dymmond / edgy Goto Github PK

This is a short note for me:

in combination with embeddables and inheritance when having in different models different named pks I noticed a missmatch.

pkcolumns contained the wrong informations which were from a superclass.

Maybe create an helper object instead, so inheritance is guaranteed

Have a look at the fixmes in:
edgy/core/db/models/metaclasses.py of my PR feature/composite_field
and run hatch run test:test tests/fields to see the crashes when the workaround is removed

The fields are carried over when dropping models (or maybe the fields should be remembered and only the class partly rewired). Somehow even the field attribute is remembered. This is not good.

Discussed in #98

db, registry = get_db_connection()

class World(Model):
    id = fields.IntegerField(primary_key=True)
    name = fields.CharField(max_length=255)

    class Meta:
        registry = registry

class Hello(Model):
    id = fields.IntegerField(primary_key=True)
    name = fields.CharField(max_length=255)
    worlds = fields.ManyToManyField(World, related_name="hellos")

    class Meta:
        registry = registry

world = await World.query.create(name='Hearth')
hello = await Hello.query.create(name='King')
await hello.worlds.add(world)                             # It works correctly

h = await Hello.query.get(id=1)
w = await World.query.get(id=1)
await h.worlds.remove(w)                                   # AttributeError: 'NoneType' object has no attribute 'remove'

We should check properly if a value in __dict__ is actually a cached property.

Currently there is no problem, because related_fields are the only users which needs to be escaped. But for future we have to rethink this. We may could take the column names and make an inclusive set of db_field_columns

We could also reduce the usage of cached_properties by redirecting calls to meta (needs some harmonization of attributes, which will be a breaking change)

Please tell me, how can I switch to another postgres schema when doing queries or making migrations?
In sqlalchemy we can do

async def get_session(schema_name):
    async with async_session() as session:
        await session.execute(f"set search_path to {schema_name}")
        await session.commit()
        yield session

And then use session in api.
How to do that with edgy or saffier?

When i want to create custom manager it gives me an error RecursionError: maximum recursion depth exceeded.

When I try makemigrations I also get Error: maximum recursion depth exceeded error or I import the model that include this manager into esmerald shell I get same error.

Edgy version: 0.4.0
Python version: 3.10.12
OS: Ubuntu 22.04
Platform: vs code

Is it possible to generate models from existing database without writing all models myself? Thanks