peterxcli commented on PR #5161:
URL: 
https://github.com/apache/datafusion-comet/pull/5161#issuecomment-5443932242

   Thanks for the pass — point by point:
   
   **Match arm ordering** — agreed it read as "exempt from everything" rather 
than "supported." Moved the arm to just above the `super.isTypeSupported` 
fallthrough (it has to stay above the super call, since the shared default 
rejects intervals) with a comment. No behavior change: an interval can never 
match the `ShortType`, collation, or struct guards.
   
   **V2 coverage** — the V2 Parquet path doesn't go through 
`CometScanTypeChecker`: `transformV2Scan` gates on 
`CometBatchScanExec.isSchemaSupported`, which uses the shared `DataTypeSupport` 
default and still rejects intervals, so a V2 interval scan falls back to Spark 
exactly as before this PR. (`CometScanTypeChecker` is shared with the V2 
Iceberg path, but Iceberg has no interval type.) The fixture's `Config:` line 
only pins Spark's default — parquet is already in the default `useV1SourceList` 
— and I've added a comment in the fixture saying so.
   
   **Native predicate coverage** — `ParquetReadSuite`'s "ANSI interval types" 
test already filters on `i2 = INTERVAL '1-2' YEAR TO MONTH AND i9 = INTERVAL '1 
02:03:04.5' DAY TO SECOND` under `checkSparkAnswerAndOperator`, asserts 
`dataFilters` is non-empty, and checks `output_rows` == 333 with pushdown vs 
1000 without, so both the literal encoding and row-group pruning are pinned 
against Spark's answer. I've additionally added `ym >` / `dt >` predicate 
queries to `ansi_interval_scan.sql` so the SQL golden harness covers it across 
Spark versions too.
   
   **Prettier** — `npx prettier --check 
docs/source/user-guide/latest/datatypes.md` passes on the branch as-is, and the 
markdown CI check is green on the latest push, so no realignment is pending.
   
   Separately, the one failing check (`PR Build (Linux) / Spark 4.1, JDK 17 
[exec]`) was an HTTP 429 fetching the Maven distribution during setup — infra 
flake; a re-run should clear it.
   
   Both targeted suites pass locally on Spark 4.1 / JDK 17 after the change: 
`ParquetReadV1Suite "ANSI interval types"` and `CometSqlFileTestSuite 
"sql-file: expressions/datetime/ansi_interval_scan.sql"`.
   


-- 
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]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to