFloatingArray. Native andNumpydtypes lead to aNumpyExtensionArray, andPyArrowdtypes lead to anArrowExtensionArray, if notArrowStringArrayPandassupports more dtypes inastypethan in constructorsIndexdoes not supportnp.float16Closes
FloatingArraydocumented but missing from stubs #1428Tests added