带有 SQLAlchemy 的 Postgres 中的枚举数组

Question

I've been using an array of enums with postgres and SQLAlchemy successfully over the past year like so:

class MyModel(BaseModel):
    enum_field = Column(postgresql.ARRAY(EnumField(MyEnum, native_enum=False)))

The EnumField is from the sqlalchemy_enum34 library, a small wrapper around the builtin enum that uses Python enums as Python representation instead of strings.

Although the docs say , array of enum is not supported, I guess it worked, because I chose 'native_enum=False'. Recently I noticed that it doesn't work anymore, I think it's due to the upgrade from SQLA 1.0 to 1.1, but I'm not sure.

The problem is, that it generates invalid DQL:

CREATE TABLE my_model (
    enum_field VARCHAR(5)[3] NOT NULL CHECK (contexts IN ('ONE', 'TWO', 'THREE'))
)

The error I get is:

ERROR:  malformed array literal: "ONE"
DETAIL:  Array value must start with "{" or dimension information.

Any idea how I can get back my enum array?
By the way: when it worked, no CHECK constraint was actually created, just an array of varying. I'm ok with that as long as I can use enums in my Python code (eg query.filter(enum_field==MyEnum.ONE) )

Answer 1

I found nice workaround in SqlAlchemy source code:

import re

from sqlalchemy import TypeDecorator, cast
from sqlalchemy.dialects.postgresql import ARRAY


class ArrayOfEnum(TypeDecorator):

    impl = ARRAY

    def bind_expression(self, bindvalue):
        return cast(bindvalue, self)

    def result_processor(self, dialect, coltype):
        super_rp = super(ArrayOfEnum, self).result_processor(dialect, coltype)

        def handle_raw_string(value):
            inner = re.match(r"^{(.*)}$", value).group(1)

            return inner.split(",") if inner else []

        def process(value):
            if value is None:
                return None

            return super_rp(handle_raw_string(value))

        return process

And now:

achievements = Column(ArrayOfEnum(Enum(AchievementsType)))

And then:

career.achievements = [AchievementsType.world, AchievementsType.local]

Answer 2

When I needed an array of enums I used the recipe from Mike Bayer here: https://bitbucket.org/zzzeek/sqlalchemy/issues/3467/array-of-enums-does-not-allow-assigning#comment-19370832

EDIT: Issue moved to https://github.com/sqlalchemy/sqlalchemy/issues/3467

This is, create a custom type like this:

import sqlalchemy as sa

class ArrayOfEnum(ARRAY):

    def bind_expression(self, bindvalue):
        return sa.cast(bindvalue, self)

    def result_processor(self, dialect, coltype):
        super_rp = super(ArrayOfEnum, self).result_processor(dialect, coltype)

        def handle_raw_string(value):
            inner = re.match(r"^{(.*)}$", value).group(1)
            return inner.split(",")

        def process(value):
            return super_rp(handle_raw_string(value))
        return process

I haven't used this for a while so I'm not certain that it continues to work.

It's not the same code as your enum34 library so maybe it won't have the same problems?

Answer 3

In modern SqlAlchemy, you don't have to define a custom type for this:

import sqlalchemy.dialects.postgresql as pg

class MyModel(Base):
    ...
    flags = Column(pg.ARRAY(sa.Enum(MyEnum, 
                   create_constraint=False, native_enum=False)))

Answer 4

Mike Bayer answered on the sqlalchemy mailing list :

you probably want to add create_constraint=False, see if that works

http://docs.sqlalchemy.org/en/latest/core/type_basics.html?highlight=enum#sqlalchemy.types.Enum.params.create_constraint

I can now create the table (without any CHECK).

Answer 5

If you find your way here, updating SQLAlchemy to >=1.3.17 should sort you out.

See the release notes at: https://docs.sqlalchemy.org/en/13/changelog/changelog_13.html#change-e57f5913ab592a9c044cad747636edd8