Pydantic `GenerateSchema.match_type` does not work for

The Fix

pip install pydantic==2.9.1

Based on closed pydantic/pydantic issue #8783 · PR/commit linked

Jump to Verify Open PR/Commit

@@ -534,7 +534,6 @@ Specifically, the following config options are relevant:
 * [`title`][pydantic.config.ConfigDict.title]
 * [`json_schema_extra`][pydantic.config.ConfigDict.json_schema_extra]
-* [`schema_generator`][pydantic.config.ConfigDict.schema_generator]
 * [`json_schema_mode_override`][pydantic.config.ConfigDict.json_schema_mode_override]
 * [`field_title_generator`][pydantic.config.ConfigDict.field_title_generator]

repro.py

from datetime import date, datetime
from functools import partial
from typing import Any, cast, override, TYPE_CHECKING

from pydantic import ConfigDict, GenerateSchema, TypeAdapter
from pydantic.dataclasses import dataclass
from pydantic_core import core_schema
from pydantic._internal._std_types_schema import InnerSchemaValidator

class ApiSchema(GenerateSchema):
    @override
    def match_type(obj: object):
       print(obj)
       return super().match_type(obj)

@api_dataclass
class DateTimeOrNone:
    none: None  # works
    date_none: date | None  # only works for date
    datetime_none: datetime | None  # only works for datetime
    time_none: time | None  # only works for time
    fail_date: date  # doesn't works
    fail_datetime: datetime  # doesn't works
    fail_time: time  # doesn't works



# Example of overriding '_get_prepare_pydantic_annotations_for_known_type' and '_union_schema' (separate file, use same imports)

def test_date_or_None():
    test_none_dict: dict[str, str] = {"dn": ""}
    assert None is TypeAdapter(DateTimeOrNone).validate_python(test_none_dict).dn


class ApiSchema(GenerateSchema):
    @override
    def _get_prepare_pydantic_annotations_for_known_type(
        self, obj: Any, annotations: tuple[Any, ...]
    ) -> tuple[Any, list[Any]] | None:
        result = None
        if obj is date:
            result = core_schema.no_info_plain_validator_function(
                lambda x: datetime.strptime(x, "%Y%m%d")
            )
        if result is not None:
            return obj, [InnerSchemaValidator(result)]
        return cast(
            tuple[object, list[object]] | None,
            super()._get_prepare_pydantic_annotations_for_known_type(obj, annotations),
        )

    @override
    def _union_schema(self, union_type: Any) -> core_schema.CoreSchema:
        schema = super()._union_schema(union_type)
        if schema["type"] == "nullable":
            schema["schema"] = core_schema.union_schema([
                schema["schema"],
                core_schema.no_info_before_validator_function(
                    lambda x: None if x == "" else x, core_schema.none_schema()
                ),
            ])
        return schema

@dataclass(config=ConfigDict(schema_generator=ApiSchema))
class DateTimeOrNone:
    dn: date | None = None

verify

Re-run the minimal reproduction on your broken version, then apply the fix and re-run.

fix.md

Option A — Upgrade to fixed release\npip install pydantic==2.9.1\nWhen NOT to use: This fix should not be used if the application relies on the previous behavior of `match_type`.\n\n

Why This Fix Works in Production

Trigger: `GenerateSchema.match_type` does not work for union types or `datetime` types
Mechanism: The `match_type` method in `GenerateSchema` does not evaluate union types and datetime types correctly
Why the fix works: Addresses issues with the `GenerateSchema.match_type` method not working for union and datetime types, marking it as experimental in the documentation. (first fixed release: 2.9.1).

Why This Breaks in Prod

The `match_type` method in `GenerateSchema` does not evaluate union types and datetime types correctly
Production symptom (often without a traceback): `GenerateSchema.match_type` does not work for union types or `datetime` types

Proof / Evidence

GitHub issue: #8783
Fix PR: https://github.com/pydantic/pydantic/pull/10303
First fixed release: 2.9.1
Reproduced locally: No (not executed)
Last verified: 2026-02-09
Confidence: 0.85
Did this fix it?: Yes (upstream fix exists)
Own content ratio: 0.39

Discussion

High-signal excerpts from the issue thread (symptoms, repros, edge-cases).

Jump to Sources Open on GitHub

“@sydney-runkle additionally, None does not work correctly.”

@KotlinIsland · 2024-03-18 · source

“See https://github.com/pydantic/pydantic/pull/10303 - I don't think we plan on supporting GenerateSchema in this way in the short term, so going to close this as not…”

@sydney-runkle · 2024-09-05 · source

“Hi @sydney-runkle”

Issue thread · 2024-02-21 · repro detail · source

“Hi @Loch64, Thanks for raising this issue”

@sydney-runkle · 2024-02-20 · source

Failure Signature (Search String)

`GenerateSchema.match_type` does not work for union types or `datetime` types
`match_type` does not completely evaluate the whole union on a class that inherits `GenerateSchema`.

Copy-friendly signature

signature.txt

Failure Signature
-----------------
`GenerateSchema.match_type` does not work for union types or `datetime` types
`match_type` does not completely evaluate the whole union on a class that inherits `GenerateSchema`.

Error Message

Signature-only (no traceback captured)

error.txt

Error Message
-------------
`GenerateSchema.match_type` does not work for union types or `datetime` types
`match_type` does not completely evaluate the whole union on a class that inherits `GenerateSchema`.

Minimal Reproduction

repro.py

from datetime import date, datetime
from functools import partial
from typing import Any, cast, override, TYPE_CHECKING

from pydantic import ConfigDict, GenerateSchema, TypeAdapter
from pydantic.dataclasses import dataclass
from pydantic_core import core_schema
from pydantic._internal._std_types_schema import InnerSchemaValidator

class ApiSchema(GenerateSchema):
    @override
    def match_type(obj: object):
       print(obj)
       return super().match_type(obj)

@api_dataclass
class DateTimeOrNone:
    none: None  # works
    date_none: date | None  # only works for date
    datetime_none: datetime | None  # only works for datetime
    time_none: time | None  # only works for time
    fail_date: date  # doesn't works
    fail_datetime: datetime  # doesn't works
    fail_time: time  # doesn't works



# Example of overriding '_get_prepare_pydantic_annotations_for_known_type' and '_union_schema' (separate file, use same imports)

def test_date_or_None():
    test_none_dict: dict[str, str] = {"dn": ""}
    assert None is TypeAdapter(DateTimeOrNone).validate_python(test_none_dict).dn


class ApiSchema(GenerateSchema):
    @override
    def _get_prepare_pydantic_annotations_for_known_type(
        self, obj: Any, annotations: tuple[Any, ...]
    ) -> tuple[Any, list[Any]] | None:
        result = None
        if obj is date:
            result = core_schema.no_info_plain_validator_function(
                lambda x: datetime.strptime(x, "%Y%m%d")
            )
        if result is not None:
            return obj, [InnerSchemaValidator(result)]
        return cast(
            tuple[object, list[object]] | None,
            super()._get_prepare_pydantic_annotations_for_known_type(obj, annotations),
        )

    @override
    def _union_schema(self, union_type: Any) -> core_schema.CoreSchema:
        schema = super()._union_schema(union_type)
        if schema["type"] == "nullable":
            schema["schema"] = core_schema.union_schema([
                schema["schema"],
                core_schema.no_info_before_validator_function(
                    lambda x: None if x == "" else x, core_schema.none_schema()
                ),
            ])
        return schema

@dataclass(config=ConfigDict(schema_generator=ApiSchema))
class DateTimeOrNone:
    dn: date | None = None

Environment

Pydantic: 2

What Broke

Users experience incorrect schema generation for union and datetime types, leading to validation errors.

Why It Broke

The `match_type` method in `GenerateSchema` does not evaluate union types and datetime types correctly

Fix Options (Details)

Option A — Upgrade to fixed release Safe default (recommended)

pip install pydantic==2.9.1

When NOT to use: This fix should not be used if the application relies on the previous behavior of `match_type`.

Use when you can deploy the upstream fix. It is usually lower-risk than long-lived workarounds.

Fix reference: https://github.com/pydantic/pydantic/pull/10303

First fixed release: 2.9.1

Last verified: 2026-02-09. Validate in your environment.

When NOT to Use This Fix

This fix should not be used if the application relies on the previous behavior of `match_type`.

Verify Fix

verify

Re-run the minimal reproduction on your broken version, then apply the fix and re-run.

Did This Fix Work in Your Case?

Quick signal helps us prioritize which fixes to verify and improve.

Prevention

Add a CI check that diffs key outputs after upgrades (OpenAPI schema snapshots, JSON payload shapes, CLI output).
Upgrade behind a canary and run integration tests against the canary before 100% rollout.

Version Compatibility Table

Version	Status
2.9.1	Fixed

Related Issues

No related fixes found.

Cluster: pydantic:openapi-schema Pydantic hub Pydantic best practices All hubs All clusters

Related clusters: Configuration error Data consistency Serialization

Sources

We don’t republish the full GitHub discussion text. Use the links above for context.

Pydantic `GenerateSchema.match_type` does not work for union types or `datetime` types (Fix)