feat(query): type-referenced invalidation — .exact(TodoList()) over raw keys

[zzal]

Part II Wave 1 · #3 — invalidation restates keys the query type already owns

Invalidation was only .prefix(QueryKey) / .exact(QueryKey) / .tag(QueryTag) over raw keys. Inside one mutation the same query gets referenced type-safely in optimistic and stringly one line later:

func optimistic(_ title: String) -> [OptimisticEdit] {
    [.update(TodoList()) { $0 + [newTodo] }]      // typed
}
func invalidations(input: String, output: Todo) -> [Invalidation] {
    [.exact(["todos"])]                            // stringly — restates @Query(prefix: "todos")'s key
}

A renamed prefix silently breaks the invalidation; the compiler stays quiet.

Change

@MainActor static overloads of the case names taking the typed owner of the key, resolving eagerly to the existing cases:

static func exact<Q: Query>(_ query: Q) -> Invalidation   // .exact(query.queryKey)
static func prefix<Q: Query>(_ query: Q) -> Invalidation  // .prefix(query.queryKey)

• No new vocabulary — the same words now accept the typed source of the key, so invalidations rhymes with optimistic: [.exact(TodoList())].
• No ambiguity — an array literal can't be a Query; a test pins that raw-key spellings keep resolving to the cases.
• Raw keys remain the escape hatch for keys no Query type owns.
• @MainActor matches OptimisticEdit.update (the whole Query protocol is isolated).

Dogfood — all 4 stringly sites in the repo

• TodoCRUD ×3: [.exact(["todos"])] → [.exact(TodoList())]
• QueryDemo ×1 (parameterized): [.exact(["users", .int(id)])] → [.exact(UserByID(id: id))] — the macro (@Query(prefix:) + @Key) owns the key end-to-end
• EmbeddedTemplates.swift regenerated (both examples are embedded templates)

Tests

New Invalidation/type-referenced suite (3): typed spellings produce values identical to the raw-key spellings — so the fully-tested dispatch machinery behind the cases is provably shared — plus exact≠prefix discrimination and the no-overload-capture pin.

Verification

• Full host suite: 1379 tests, 277 suites green (incl. the TemplateEmbedder byte-pin on the regen)
• TodoCRUD + QueryDemo wasm builds clean
• Release CLI rebuilt with the new templates, then TodoCRUD Playwright e2e 5/5 against the real Bun backend (freshly scaffolded — optimistic-add-reconcile + rollback exercise the changed path)

🤖 Generated with Claude Code

[github-actions]

📦 Bundle size

Artifact	Baseline	This PR	Δ
App.wasm	5.06 MB	5.15 MB	+1.91%
App.wasm (gzip)	1.80 MB	1.84 MB	+2.20%
JS runtime	54.5 KB	54.5 KB	+0.00%
JS runtime (gzip)	11.3 KB	11.3 KB	+0.00%
Total (gzip)	1.81 MB	1.85 MB	+2.19%

✅ Within budget (≤5% growth allowed).

_{Baseline: Swift 6.3, WASM SDK 6.3-RELEASE, measured 2026-06-18.}