This PR proposes we use the `@internal` JSDoc tag, and `--stripInternal`
typescript compiler option to remove internal keys from the public
`.d.ts` files.
Docs for `stripInternal`:
https://www.typescriptlang.org/tsconfig/#stripInternal
This helps ensuring internal values aren't considered as part of the
*public API surface* by users, allowing us more flexibility in
refactoring.
> [!WARNING]
> This PR proposes we add `stripInternal` to the entire repo, not just
`router-core`. However there are already some values marked as internal
that weren't added by this PR:
> - `InferStructuralSharing` in `react-router`
> - `handleHashScroll` in `router-core > scrollRestoration`
> - many `InferFoo` types in `router-core > typePrimitives`
>
> as a result of this, we might have to switch some of those `@internal`
tags to something else (like `@private` maybe?) because they are, in
fact, necessary for the build to be correct.
Example:
```ts
export type Foo = {
a: number,
/** @internal */
b: number,
}
/** @internal */
export function hello() {
return 'world'
}
export const answer = 42
```
gets compiled as `.d.ts` to
```ts
export type Foo = {
a: number
}
export const answer = 42
```
---------
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
The `start-bare` example includes Tailwind with a broken config (missing
`tailwindcss()` in vite.config.ts plugins), which contradicts the 'bare'
concept. Since `start-tailwind-v4` already provides a properly
configured Tailwind example, this PR removes Tailwind from `start-bare`
to make it actually bare.
When ordering routes to determine "which one will match in priority", we
try to follow the "more specific 1st" principle.
Currently, 2 routes with the same pattern, but with prefixes of
different lengths, are counted as having the same specificity, and thus
are not re-ordered. For example:
- `/f{$param}`
- `/foo{$param}`
In this case, assuming `f{$param}.tsx` is returned 1st by the OS when
listing route files (which it is by alphabetical order), a path
`/foobar` will match `/f{$param}`.
This PR proposes we take the `length` of the `prefix` and `suffix` into
account when sorting routes. In this example this would mean
`/foo{$param}` takes priority because it is considered more specific.
This change is a no-op, even in terms of performance. But it's more of
an "anti-slippery-slope" measure, so we don't end up with array copying
methods like `.map` or `.filter` being used instead of simpler methods
like `.find` or `.some`
`updateMatch` can trigger some heavy synchronous work (it updates the
store, which in turn, will re-run a bunch of things), so it's nice if
itself can avoid doing unnecessary work.
- it doesn't need to return anything, the return value is not used
- it doesn't need to search through the 3 arrays of matches if it found
a result in the 1st or 2nd.
---------
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
Use De Morgan's contrapositive to cleanup the return condition of the
`clearExpiredCache` filter function. This should be a no-op and is
mathematically equivalent to what was written before, just more readable
for future refactoring.