test: consolidate duplicated live-query test harness

[mbret]

Consolidation: live-query test harness (net −162 LOC)

What was duplicated

• src/lib/queries/useQuery$.reactivity.test.tsx (6 copies)
• src/lib/queries/useQuery$.unmount.test.tsx (4 copies)

Every test in these two suites rebuilt the exact same scaffold by hand:

const liveQuery$ = new BehaviorSubject([...])
const db$ = new BehaviorSubject<object | undefined>({})
const queryClient = createQueryClient()

function Comp() {
  const { data } = useQuery$({
    ...liveQueryOptions,
    queryKey: [...],
    queryFn: () =>
      db$.pipe(filter(isDefined), switchMap(() => liveQuery$), map((items) => [...items])),
  })
  return <span data-testid="data">{JSON.stringify(data)}</span>
}

render(<Comp />, { wrapper: createWrapper(queryClient) })

plus, in the unmount suite, the same Host/toggle visibility wrapper repeated in all 4 tests, and the same expect(screen.getByTestId("data").textContent).toBe(JSON.stringify(...)) assertion everywhere.

What it became

• createLiveQuerySource(initialItems) in src/tests/liveQuery.tsx — the shared home these suites already import from (createQueryClient, createWrapper, liveQueryOptions were consolidated there by test: dedupe isDefined helper against the library util #144's follow-ups). It builds the liveQuery$/db$ pair and the filter → switchMap → map queryFn that emulates a live database query.
• One setup() helper per suite (component shape differs: plain component vs. hide/show host), plus a local expectData assertion helper.

Why they are truly the same concept

All 10 copies exercise the same fixture — "a live query backed by a db-readiness gate that re-emits a fresh array on every push" — against the same rendered output. They were copy-pasted, differing only in query key, initial items, and (in two tests) query options, which are now plain parameters. No behavior changes: query keys, options (including the two tests that intentionally omit staleTime: Infinity), emission sequences, timings, and assertions are byte-for-byte equivalent, and the two-observers test keeps its bespoke double-hook component.

Gates (same commands CI runs, all green before and after; 139/139 tests pass both times)

• npm run check
• npm run build (tsc + vite)
• npm run test:ci

Other candidates surveyed and rejected

• useSwitchMutation$ / useConcatMutation$: shared adapters were already extracted (mutationOptions.ts); the remaining similarity is genuinely divergent cancellation semantics (abort vs. ready-gate).
• createLocalStorageAdapter / createLocalforageAdapter: superficially similar JSON (de)serialization but different backends with intentionally different removeItem/clear semantics.
• Remaining jscpd hits (persistSignals.test.ts, useObserve.compare.test.tsx, useSwitchMutation$.test.tsx) are small intra-file test-case variations; consolidating them would trade readability for little LOC — possible future runs if they grow.

🤖 Generated with Claude Code

https://claude.ai/code/session_015zuCUHEHA1tWwYQ28x4rgB

---

Generated by Claude Code