peterxcli opened a new issue, #5716:
URL: https://github.com/apache/datafusion-comet/issues/5716
### Describe the bug
Spark accepts a leading `+` on a timestamp year, but Comet's native
string-to-timestamp parser returns NULL. Widening the timestamp fuzz alphabet
in PR #5682 exposed this on Spark 4.1.3. Observed inputs include `+7528`,
`+00463`, `+79821`, and `+2976`.
### To reproduce
Use Parquet-backed column input to avoid constant folding:
```sql
SET spark.sql.session.timeZone=UTC;
SET spark.sql.ansi.enabled=false;
CREATE TABLE positive_year_probe USING parquet AS SELECT '+7528' AS s;
SELECT s, CAST(s AS TIMESTAMP) FROM positive_year_probe;
```
With native execution disabled, Spark returns the start of year 7528. With
Comet enabled, the cast returns NULL. The failing differential run on PR head
`0a1fc76a0` plus the review follow-up used 10,000 seeded random strings and
exposed nine such bare positive-year inputs. No other mismatch appeared in that
run.
### Expected behavior / proposed fix
Accept explicit positive years according to Spark's timestamp segment rules.
Current patterns use `-?[0-9]{4,6}`, excluding `+`. Audit both timestamp and
timestamp_ntz paths and numeric decoding before broadening the sign grammar;
add column-input Legacy/TRY/ANSI coverage, including signed dates and malformed
signs. Remove the narrow positive-bare-year fuzz exclusion in
`CometNativeCastSuite` once fixed.
--
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]
