raghavyadav01 commented on issue #26700: URL: https://github.com/apache/datafusion/issues/26700#issuecomment-6718037611
Ran the full corpus locally at batch sizes 1 and 7 with an env override (8192 baseline clean). 229 files differ at 1 and 84 at 7, but nearly all of it is output order without `ORDER BY`, `LIMIT` without `ORDER BY`, ties, `EXPLAIN`, or float last digits, which supports comparing only order-insensitive results for non-default sizes. Real findings so far: `array_concat(2D, 1D)` returns `[NULL]` or `NULL` for the same all-NULL row depending on the other rows in its batch (reproduces at the default size with a `WHERE`), a rejected `INSERT` leaves earlier batches written at size 1 (`listing_table_partitions.slt:252`), and `nested_loop_join_spill` / `push_down_filter_regression` go from seconds to over 3 minutes at size 1. I'll file the `array_concat` one separately. -- 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]
