XuQianJin-Stars opened a new pull request, #3574:
URL: https://github.com/apache/fluss/pull/3574
Introduce `FlussRecordAsIcebergRecordTest` that mirrors the coverage matrix
of `FlussRecordAsPaimonRowTest`, focusing on behaviors specific to the tiering
wrapper on top of `FlussRowAsIcebergRecord`:
- **All-types log-table row**: dispatch of `__bucket` / `__offset` /
`__timestamp` system columns via both positional `get(pos)` and name-based
`getField(name)`, alongside the existing 14 primitive/business columns
delegated to the parent wrapper.
- **Primary-key table with all `ChangeType` values** (`INSERT` /
`UPDATE_BEFORE` / `UPDATE_AFTER` / `DELETE` / `APPEND_ONLY`): asserts that
`offset` / `timestamp` are refreshed on every `setFlussRecord` call and system
columns remain consistent.
- **Field count mismatch**: `setFlussRecord` throws `IllegalStateException`
when Fluss row field count does not match `struct.size - system`.
- **Complex-type propagation via the parent converter**: `ARRAY<INT>`,
`ARRAY<ARRAY<INT>>`, `ARRAY` with nullable / empty / null variants,
`MAP<STRING,INT>` populated / empty / null, and `ROW<INT,STRING>` including a
null nested row.
This mirrors the guardrails present in `fluss-lake-paimon`'s
`FlussRecordAsPaimonRowTest` so log-table tiering into Iceberg has equivalent
unit-test coverage.
## Purpose
`fluss-lake-iceberg`'s tiering wrapper `FlussRecordAsIcebergRecord` sits
between the raw Fluss `LogRecord` stream and Iceberg's `Record` interface,
injecting the three system columns (`__bucket`, `__offset`, `__timestamp`) on
top of the field-by-field conversion done by `FlussRowAsIcebergRecord`. Prior
to this PR, only the parent converter had unit-test coverage; the wrapper's
system-column dispatch, field-count guardrail, per-record `offset`/`timestamp`
refresh semantics and complex-type propagation were tested only indirectly
through end-to-end IT cases. Meanwhile, the sibling `fluss-lake-paimon` module
already carries a dedicated `FlussRecordAsPaimonRowTest` that exercises exactly
this contract for its Paimon counterpart.
This PR closes the gap by adding an Iceberg equivalent test class that pins
down the same contract, so future refactors of the Iceberg tiering path (adding
new complex types, changing system-column ordering, tightening the mismatch
check, etc.) will surface regressions at unit-test speed rather than IT-test
speed.
Linked issue: close #xxx
## Brief change log
Test-only change. No production code is touched.
- Add
`fluss-lake/fluss-lake-iceberg/src/test/java/org/apache/fluss/lake/iceberg/tiering/FlussRecordAsIcebergRecordTest.java`
(+477 lines, 8 `@Test` methods):
- `testLogTableRecordAllTypes` — all-types (14 primitive/business columns
+ 3 system columns) log-table row; asserts both positional `get(pos)` and
name-based `getField(name)` for every column, including `__bucket` / `__offset`
/ `__timestamp`.
- `testPrimaryKeyTableRecordAllChangeTypes` — iterates over `INSERT` /
`UPDATE_BEFORE` / `UPDATE_AFTER` / `DELETE` / `APPEND_ONLY`, calling
`setFlussRecord` with distinct offsets/timestamps each time, and asserts the
system columns are refreshed and business columns remain consistent per record.
- `testFieldCountMismatchThrows` — feeds a row whose field count does not
match `struct.size - system` and asserts `IllegalStateException` is thrown by
`setFlussRecord`.
- `testArrayTypeWithIntElements` — `ARRAY<INT>` propagation through the
parent converter.
- `testNestedArrayType` — `ARRAY<ARRAY<INT>>` propagation.
- `testArrayWithNullableElementsAndEmptyAndNullArray` — `ARRAY` with
nullable elements, empty array, and top-level null array.
- `testMapTypeWithStringToIntAndEmptyAndNullMap` — `MAP<STRING,INT>`
populated / empty / null.
- `testNestedRowType` — `ROW<INT,STRING>` including a top-level null
nested row.
The 8 test methods together mirror the coverage matrix of
`fluss-lake-paimon`'s `FlussRecordAsPaimonRowTest`, adapted to the Iceberg
`Record` / `Types.NestedField` API.
## Tests
New tests are the deliverable of this PR. Verified locally:
- `./mvnw -pl fluss-lake/fluss-lake-iceberg test
-Dtest=FlussRecordAsIcebergRecordTest -Dspotless.check.skip` — all 8 tests pass.
- `./mvnw -pl fluss-lake/fluss-lake-iceberg test -Dspotless.check.skip` —
full module test suite still green (no interference with existing tests).
- `./mvnw -pl fluss-lake/fluss-lake-iceberg checkstyle:check` — 0 violations.
No production behavior change; existing tests are unchanged.
## API and Format
None. This PR only adds tests under `src/test/java`. There are no changes to
public Java API, table properties, wire format, or Iceberg on-disk format.
## Documentation
None. Test-only change; no user-facing documentation impact.
--
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]
