agama-project / agama / 26279477785

81%

push

github

refactor(web): standardize form organization (#3521)

## Summary

Standardizes all TanStack Form implementations to consistent naming
conventions and directory structure, consolidating form concerns (types,
defaults, validation) into unified `fields.ts` modules.

**Affected forms**:

- **connection-form**: Moved into dedicated subdirectory, renamed field
group components, consolidated validation
- **system-form**: Moved into dedicated subdirectory, consolidated
validation
- **registration-form**: Moved into dedicated subdirectory, extracted
inline validation
- **patterns-form**: Moved into dedicated subdirectory

## Background

This branch initially explored [Valibot](https://valibot.dev/), a
[Standard Schema](https://standardschema.dev/) validation library
(~1.5kb), for declarative schema-based validation in the
connection-form:

```typescript
import * as v from "valibot";

const ipFieldsSchema = v.object({
  addresses4: v.pipe(
    v.array(v.string()),
    v.check((arr) => arr.every(isValidIPv4)),
  ),
  // ...
});
```

After implementation (commit
https://github.com/agama-project/agama/pull/3521/commits/8f89746d5)
it was found that the abstraction added more overhead than value.
Valibot's API is nice, but introduced a learning curve, and imperative
validation functions were clearer for the complex field
interdependencies present in these forms.

The approach was rolled back (commit
https://github.com/agama-project/agama/pull/3521/commits/d594ee7f5)
in favor of plain TypeScript with reusable helpers. The full reasoning
is documented in
[`conventions.md`](https://github.com/agama-project/agama/blob/f558792a1/web/src/components/form/conventions.md).

## Motivation

After the Valibot rollback, the codebase settled on a clear pattern but
it hadn't been applied consistently across all forms:

- Connection-form files were flat inside `network/` with no subdirector... (continued)

4199 of 6288 branches covered (66.78%)

Branch coverage included in aggregate %.

130 of 133 new or added lines in 21 files covered. (97.74%)

23551 of 27996 relevant lines covered (84.12%)

99.04 hits per line