feat: add support for SQLAlchemy dataclass-mapped tables

[barsa-net]

Resolves #517

This hopefully add support for SQLAlchemy dataclass-mapped tables without breaking anything else

The following code should run without errors:

from typing import Optional

import sqlalchemy as sa
from sqlalchemy import String, create_engine, JSON
from sqlalchemy.orm import DeclarativeBase, MappedAsDataclass, Session
from sqlalchemy.orm import Mapped
from sqlalchemy.orm import mapped_column

from serde import serde, from_dict
from serde.json import to_json


class Base(MappedAsDataclass, DeclarativeBase):
    pass


@serde
class User(Base):
    __tablename__ = "user"

    id: Mapped[int] = mapped_column(primary_key=True)
    name: Mapped[str]
    fullname: Mapped[str] = mapped_column(String(30))
    nickname: Mapped[Optional[str]]
    meta: Mapped[Optional[JSON]] = mapped_column(type_=JSON)


engine = create_engine("sqlite:///:memory:")
Base.metadata.create_all(engine)

user = from_dict(User, {"id": 1, "name": "john", "fullname": "John Doe"})
with Session(engine) as session:
    session.add(user)
    session.commit()

    user = session.scalar(sa.select(User))

to_json(user)

[codecov]

Codecov Report

Attention: Patch coverage is 85.00000% with 3 lines in your changes are missing coverage. Please review.

Project coverage is 91.34%. Comparing base (fb93ed7) to head (1a9f937).
Report is 13 commits behind head on main.

Files	Patch %	Lines
serde/sqlalchemy.py	76.92%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #518      +/-   ##
==========================================
- Coverage   91.42%   91.34%   -0.08%     
==========================================
  Files          12       13       +1     
  Lines        1901     1919      +18     
  Branches      401      402       +1     
==========================================
+ Hits         1738     1753      +15     
- Misses        107      110       +3     
  Partials       56       56              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[barsa-net]

I added only some simple tests at the moment, but reaching codecov target is unfeasible without bloat testing everything after except ImportError:
@yukinarit may I ask for some advices on what other cases than simples one should be covered by tests?

[yukinarit]

Hi @barsa-net

Thank you for contributing to pyserde! 🙏

Please do the followings and the PR can be merged.

• Run test without sqlalchemy (See the comments in test_sqlalchemy.py)
• Add simple example (See 👥 Add @JWSong as a contributor #522)
• Add documentation in https://github.com/yukinarit/pyserde/blob/main/docs/en/types.md

[yukinarit]

Can you run tests without sqlalchemy?

You can manually uninstall sqlalchemy and run tests here IMO
https://github.com/yukinarit/pyserde/blob/main/.github/workflows/test.yml#L32-L38

[barsa-net]

Sure can do, but I have a unrelated doubt regarding https://github.com/yukinarit/pyserde/blob/main/.github/workflows/test.yml#L34
All tests are being run through poetry run ... while orjson is being installed (seemingly) outside the virtual enviroment with pip install orjson, it should be poetry install --extras orjson or do I miss something?

[yukinarit]

Good catch. This line had been added two years ago by this commit but apparently pyserde was never tested with orjson 🤣

[barsa-net]

Just to make it clear and avoiding misunderstanding about the scope of this PR, SQLAlchemy is a HUGE project with countless ways to declare tables, at the moment I plan to add only "basic" support to declarative dataclass mapping.

As such, I suggest it should be annotated as an "experimental" feature, as some features at the moment don't work, e.g.:

• @serde(type_check=strict) does nothing, no errors on wrong types, so I guess no type checking is being performed at Beartype level.
• As SQLAlchemy want fields "annotated" with its mapped_column() function I'm unsure if it's even possible to make serde.core.field() and sqlalchemy.orm.mapped_column() working along each other, that means no field attributes.

[yukinarit]

LGTM!

[yukinarit]

@barsa-net is it ready for merge, or you would be adding some more changes?

[barsa-net]

Sorry, I have been busy those days.
I want to add a notice in Types docs about that is not feature complete (refer to my previous comment)
Also I'm trying something else but I don't know if it works at this time, I'll get in touch by the end of the week.

[barsa-net]

Sorry for the delay, I added an additional test for nested SQLAlchemy models and I made clear in the documentation that the feature has to be considered experimental and as such some features could not work.

I consider the PR "complete", if I manage to get some other feature working together with SQLAlchemy I'll eventually open another PR.