xylaaaaa commented on issue #69884: URL: https://github.com/apache/doris/issues/69884#issuecomment-6659316728
Reader review at [`a8954bdc70`](https://github.com/apache/doris/pull/69904/commits/a8954bdc70dd0d450e39dabe6201a6042544dfb5) found a file-pruning correctness issue on the existing Spark golden table `delta_golden.type_widening`. It reproduces through the dedicated filesystem and HMS catalogs and through the Hive HMS gateway. ```sql SELECT decimal_decimal_same_scale, CAST(decimal_decimal_same_scale AS BIGINT) FROM delta_golden.type_widening; -- Includes (123.45, 123). SELECT decimal_decimal_same_scale FROM delta_golden.type_widening WHERE CAST(decimal_decimal_same_scale AS BIGINT) = 123; -- Expected 123.45. All three entry points return zero rows and EXPLAIN selects 0/2 files. SELECT decimal_decimal_same_scale FROM delta_golden.type_widening WHERE decimal_decimal_same_scale >= 123 AND decimal_decimal_same_scale < 124; -- Returns 123.45 and selects 1/2 files. ``` The retained Parquet file and Delta add stats both put the first file's decimal min/max at 123.45. The FE expression converter unwraps `CastExpr`. Delta inherits the SPI's permissive cast-pushdown default, so its Kernel filter can compare the original decimal to 123 and exclude the file before BE evaluates the original predicate. The BE receives no data split for the failing query. This needs a correctness fix and coverage for the three entry points before broadening predicate pushdown. The reader review has not changed product 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
