OpenAPI/Pydantic Validation Support

[mplemay]

Is your feature request related to a problem? Please describe.
Many backends for svelte apps are written in python using FastAPI and Pydantic. When a project uses this stack, Sveltekit Superforms is not able to natively display the validation errors returned by this backend.

Describe the solution you'd like
While pydantic is not a javascript validation library, it does do validation (and return validation errors) and a schema is available through the OpenAPI Specification. As such, I was wondering if it would be possible to parse and show the validation errors for form fields returned by pydantic.

Describe alternatives you've considered
Currently, I either replicate the schema I originally defined in Pydantic using zod or I don't use Sveltekit Superforms for validation.

[ciscoheat]

Do you use Superforms in SPA mode for this? (As it's not a Svelte library, but SvelteKit)

[mplemay]

@ciscoheat Sorry for the delayed response - I have been out traveling (and away from my computer).

To answer your question, yes, I have been using Superforms in SPA mode. In the +page.ts I do something very similar to this example. Currently, I used a zod model for validation, but ideally I would love to remove all of the validation from zod and return the messages sent from API endpoint; thus effectively treating the endpoint as a validator of sorts.

For instance, if we were to setup a FastAPI (i.e. OpenAPI) server to handle the example above, we would have a schema struct (automatically generated by https://openapi-ts.pages.dev/ or in general) that looks like:

interface UserSchema {
    id: integer
    name: string
    email: string
}

When we called the API, it would return a json object with validation errors which at least in the case of fastapi looks something like this:

{ 
    "detail":[ 
        { 
            "loc":[ 
                "name"
            ],
            "msg":"field required",
            "type":"value_error.missing"
        },
        { 
            "loc":[ 
                "id"
            ],
            "msg": "value is not a valid integer",
            "type": "type_error.float"
        }
    ]
}

Given this paradigm, where the "schema" is separated from the "validator", how would one go about integrating this into superforms (if this is a workflow that superforms supports)?

[ciscoheat]

I'm thinking that you can use the onUpdate event, call the API there and loop through its validation errors with setError, similar to https://superforms.rocks/concepts/spa#displaying-the-form