Schema generations doesn't take well custom postgres types

[Fabricio-191]

It's a small issue, in my postgres database (through supabase), i have a custom type:

CREATE DOMAIN api.rating_value AS FLOAT CHECK (VALUE BETWEEN 0 AND 5);

And the generated schema column for columns with that data type turns out to be Column.text instead of Column.real (i'm using dart/flutter)

It would be great if it could detect the subtype of the custom type, if not it would be nice a warning/comment while generating the schema like: x column datatype not recognized, assigned text type

[stevensJourney]

Hi @Fabricio-191, Thanks for the feedback.

We currently default the column type to TEXT if the detected column does not fall in to a defined set of types.

I agree that your suggestions are could be valuable. We'll use this feedback to schedule investigating these improvements.

[rkistner]

FYI, I ran into the same issue when I tried to define a custom domain to allow uploading base64-encoded data to a BYTEA column via the supabase APIs. It would be good to solve this, but could also be considered a breaking change, so we may need to make this opt-in if we add support.

[rkistner]

To clarify, the primary issue is not just with the schema generation (it would be easy to just modify the schema for your app), but in the way we replicate the data - the replication process sees these as unknown types, and keep as text as we receive it from postgres, even if it should be something different such as a number, date or binary data.