Open mberdyshev opened 1 month ago
CaselIT
commented
1 month ago Hi,
Managing all attributes of enum with a type decorator is not easy, so it's likely missing something here.
If your use case I would more simply do something like
def Enum1(*args,**kwargs):
kwargs.set_default('validate_strings', True)
return Enum(*args, **kwargs)In any case it's a valid bug
mberdyshev
commented
1 month ago I've looked up the source code and found that type rendering depends on __repr__(). So, inside SQLAlchemy I managed to find that Enum and TypeDecorator representations are different. To sum it up, the difference is the following:
enum = Enum(TestEnum)
print(util.generic_repr(enum)) # TypeDecorator
# Enum('FIRST', 'SECOND')
print(util.generic_repr(enum, to_inspect=[Enum, SchemaType])) # Enum
# Enum('FIRST', 'SECOND', name='testenum')It seems now that this bug is more related to SQLAlchemy. Probably TypeDecorator's __repr__() should respect the contained type's __repr__().
CaselIT
commented
1 month ago nice finding. Seems like that may be the issue. I'm not sure why there is a reason for that repr in type decorator (there usually is).
@zzzeek maybe a type should have a generic method to format itself that type decorator could hook into?
zzzeek
commented
1 month ago But the TypeDecorator does not repr() as the inner type does, it has its ownr repr() that does something different:
from sqlalchemy import *
print(repr(Enum("a", "b", "c", name="e1")))
class MyType(TypeDecorator):
impl = Enum
print(repr(MyType("a", "b", "c", name="e2")))prints:
Enum('a', 'b', 'c', name='e1')
MyType('a', 'b', 'c')
note it uses MyType. so that's not just calling Enum.__repr__(). also no I'm not going to take the string from Enum.__repr__() and manipulate it, that's too much guessing.
so there's no quick fix here. choices include:
__repr__() into commands that are used to build out a composed reprzzzeek
commented
1 month ago this would be the most expedient fix for the moment, but it wouldn't get the extra arguments used by enum beyond those used for schema type:
diff --git a/lib/sqlalchemy/sql/sqltypes.py b/lib/sqlalchemy/sql/sqltypes.py
index bc2d898ab..6676a61e0 100644
--- a/lib/sqlalchemy/sql/sqltypes.py
+++ b/lib/sqlalchemy/sql/sqltypes.py
@@ -3826,3 +3826,4 @@ type_api.MATCHTYPE = MATCHTYPE
type_api.INDEXABLE = INDEXABLE = Indexable
type_api.TABLEVALUE = TABLEVALUE
type_api._resolve_value_to_type = _resolve_value_to_type
+type_api.SchemaType = SchemaType
diff --git a/lib/sqlalchemy/sql/type_api.py b/lib/sqlalchemy/sql/type_api.py
index 9f40905fa..f484e26db 100644
--- a/lib/sqlalchemy/sql/type_api.py
+++ b/lib/sqlalchemy/sql/type_api.py
@@ -58,6 +58,7 @@ if typing.TYPE_CHECKING:
from .sqltypes import NUMERICTYPE as NUMERICTYPE # noqa: F401
from .sqltypes import STRINGTYPE as STRINGTYPE # noqa: F401
from .sqltypes import TABLEVALUE as TABLEVALUE # noqa: F401
+ from .sqltypes import SchemaType as SchemaType # noqa: F401
from ..engine.interfaces import Dialect
from ..util.typing import GenericProtocol
@@ -2276,7 +2277,13 @@ class TypeDecorator(SchemaEventTarget, ExternalType, TypeEngine[_T]):
return self.impl_instance.sort_key_function
def __repr__(self) -> str:
- return util.generic_repr(self, to_inspect=self.impl_instance)
+ inst = self.impl_instance
+ if isinstance(inst, SchemaType):
+ to_inspect=[inst, SchemaType]
+ else:
+ to_inspect = [inst]
+
+ return util.generic_repr(self, to_inspect=to_inspect)
class Variant(TypeDecorator[_T]):
But the TypeDecorator does not repr() as the inner type does, it has its ownr repr() that does something different:
I know, that's why I suggested to have new methods for this. regarding the fix you proposed it's likely better to have a proper method for it. the effort should be similar
abhiaagarwal
commented
2 weeks ago For what it's worth, I ran into a similar issue. I have a PydanticModelType generic
class PydanticModelType(TypeDecorator, Generic[T]):
"""Generic class representing a pydantic model that can be serialized to Python."""
cache_ok = True
impl = JSON()
def __init__(self, pydantic_type: type[T]) -> None:
"""Initializes class with the type of the pydantic model."""
self.pydantic_type = pydantic_type
super().__init__()
@override
def load_dialect_impl(self, dialect: Dialect) -> TypeEngine:
# Use JSONB for PostgreSQL and JSON for other databases.
if dialect.name == "postgresql":
return dialect.type_descriptor(JSONB())
return dialect.type_descriptor(JSON())
@override
def process_bind_param(self, value: T | None, _dialect: Dialect) -> str | None:
if value is not None:
value = self.pydantic_type.model_dump_json(value)
return value
@override
def process_result_value(
self, value: str | None, _dialect: Dialect
) -> BaseModel | None:
if value is not None:
value = self.pydantic_type.model_validate_json(value)
return valueAnd when I declare it in my class
configuration: Mapped[Config] = mapped_column(
PydanticModelType(Config), nullable=False
)Alembic generates this:
sa.Column('configuration', app.database.types.PydanticModelType(), nullable=False),
Not sure if it's the same bug. I'll try overriding repr later.
Describe the bug A migration doesn't provide the
nameproperty of Enum ifsqlalchemy.Enumis augmented withsqlalchemy.types.Decorator. Even if thenameproperty is explicitly given.Expected behavior The property has to be provided in Enum's definition inside a migration.
To Reproduce
sql_types.pytables.pyalembic revision --autogenerate -m "Enums"provides (shortened):Error As you can see from the migration enums are defined differently - the second one misses its
nameproperty. If I try to run it on PostgreSQL I get the error:Stacktrace
```cmd INFO [alembic.runtime.migration] Context impl PostgresqlImpl. INFO [alembic.runtime.migration] Generating static SQL INFO [alembic.runtime.migration] Will assume transactional DDL. BEGIN; CREATE TABLE alembic_version ( version_num VARCHAR(32) NOT NULL, CONSTRAINT alembic_version_pkc PRIMARY KEY (version_num) ); INFO [alembic.runtime.migration] Running upgrade -> 264b6f57ae76, Enums -- Running upgrade -> 264b6f57ae76 CREATE TYPE testenum AS ENUM ('FIRST', 'SECOND'); Traceback (most recent call last): File "C:\Program Files\WindowsApps\PythonSoftwareFoundation.Python.3.10_3.10.3056.0_x64__qbz5n2kfra8p0\lib\runpy.py", line 196, in _run_module_as_main return _run_code(code, main_globals, None, File "C:\Program Files\WindowsApps\PythonSoftwareFoundation.Python.3.10_3.10.3056.0_x64__qbz5n2kfra8p0\lib\runpy.py", line 86, in _run_code exec(code, run_globals) File "C:\Users\Михаил\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0\LocalCache\local-packages\Python310\site-packages\alembic\__main__.py", line 4, inVersions.
Additional context The docs state that subclassing from
TypeDecoratorshould be preferred:Have a nice day!