Reject an overload that violates a restriction under a gradual expected type (#5240) by kinto0 · Pull Request #5240 · facebook/pyrefly

kinto0 · 2026-10-14T21:05:01Z

Summary:

Stack context: Make a restriction violation under a gradual expected type reject the overload that caused it.

Problem: a constructor call with a gradual expected type, such as return pd.Series(pd.NA, dtype="Int64") under a bare -> pd.Series, reports bad-specialization. The same call without the annotation type checks clean. The trigger is an overloaded __new__ or __init__ where an earlier overload accepts the argument only through the class type parameter (data: S1 | ...), and a later overload accepts it directly (data: NAType).
Cause: since D115555748, the expected type solves S1 to Any, and argument matching records a restriction violation on that answer. The constructor reports the violation only when it finishes, so overload selection does not see it. The earlier overload matches, and the call reports the violation instead of choosing the later overload.
Fix: each call that matches arguments against a restricted answer now takes the violations that it caused into its own specialization errors. Overload selection rejects the overload and continues, as it does without the expected type.
Side effect: for a constructor without overloads, the violation is now a hard error of the hinted attempt, so construct_with_hint discards the expected type. The call then reports the same errors as the unhinted call. This matches the intent of D115555748: an Any from the expected type says nothing, and argument matching checks the call normally.

Reviewed By: rchen152

Differential Revision: D125293977

meta-codesync · 2026-10-14T21:05:13Z

@kinto0 has exported this pull request. If you are a Meta employee, you can view the originating Diff in D125293977.

rchen152

Review automatically exported from Phabricator review in Meta.

…ded constructors Summary: Stack context: Fix overload selection errors that cause false positives in pandas-stubs. - Add two bug-marked tests for constructor calls that violate a restriction under a gradual expected type. With overloads, `Series(NA(), dtype="Int64")` under a bare `-> Series` reports `bad-specialization`, but the unhinted call picks the `NA` overload and reports nothing. Without overloads, the hinted call hides a bad argument that the unhinted call reports. - The first test is the pattern that `pd.Series(pd.NA, dtype="Int64")` hits with pandas-stubs since D115555748. The next diff fixes both tests. Differential Revision: D125293976

…ed type (facebook#5240) Summary: Stack context: Make a restriction violation under a gradual expected type reject the overload that caused it. - Problem: a constructor call with a gradual expected type, such as `return pd.Series(pd.NA, dtype="Int64")` under a bare `-> pd.Series`, reports `bad-specialization`. The same call without the annotation type checks clean. The trigger is an overloaded `__new__` or `__init__` where an earlier overload accepts the argument only through the class type parameter (`data: S1 | ...`), and a later overload accepts it directly (`data: NAType`). - Cause: since D115555748, the expected type solves `S1` to `Any`, and argument matching records a restriction violation on that answer. The constructor reports the violation only when it finishes, so overload selection does not see it. The earlier overload matches, and the call reports the violation instead of choosing the later overload. - Fix: each call that matches arguments against a restricted answer now takes the violations that it caused into its own specialization errors. Overload selection rejects the overload and continues, as it does without the expected type. - Side effect: for a constructor without overloads, the violation is now a hard error of the hinted attempt, so `construct_with_hint` discards the expected type. The call then reports the same errors as the unhinted call. This matches the intent of D115555748: an `Any` from the expected type says nothing, and argument matching checks the call normally. Reviewed By: rchen152 Differential Revision: D125293977

…ed type (facebook#5240) Summary: Pull Request resolved: facebook#5240 Stack context: Make a restriction violation under a gradual expected type reject the overload that caused it. - Problem: a constructor call with a gradual expected type, such as `return pd.Series(pd.NA, dtype="Int64")` under a bare `-> pd.Series`, reports `bad-specialization`. The same call without the annotation type checks clean. The trigger is an overloaded `__new__` or `__init__` where an earlier overload accepts the argument only through the class type parameter (`data: S1 | ...`), and a later overload accepts it directly (`data: NAType`). - Cause: since D115555748, the expected type solves `S1` to `Any`, and argument matching records a restriction violation on that answer. The constructor reports the violation only when it finishes, so overload selection does not see it. The earlier overload matches, and the call reports the violation instead of choosing the later overload. - Fix: each call that matches arguments against a restricted answer now takes the violations that it caused into its own specialization errors. Overload selection rejects the overload and continues, as it does without the expected type. - Side effect: for a constructor without overloads, the violation is now a hard error of the hinted attempt, so `construct_with_hint` discards the expected type. The call then reports the same errors as the unhinted call. This matches the intent of D115555748: an `Any` from the expected type says nothing, and argument matching checks the call normally. Reviewed By: rchen152 Differential Revision: D125293977

github-actions · 2026-10-14T21:59:48Z

Diff from mypy_primer, showing the effect of this PR on open source code:

============================================================
SUMMARY
============================================================
Total: +15 new errors, -31 fixed errors
By preset: +15/-29 (default), +15/-29 (strict)

Projects with changes (6):
  steam.py: +1 -1
  spark: +0 -14
  pandas-stubs: +0 -1
  pandas: +1 -0
  meson: +13 -13
  xarray: +0 -2
============================================================

FULL DIFF DETAILS
------------------------------------------------------------

steam.py (https://github.com/Gobot1234/steam.py)
- ERROR steam/abc.py:552:30-561:10: `PartialApp | tuple[str, int]` is not assignable to upper bound `PartialApp` of type variable `AppT` [bad-specialization]
+ ERROR steam/abc.py:552:30-561:10: `PartialApp[None] | tuple[str, int]` is not assignable to upper bound `PartialApp` of type variable `AppT` [bad-specialization]

spark (https://github.com/apache/spark)
- ERROR python/pyspark/eval_handlers/_pandas.py:812:47-56: `Unknown | None` is not assignable to upper bound `BaseOffset | CategoricalDtype | ExtensionDtype | Interval | Period | bool | bytes | complex | date | datetime | dtype | float | int | list[str] | str | time | timedelta | type[bool | complex | object | str]` of type variable `S1` [bad-specialization]
- ERROR python/pyspark/tests/upstream/pandas/test_pandas_series_astype.py:99:44-72: `object` is not assignable to upper bound `BaseOffset | CategoricalDtype | ExtensionDtype | Interval | Period | bool | bytes | complex | date | datetime | dtype | float | int | list[str] | str | time | timedelta | type[bool | complex | object | str]` of type variable `S1` [bad-specialization]
- ERROR python/pyspark/tests/upstream/pandas/test_pandas_series_astype.py:100:44-79: `object` is not assignable to upper bound `BaseOffset | CategoricalDtype | ExtensionDtype | Interval | Period | bool | bytes | complex | date | datetime | dtype | float | int | list[str] | str | time | timedelta | type[bool | complex | object | str]` of type variable `S1` [bad-specialization]
- ERROR python/pyspark/tests/upstream/pandas/test_pandas_series_astype.py:101:51-84: `object` is not assignable to upper bound `BaseOffset | CategoricalDtype | ExtensionDtype | Interval | Period | bool | bytes | complex | date | datetime | dtype | float | int | list[str] | str | time | timedelta | type[bool | complex | object | str]` of type variable `S1` [bad-specialization]
- ERROR python/pyspark/tests/upstream/pandas/test_pandas_series_astype.py:102:43-74: `object` is not assignable to upper bound `BaseOffset | CategoricalDtype | ExtensionDtype | Interval | Period | bool | bytes | complex | date | datetime | dtype | float | int | list[str] | str | time | timedelta | type[bool | complex | object | str]` of type variable `S1` [bad-specialization]
- ERROR python/pyspark/tests/upstream/pandas/test_pandas_series_astype.py:103:43-96: `object` is not assignable to upper bound `BaseOffset | CategoricalDtype | ExtensionDtype | Interval | Period | bool | bytes | complex | date | datetime | dtype | float | int | list[str] | str | time | timedelta | type[bool | complex | object | str]` of type variable `S1` [bad-specialization]
- ERROR python/pyspark/tests/upstream/pandas/test_pandas_series_astype.py:104:41-86: `object` is not assignable to upper bound `BaseOffset | CategoricalDtype | ExtensionDtype | Interval | Period | bool | bytes | complex | date | datetime | dtype | float | int | list[str] | str | time | timedelta | type[bool | complex | object | str]` of type variable `S1` [bad-specialization]
- ERROR python/pyspark/tests/upstream/pyarrow/test_pyarrow_array_from_pandas_default.py:165:54-94: `object` is not assignable to upper bound `BaseOffset | CategoricalDtype | ExtensionDtype | Interval | Period | bool | bytes | complex | date | datetime | dtype | float | int | list[str] | str | time | timedelta | type[bool | complex | object | str]` of type variable `S1` [bad-specialization]
- ERROR python/pyspark/tests/upstream/pyarrow/test_pyarrow_array_from_pandas_default.py:171:47-75: `object` is not assignable to upper bound `BaseOffset | CategoricalDtype | ExtensionDtype | Interval | Period | bool | bytes | complex | date | datetime | dtype | float | int | list[str] | str | time | timedelta | type[bool | complex | object | str]` of type variable `S1` [bad-specialization]
- ERROR python/pyspark/tests/upstream/pyarrow/test_pyarrow_array_from_pandas_default.py:242:47-67: `object` is not assignable to upper bound `BaseOffset | CategoricalDtype | ExtensionDtype | Interval | Period | bool | bytes | complex | date | datetime | dtype | float | int | list[str] | str | time | timedelta | type[bool | complex | object | str]` of type variable `S1` [bad-specialization]
- ERROR python/pyspark/tests/upstream/pyarrow/test_pyarrow_array_from_pandas_default.py:243:54-245:10: `object` is not assignable to upper bound `BaseOffset | CategoricalDtype | ExtensionDtype | Interval | Period | bool | bytes | complex | date | datetime | dtype | float | int | list[str] | str | time | timedelta | type[bool | complex | object | str]` of type variable `S1` [bad-specialization]
- ERROR python/pyspark/tests/upstream/pyarrow/test_pyarrow_array_from_pandas_default.py:246:48-101: `object` is not assignable to upper bound `BaseOffset | CategoricalDtype | ExtensionDtype | Interval | Period | bool | bytes | complex | date | datetime | dtype | float | int | list[str] | str | time | timedelta | type[bool | complex | object | str]` of type variable `S1` [bad-specialization]
- ERROR python/pyspark/tests/upstream/pyarrow/test_pyarrow_array_from_pandas_default.py:359:57-361:10: `object` is not assignable to upper bound `BaseOffset | CategoricalDtype | ExtensionDtype | Interval | Period | bool | bytes | complex | date | datetime | dtype | float | int | list[str] | str | time | timedelta | type[bool | complex | object | str]` of type variable `S1` [bad-specialization]
- ERROR python/pyspark/tests/upstream/pyarrow/test_pyarrow_array_from_pandas_default.py:362:57-364:10: `object` is not assignable to upper bound `BaseOffset | CategoricalDtype | ExtensionDtype | Interval | Period | bool | bytes | complex | date | datetime | dtype | float | int | list[str] | str | time | timedelta | type[bool | complex | object | str]` of type variable `S1` [bad-specialization]

pandas-stubs (https://github.com/pandas-dev/pandas-stubs)
- ERROR tests/test_resampler.py:306:22-27: `DatetimeIndexResampler[Series]` is not assignable to upper bound `BaseOffset | CategoricalDtype | ExtensionDtype | Interval | Period | bool | bytes | complex | date | datetime | dtype | float | int | list[str] | str | time | timedelta | type[bool | complex | object | str]` of type variable `S1` [bad-specialization]

pandas (https://github.com/pandas-dev/pandas)
+ ERROR pandas/core/arrays/interval.py:699:35-40: Argument `DatetimeArray | NaTType | Period | Timedelta | TimedeltaArray | Timestamp | ndarray` is not assignable to parameter `right` with type `NaTType | Period | Timedelta | Timestamp` in function `pandas._libs.interval.Interval.__init__` [bad-argument-type]

meson (https://github.com/mesonbuild/meson)
+ ERROR mesonbuild/mparser.py:859:36-88: `+` is not supported between `str` and `None` [unsupported-operation]
- ERROR mesonbuild/mparser.py:859:27-32: The type of attribute `value` is unknown; it is inferred as an implicit `Any` [unknown-attribute-access]
- ERROR mesonbuild/mparser.py:859:83-88: The type of attribute `value` is unknown; it is inferred as an implicit `Any` [unknown-attribute-access]
- ERROR unittests/versiontests.py:302:43-63: `int` is not assignable to upper bound `Comparable` of type variable `_V` [bad-specialization]
- ERROR unittests/versiontests.py:303:43-64: `int` is not assignable to upper bound `Comparable` of type variable `_V` [bad-specialization]
- ERROR unittests/versiontests.py:307:43-84: `int` is not assignable to upper bound `Comparable` of type variable `_V` [bad-specialization]
- ERROR unittests/versiontests.py:310:44-65: `int` is not assignable to upper bound `Comparable` of type variable `_V` [bad-specialization]
- ERROR unittests/versiontests.py:311:44-64: `int` is not assignable to upper bound `Comparable` of type variable `_V` [bad-specialization]
- ERROR unittests/versiontests.py:314:45-65: `int` is not assignable to upper bound `Comparable` of type variable `_V` [bad-specialization]
- ERROR unittests/versiontests.py:315:45-65: `int` is not assignable to upper bound `Comparable` of type variable `_V` [bad-specialization]
- ERROR unittests/versiontests.py:318:45-66: `int` is not assignable to upper bound `Comparable` of type variable `_V` [bad-specialization]
- ERROR unittests/versiontests.py:319:45-67: `int` is not assignable to upper bound `Comparable` of type variable `_V` [bad-specialization]
- ERROR unittests/versiontests.py:322:49-69: `int` is not assignable to upper bound `Comparable` of type variable `_V` [bad-specialization]
- ERROR unittests/versiontests.py:327:44-64: `int` is not assignable to upper bound `Comparable` of type variable `_V` [bad-specialization]
+ ERROR unittests/versiontests.py:302:48-49: Argument `Literal[3]` is not assignable to parameter `min` with type `@_ | None` in function `mesonbuild.utils.universal.Range.__init__` [bad-argument-type]
+ ERROR unittests/versiontests.py:303:48-50: Argument `Literal[12]` is not assignable to parameter `max` with type `@_ | None` in function `mesonbuild.utils.universal.Range.__init__` [bad-argument-type]
+ ERROR unittests/versiontests.py:307:48-49: Argument `Literal[5]` is not assignable to parameter `min` with type `@_ | None` in function `mesonbuild.utils.universal.Range.__init__` [bad-argument-type]
+ ERROR unittests/versiontests.py:307:68-70: Argument `Literal[10]` is not assignable to parameter `max` with type `@_ | None` in function `mesonbuild.utils.universal.Range.__init__` [bad-argument-type]
+ ERROR unittests/versiontests.py:310:49-51: Argument `Literal[11]` is not assignable to parameter `min` with type `@_ | None` in function `mesonbuild.utils.universal.Range.__init__` [bad-argument-type]
+ ERROR unittests/versiontests.py:311:49-50: Argument `Literal[4]` is not assignable to parameter `max` with type `@_ | None` in function `mesonbuild.utils.universal.Range.__init__` [bad-argument-type]
+ ERROR unittests/versiontests.py:314:50-51: Argument `Literal[7]` is not assignable to parameter `min` with type `@_ | None` in function `mesonbuild.utils.universal.Range.__init__` [bad-argument-type]
+ ERROR unittests/versiontests.py:315:50-51: Argument `Literal[8]` is not assignable to parameter `max` with type `@_ | None` in function `mesonbuild.utils.universal.Range.__init__` [bad-argument-type]
+ ERROR unittests/versiontests.py:318:50-51: Argument `Literal[5]` is not assignable to parameter `min` with type `@_ | None` in function `mesonbuild.utils.universal.Range.__init__` [bad-argument-type]
+ ERROR unittests/versiontests.py:319:50-52: Argument `Literal[10]` is not assignable to parameter `max` with type `@_ | None` in function `mesonbuild.utils.universal.Range.__init__` [bad-argument-type]
+ ERROR unittests/versiontests.py:322:54-55: Argument `Literal[5]` is not assignable to parameter `min` with type `@_ | None` in function `mesonbuild.utils.universal.Range.__init__` [bad-argument-type]
+ ERROR unittests/versiontests.py:327:49-50: Argument `Literal[5]` is not assignable to parameter `min` with type `@_ | None` in function `mesonbuild.utils.universal.Range.__init__` [bad-argument-type]

xarray (https://github.com/pydata/xarray)
- ERROR xarray/tests/test_pandas_to_xarray.py:56:20-70: `object` is not assignable to upper bound `BaseOffset | CategoricalDtype | ExtensionDtype | Interval | Period | bool | bytes | complex | date | datetime | dtype | float | int | list[str] | str | time | timedelta | type[bool | complex | object | str]` of type variable `S1` [bad-specialization]
- ERROR xarray/tests/test_pandas_to_xarray.py:73:25-58: `object` is not assignable to upper bound `BaseOffset | CategoricalDtype | ExtensionDtype | Interval | Period | bool | bytes | complex | date | datetime | dtype | float | int | list[str] | str | time | timedelta | type[bool | complex | object | str]` of type variable `S1` [bad-specialization]

meta-codesync · 2026-10-14T22:43:20Z

This pull request has been merged in ea8421d.

meta-cla Bot added the cla signed label Oct 14, 2026

github-actions Bot assigned kinto0 Oct 14, 2026

meta-codesync Bot added the meta-exported label Oct 14, 2026

rchen152 approved these changes Oct 14, 2026

View reviewed changes

kinto0 added this to the v1.4 milestone Oct 14, 2026

github-actions Bot added the size/m label Oct 14, 2026

This comment has been minimized.

Sign in to view

meta-codesync Bot changed the title ~~Reject an overload that violates a restriction under a gradual expected type~~ Reject an overload that violates a restriction under a gradual expected type (#5240) Oct 14, 2026

kinto0 force-pushed the export-D125293977 branch from 9ba0fbb to 53faf85 Compare October 14, 2026 21:33

github-actions Bot added size/m and removed size/m labels Oct 14, 2026

kinto0 force-pushed the export-D125293977 branch from 53faf85 to 7f4c935 Compare October 14, 2026 21:36

github-actions Bot added size/m and removed size/m labels Oct 14, 2026

meta-codesync Bot closed this in ea8421d Oct 14, 2026

meta-codesync Bot added the Merged label Oct 14, 2026

kinto0 deleted the export-D125293977 branch October 15, 2026 15:16

Navigation Menu

Repository navigation

Reject an overload that violates a restriction under a gradual expected type (#5240) - #5240

Reject an overload that violates a restriction under a gradual expected type (#5240)#5240
kinto0 wants to merge 2 commits into
facebook:mainfrom
kinto0:export-D125293977

kinto0 commented Oct 14, 2026 •

edited by meta-codesync Bot

Loading

Uh oh!

meta-codesync Bot commented Oct 14, 2026

Uh oh!

rchen152 left a comment

Uh oh!

This comment has been minimized.

github-actions Bot commented Oct 14, 2026

Uh oh!

meta-codesync Bot commented Oct 14, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

kinto0 commented Oct 14, 2026 • edited by meta-codesync Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

meta-codesync Bot commented Oct 14, 2026

Uh oh!

rchen152 left a comment

Choose a reason for hiding this comment

Uh oh!

This comment has been minimized.

github-actions Bot commented Oct 14, 2026

Uh oh!

meta-codesync Bot commented Oct 14, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

kinto0 commented Oct 14, 2026 •

edited by meta-codesync Bot

Loading