Improve EndpointCreator flexibility by adding methods to be overriden #80
Conversation
JakNowy
force-pushed
the
feat/endpoint-creator-overrides
branch
2 times, most recently
from
6d5de93
to
3a78972
Compare
| def _create_item(self): | ||
| """Creates an endpoint for creating items in the database.""" | ||
|
|
||
| async def endpoint( | ||
| db: AsyncSession = Depends(self.session), | ||
| item: self.create_schema = Body(...), # type: ignore | ||
| ): | ||
| unique_columns = _extract_unique_columns(self.model) |
There was a problem hiding this comment.
I'm not a fan of this exists check - it generates a number of database calls. In fact, I just tested that every "unique" field in SQLAlchemy introduces an index on that field, so the responses are immediate. Anyway, I still find it sufficient to just handle generic IntegrityError exception as presented.
| @@ -249,9 +269,13 @@ def _read_item(self): | |||
|
|
|||
| @apply_model_pk(**self._primary_keys_types) | |||
| async def endpoint(db: AsyncSession = Depends(self.session), **pkeys): | |||
| item = await self.crud.get(db, **pkeys) | |||
| if stmt := self._base_read_statement(): | |||
There was a problem hiding this comment.
This way, while we're trying to make FastCRUD as useful as possible, we're not forcing anyone to use either FastCRUD or EndpointCreator. The latter may be used with raw SQLAlchemy querries and the former can be used in your usual single FastAPI endpoints. And of course you will probably get most by combining the 2 features, but you're never going into a situation where you're limited by the library with something you could achieve with raw SQLAlchemy.
We could do the same with other methods, especially to easily implement complex joins and filters.
JakNowy
force-pushed
the
feat/endpoint-creator-overrides
branch
from
3a78972
to
d4a34df
Compare
This is general demonstration of my concept of EndpointCreator changes discussed briefly in #79 (comment)
It hopefully helps to draw some general concept of FastCRUD-EndpointCreator responsibilities.