The filtered CTE was referenced twice (count + data), preventing
Postgres 12+ from inlining it. This forced full materialization of
the filtered set before LIMIT could short-circuit. Split into two
separate queries run via Promise.all so the list CTE is only
referenced once and the planner can short-circuit on LIMIT.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Fredrik Adelöw <freben@gmail.com>
Now that renderInTestApp properly sets up Route matching, the test can
use the real in-memory catalog client instead of jest stubs. The
in-memory client does actual ref-based lookups, verifying that
useEntityFromUrl receives correct route params end-to-end.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Fredrik Adelöw <freben@gmail.com>
Only wrap the SELECT+UPDATE in a transaction for MySQL/PostgreSQL
where FOR UPDATE SKIP LOCKED is actually used. For sqlite3 the
transaction adds unnecessary BEGIN/COMMIT overhead.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Fredrik Adelöw <freben@gmail.com>
renderInTestApp disables app/routes, so the test element was rendered
without any Route matching. This meant useParams() always returned {}
even when initialRouteEntries and mountedRoutes were set — the mounted
routes only fed the route resolution API for link generation, never
created actual React Router <Route> matching for param extraction.
Wrap the test element in <Routes> with a <Route> for each mounted path
(plus a path="*" fallback), mirroring what the real AppRoutes extension
does via useRoutes(). This makes useParams() return real params when
initialRouteEntries matches a mounted route path.
Update entity page test href expectations to match the now-correct
production behavior where sub-route links resolve relative to the
matched parent Route, and replace synchronous getByTestId/getByText
with async findBy queries for lazy-loaded content.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Fredrik Adelöw <freben@gmail.com>
The entity page resolves route params via useRouteRefParams. Without
initialRouteEntries, the MemoryRouter starts at '/' which doesn't
match the entity route pattern. On fast machines the mock entity
resolves before the useEffect error check fires, but on slow CI
runners the timing differs and the "No name provided!" error path
wins, causing the test to flake.
Adding initialRouteEntries ensures the router starts at a URL that
matches /catalog/:namespace/:kind/:name, so route params are always
available regardless of timing.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Fredrik Adelöw <freben@gmail.com>
The InMemoryCatalogClient + initialRouteEntries approach didn't work
because route params weren't being resolved by the test harness,
causing getEntityByRef to receive an empty ref. The original mock
with getEntityByRef: async () => entityMock works correctly. The
only change from master is adding the Entity type annotation.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Fredrik Adelöw <freben@gmail.com>
Same FOR UPDATE SKIP LOCKED race: the processing loop's claim query
had the SELECT and UPDATE as separate auto-committed statements.
The existing caller already wraps in a transaction, but this makes
the function self-protecting if called without one.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Fredrik Adelöw <freben@gmail.com>
The test rendered at '/' instead of a URL matching the entity route
pattern, causing useEntityFromUrl to sometimes fail to extract route
params. Use initialRouteEntries to start at the correct entity URL,
and switch from catalogApiMock.mock() to catalogApiMock() for a
working in-memory catalog client.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Fredrik Adelöw <freben@gmail.com>
The env vars used lowercase postgres18/postgres14 instead of uppercase
POSTGRES18/POSTGRES14, which didn't match the names expected by
TestDatabases. This caused all postgres-only test suites to see no
available databases and crash.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Fredrik Adelöw <freben@gmail.com>
When eachSupportedId() returned an empty array, describe.each([]) would
throw and crash the entire test suite. Return a placeholder ID instead,
so individual tests fail with a clear error rather than preventing the
entire suite from running.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Fredrik Adelöw <freben@gmail.com>
The search table is an EAV schema where key and value are highly
correlated, but standard single-column statistics can't model this.
This caused the planner to underestimate compound filter queries by
up to 1300x, choosing materialize-then-sort plans instead of LIMIT
short-circuit index scans.
CREATE STATISTICS captures the (key, value) correlation via
dependencies, ndistinct, and most-common-values metadata. The planner
then correctly estimates row counts and chooses the index-driven plan.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Fredrik Adelöw <freben@gmail.com>
The SELECT FOR UPDATE SKIP LOCKED and the subsequent UPDATE of
next_stitch_at now run inside a single transaction. Previously the
row locks were released after the SELECT auto-committed, allowing
another worker to claim the same rows before the UPDATE ran.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Fredrik Adelöw <freben@gmail.com>
This change aligns the MCP tool execution responses with the specification by returning plain text in the 'content' array and providing the raw JSON in a new 'structuredContent' field. It also removes the unnecessary Markdown code block formatting that was previously added to tool results.
Fixes#34052
Signed-off-by: Love Kumar Chauhan <lovechauhan6564@gmail.com>
Co-authored-by: benjdlambert <3645856+benjdlambert@users.noreply.github.com>
Signed-off-by: benjdlambert <ben@blam.sh>
We should mostly get these for free because of the
GitHub conditional request caching, but this is a
nice safety net and also probably saves us some
time.
Signed-off-by: MT Lewis <mtlewis@users.noreply.github.com>
Keeping this change as minimal as possible for
this first phase, will restore the parallel
processing in a later iteration.
Signed-off-by: MT Lewis <mtlewis@users.noreply.github.com>
Only delete the cache entry on error if the map still holds the same
promise, preventing a stale rejection from evicting a newer entry.
Also switch the test to createDeferred for readability.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Fredrik Adelöw <freben@gmail.com>
Fredrik Adelöw