Support for overriding types

[oliverlambson]

It would be really helpful to have the override functionality available in the go version of sqlc. In combination with the existing pydantic functionality, this would be very useful.

e.g., currently JSONB columns are returned with Any type. I would like to be able to override this behaviour on a per-column basis. This would be useful not only for "Any" types, but also for types where a different primitive is desired (e.g. a 3rd party UUID instead of the builtin, like the go example)

Given this schema

-- schema.sql
CREATE TABLE some_table (
    id SERIAL NOT NULL,
    payload JSONB NOT NULL
);

Current output of sqlc generate (uses Any for payload column)

# db/models.py
# Code generated by sqlc. DO NOT EDIT.
# versions:
#   sqlc v1.28.0
import pydantic
from typing import Any


class SomeTable(pydantic.BaseModel):
    id: int
    payload: Any

Desired output (uses custom type for payload column)

# db/models.py
# Code generated by sqlc. DO NOT EDIT.
# versions:
#   sqlc v1.28.0
import pydantic
import my_lib.models


class SomeTable(pydantic.BaseModel):
    id: int
    payload: my_lib.models.Payload

How it could work

I would imagine something in the config like:

version: "2"
plugins:
  - name: py
    wasm:
      url: https://downloads.sqlc.dev/plugin/sqlc-gen-python_1.3.0.wasm
      sha256: fbedae96b5ecae2380a70fb5b925fd4bff58a6cfb1f3140375d098fbab7b3a3c
sql:
  - schema: "./schemq.sql"
    queries: "./query"
    engine: postgresql
    codegen:
      - out: ./packages/core/src/core/db
        plugin: py
        options:
          package: core.db
          emit_pydantic_models: true
          overrides:  # similar to the go overrides
            - column: "some_table.payload"
              py_import: "my_lib.models"
              py_type: "Payload"

[thezachdrake]

This would also enable working with vectors and geoms as requested in #61 and #62. Can this be merged soon?

[thezachdrake]

@kyleconroy Is the PR on this something that can be rolled out into a release soon? Once its out I would like to start experimenting with making sure we can use it for spatial and vector types as mentioned and updated the README with examples of that working.