jayzhan211 commented on issue #26661: URL: https://github.com/apache/datafusion/issues/26661#issuecomment-6384279512
I ran all 409 queries against `main` with the benchmark's schema and a few rows per table, so this covers queries that fail, not wrong results or timeouts. Of the 78 queries that fail on 55.0.0, 31 work on `main`. Three of those 31 failed on 55.0.0 while comparing decimals at run time, which needs the real data to confirm. What still fails on `main`, plus one query that broke after 55.0.0: | Cause | Queries | Tracked in | |---|---|---| | Subquery inside an aggregate (`GROUP BY` key, aggregate argument or `FILTER`) | 13 | #26704 | | Correlated scalar subquery with `ORDER BY ... LIMIT 1`, mostly "the exchange rate as of a date" | 11 | #26710 | | Correlated subquery whose correlation is a range rather than an equality | 8 | #14554, #5492 | | Correlated scalar subquery without an aggregate | 5 | #16137 | | Correlated subquery that references a table two levels up (`Schema error: No field named p.id`) | 3 | #15558, #21727 | | Integer or float times an interval or duration | 3 | #13850 | | `json_value` | 2 | not built in | | Scalar subquery in a `JOIN ... ON` condition | 1 | | | `abs(COALESCE(...))` in a window `ORDER BY` fails `SanityCheckPlan` | 1 | #26430 | | Decimal `sum` next to `count(DISTINCT)`, broken since #24859 | 1 | #26193, fix in #26211 | 18 queries correlate on a range: the 8 in the range row, 8 of the 11 `LIMIT 1` queries and two in the aggregate row. Nine of them are the same exchange-rate lookup (`er.as_of <= o.closed_at ORDER BY er.as_of DESC LIMIT 1`), which the ASOF join on `main` already computes, see #26710. The other nine need general range decorrelation. <details> <summary>Queries per row</summary> | Cause | Queries | |---|---| | Subquery inside an aggregate | q057 q084 q085 q091 q135 q136 q156 q178 q182 q312 q315 q370 q409 | | `ORDER BY ... LIMIT 1` | q077 q174 q176 q181 q183 q194 q365 q368 q383 q386 q388 | | Range correlation | q086 q113 q162 q167 q273 q274 q289 q403 | | Scalar subquery without an aggregate | q098 q179 q263 q363 q367 | | Two levels up | q353 q358 q393 | | Interval or duration arithmetic | q250 q341 q379 | | `json_value` | q175 q399 | | Scalar subquery in `JOIN ... ON` | q064 | | `abs(COALESCE(...))` window | q121 | | Decimal `sum` | q158 | </details> Anyone is welcome to pick up a row; feel free to ping me for the exact errors of each query. -- 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]
