Table with name <NAME> does not exist for ReflectModel?

[golgor]

Hey,
First of all, great work with edgy. It seems to be very intuitive and it works great in VS Code with code completions etc.

I have a legacy database I want to use for data analysis and I used the edgy inspectdb and it crashed with an error:

$ edgy inspectdb --database "mysql://<USER>:<PASS>@localhost:62999/api" > models.py
2024-03-05 22:26:15.400 | INFO     | edgy.utils.inspect:reflect:234 - Collecting database tables information...
Exception ignored in: <function Connection.__del__ at 0x7fdce7771080>
Traceback (most recent call last):
  File "/home/golgor/Code/toolsense-flespi/.venv/lib/python3.12/site-packages/aiomysql/connection.py", line 1131, in __del__
    self.close()
  File "/home/golgor/Code/toolsense-flespi/.venv/lib/python3.12/site-packages/aiomysql/connection.py", line 339, in close
    self._writer.transport.close()
  File "/home/golgor/.rye/py/[email protected]/install/lib/python3.12/asyncio/selector_events.py", line 1206, in close
    super().close()
  File "/home/golgor/.rye/py/[email protected]/install/lib/python3.12/asyncio/selector_events.py", line 871, in close
    self._loop.call_soon(self._call_connection_lost, None)
  File "/home/golgor/.rye/py/[email protected]/install/lib/python3.12/asyncio/base_events.py", line 792, in call_soon
    self._check_closed()
  File "/home/golgor/.rye/py/[email protected]/install/lib/python3.12/asyncio/base_events.py", line 539, in _check_closed
    raise RuntimeError('Event loop is closed')
RuntimeError: Event loop is closed

Though, it did create the models.py and all of the models where generated, so I let that go. I tried to implement some code, but kept running into a weird issue. I have the following code:

import asyncio

import edgy


database = edgy.Database("mysql+aiomysql://<USERNAME>:<PASSWORD>@127.0.0.1:62999/api")
api_database = edgy.Registry(database=database)

class Assetdata(edgy.ReflectModel):
    id = edgy.IntegerField(null=False, primary_key=True)
    data = edgy.JSONField(null=True)
    imei = edgy.CharField(max_length=16, collation="utf8mb4_unicode_ci", null=True)

    class Meta:
        registry = api_database
        tablename = "AssetData"


async def main() -> None:
    try:
        await database.connect()
        result = await Assetdata.query.get(id=12079021)
        print(result)
    except Exception as e:
        print(e)
    finally:
        await database.disconnect()


if __name__ == "__main__":
    asyncio.run(main())

Running this code I keep getting an error that says Table with the name AssetData does not exist. and I really couldn't wrap my head around what was the issue. In tried to copy paste and modify one of the code examples and and I got that working with a SQLITE-database and some other models from the example.

After a bit of migrating the code example to my other database, I noticed that one of the differences for me was that I used a ReflectModel and not a Model. I tried to change this, and VOILA! It worked right away.

I tried to read up on ReflectModel, but from what I understand it should pretty much be the exact same thing as Model. The only difference is that ReflectModel won't create any migrations, is that right?

So is there any additional configuration required for ReflectModel that is different from a regular Model, or what can be the issue in this case?

I also got some typing issues with Mypy with a query written as: await AssetData.query.all().limit(10) saying Missing positional argument "self" in call to "all" of "QuerySet"Mypy(call-arg)

Just for some unresponsible testing I added:

    def __call__(self, *args: Any, **kwds: Any) -> Any:
        return self

in the Manager-class (edgy/core/db/models/). If I now use await AssetData.query().all().limit(10) I don't get any typing issues, and the completions works much better. There are probably ways to make that better, just figured I could mention it.

[tarsil]

Hi @golgor so first, thank you so much for using edgy and for the feedback. I do really hope you enjoy.

In a nutshell you pretty much nailed the differences between ReflectModel and Model. Although what in trying to understand is in the reflect model in theory if you pass the exact table name of the database you are connecting, in theory it should recognise the table.

The inspectdb uses the reflect model for safety reasons, to make sure no migrations are generated.

This can be indeed a bug that needs addressing but I've been using the reflect model with existing databases where I just map fields and the exact name of the table in the dB that I'm connecting and it works.

Maybe a specific use case could be tested and we could address if is indeed an issue.

[golgor]

As mentioned, the only thing I change in the code above is edgy.ReflectModel -> edgy.Model, nothing else. The table name is exactly like it is in the database. I might also add that the table itself contains much more columns than I have specified in the model as I'm only interested in a few of them. I don't know if that could be the issue? That it somehow doesn't find a table that fits the signature of the model?

Is there any way I can troubleshoot or provide more information that could help narrow down the issue?

[tarsil]

Hmm not sure actually because it should work like the model but without the migrations. Although this indeed needs looking and see what can be done.