[Enhancement]: Passing data through form fields changes their requiredness

[acidjunk]

Currently on the python side we misuse a fields default value: for injecting the data on Form that are used to edit stuff.
This has a side effect: all form fields that have data, have a default -> so are not required anymore.

Currently we work around this with extra validators, which is error prone and leads to code duplication in cases where you have a Create and Edit Form.

Solution direction:

1. it would be nice to embed the form data in the JSON scheme
2. check if it possible to use the "uniforms" as a nested dict in the form field property: to provide the data (see screen, which shows an extra "value" key.)
3. if 2 is not possible: try to return the data inside the JSON scheme on another location: frontend will need changes to populate the fields

Questions:

• what to do with edit form when no data can be found for a field; but a default does exist on the form definition?

[acidjunk]

I fiddled a bit with some forms. Best way forward seems to be to delegate form data loading to the JS part, instead of polluting our pydantic models with defaults that hold the value.

JS part (example)

<AutoForm
    schema={bridge}
    onSubmit={this.submit}
    showInlineError={true}
    validate="onSubmit"
    model={{"str_field": "Data from backend", "number_field": 3, ...userInput}}
>

This works for all fields, except for "str": when you edit a pre-poulated string field: uniforms submits: "" -> which pydantic considers a valid string. When a required field isn't populated at all with data: a submit doesn't have this field in the payload: and that triggers a required validation.

Ugly backend workaround for the textfield

Define the field like this:

str_field: str = Field(..., min_length=1)

Probably good to convert all "" to None in the backend: pydantic/pydantic#2687

[Mark90]

@DutchBen this relates to what we discussed the other day. Could you share the conclusion or link the related UI ticket?

[Mark90]

Might be worth adding documentation for this.

[Mark90]

Maybe we can implement the correct interpretation of "required" in pydantic-forms by inspecting if the json schema has a default which is not nullable.

This would give the following possibilities:

• Field has no default -> required
• Field has a default with type str -> required
• Field has a default with type str | None -> optional

Let's discuss during the code sprint.

[hanstrompert]

Hi @Mark90, I added this the upcoming sprint, in the expectation that you will finish this issue. If not, please let me know.