eugenegujing opened a new pull request, #6722:
URL: https://github.com/apache/texera/pull/6722
### What changes were proposed in this PR?
Add `AsterixDBConnUtilSpec`, unit test coverage for `AsterixDBConnUtil` —
the HTTP-based connection utility behind the AsterixDB source operator. It
previously had no tests.
Since the utility talks plain HTTP via Unirest, the spec exercises it
end-to-end against an in-process JDK `HttpServer` stub standing in for an
AsterixDB instance (the same pattern as `LiteLLMProxyAuthSpec`; no new
dependencies, no production-code changes). The stub serves the two endpoints
the utility touches: `/admin/version` with a hit counter and configurable
status/body, and `/query/service` with a swappable statement-based responder
that records every decoded form field, which is what makes cache reuse and the
outgoing `format` field observable.
The 32 characterization tests cover: per-host version-cache population and
reuse (`/admin/version` probed exactly once per host, pre-seeded entries skip
the probe); the version-dependent `format` field (`"csv"` for 0.9.5 vs
`"text/csv"` otherwise, pinned as an exact-equality check, including
caller-supplied formats); the `results` iterator on HTTP 200 and the strict `==
200` status check (201/204 are routed to the error path); the non-200
`RuntimeException` embedding status and body, plus the current failure modes
for malformed responses (a bare NPE on a non-JSON body due to Unirest's null
body on parse failure, and a `JSONException` on a 200 body without `results`);
`updateAsterixDBVersionMapping`'s status-only tolerance (silent on non-200,
throws on a 200 with an unexpected body shape); and `fetchDataTypeFields`
recursive flattening (dot-prefixed parent naming, two-level nesting,
empty-metadata tolerance, silent drop of nested fields with missing child
metadata, the exact `Try`
tolerance boundary where a field row missing `FieldType` escapes uncaught, the
case-sensitive `contains("type")` recursion heuristic, and first-row-only
consumption of multi-row results). The statement form field is also verified to
round-trip special characters and non-ASCII text intact.
The mutable singleton version cache is reset before each test and cleared
after the suite. Tests pin current behavior, including the
`NoSuchElementException` thrown when the version probe fails before a query;
behavior changes to those failure modes belong in follow-up fixes that flip the
corresponding pins.
### Any related issues, documentation, discussions?
Closes #6721.
### How was this PR tested?
This PR only adds tests. `sbt "WorkflowOperator/testOnly *ConnUtilSpec"`
runs all three ConnUtil suites: 44 tests, all passing (AsterixDB 32 + MySQL 8 +
PostgreSQL 4). `sbt "WorkflowOperator/Test/scalafmtCheck"` passes. Exception
types asserted for the malformed-response cases were first observed against the
real Unirest behavior, then pinned.
### Was this PR authored or co-authored using generative AI tooling?
Co-authored by: Claude Code
--
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]
