aglinxinyuan opened a new issue, #8399:
URL: https://github.com/apache/texera/issues/8399
### Task Summary
`DatasetResourceSpec.beforeAll` sets the JVM-global
`FulltextSearchQueryUtils.usePgroonga` to `false`
and never puts it back:
```scala
override protected def beforeAll(): Unit = {
initializeDBAndReplaceDSLContext()
FulltextSearchQueryUtils.usePgroonga = false // disable pgroonga
...
}
```
The write has no effect on this suite. `usePgroonga` is read at exactly one
place in `src/main`,
`FulltextSearchQueryUtils.scala:52`, inside `getFullTextSearchFilter`. All
four of this suite's tests
reach that function — if at all — with an empty keyword list: its two search
tests call
`DatasetSearchQueryBuilder.constructQuery(uid,
SearchQueryParams(resourceType = DATASET_RESOURCE_TYPE),
includePublic = true)`, leaving `params.keywords` at its empty-`ArrayList`
default
(`DashboardResource.scala:72`), so `getFullTextSearchFilter` returns
`noCondition()` at `:46` before
the `:52` read. The other two tests only use `UserDao`.
Confirmed by measurement: arming the read site with a `sys.error` leaves
`DatasetResourceSpec` at
4 tests / 0 failures, while `FulltextSearchQueryUtilsSpec` — which does read
the flag — loses exactly
its three flag-reading tests.
So the write's only effect is the leak. amber has no `Test / fork` and
serialises its suites in one
JVM, so `false` stays set for every suite scheduled after this one, silently
moving their full-text
rendering onto the `to_tsvector`/`to_tsquery` arm instead of the production
`pgroonga_condition` arm.
An ordered probe (both suites in one sbt invocation, order pinned) observes
`true` before
`DatasetResourceSpec` runs and `false` after it finishes.
The same block also declares `private def getKeywordsArray`, which nothing
calls — it and the flag
write look like copy-paste residue from `WorkflowResourceSpec`, brought over
for search tests that
were never written here.
Proposed cleanup:
- delete the `usePgroonga = false` write, leaving the flag at its production
default (a
capture-and-restore would still leave the value wrong during the suite and
would depend on sbt's
scheduling, which is not even stable across invocations of the same
command); leave a short comment
saying why, and what a keyword test added here would need;
- delete the unused `getKeywordsArray` and the two imports that only these
two members needed;
- update the header paragraph in `DatasetSearchQueryBuilderSpec`, which
documents this leak as fact
while justifying a standing instruction about that spec's keyword
assertions, so the instruction is
justified by the flag being shared mutable state rather than by any
particular suite's behaviour —
and so that it states precisely what survives onto both arms of the flag
(the concatenated
`coalesce` expression and each individual keyword token) and what does not
(their joining: the
`true` arm space-joins the keyword list into one literal, the `false` arm
emits one predicate per
keyword and joins the words inside a keyword with ` & `).
Out of scope: `WorkflowResourceSpec` sets the same flag `false` and also
never restores it, but it
genuinely needs the `to_tsvector` arm — it runs real keyword searches
against the embedded Postgres,
which has no pgroonga extension because `MockTexeraDB` strips the full-text
index block out of the
DDL. Fixing that one needs a scoped override, not a deletion.
### Task Type
- [x] Refactor / Cleanup
### Was this issue authored using generative AI tooling?
Generated-by: Claude Code (Opus 5)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
