weiqingy commented on code in PR #2438:
URL: https://github.com/apache/auron/pull/2438#discussion_r3694512949
##########
spark-extension-shims-spark/src/test/scala/org/apache/auron/AuronQuerySuite.scala:
##########
@@ -883,6 +884,51 @@ class AuronQuerySuite extends AuronQueryTest with
BaseAuronSQLSuite with AuronSQ
}
}
+ test("native broadcast hash join supports inner residual condition") {
Review Comment:
This pins the inner accept path for both build sides, which is the half the
PR changes. The other half of the contract in #2437 is "continue rejecting
residual conditions for outer, semi, anti, and other broadcast join types", and
there are now four guards enforcing it with no test behind any of them:
`AuronConverters.scala:706`, `NativeBroadcastJoinBase.scala:129`,
`planner.rs:433-437`, and `broadcast_join_exec.rs:110-112`. The Rust side is
inner-only as well, since `bhj_collect_with_filter` pins `Inner` at
`test.rs:399` and `:430`.
#2197 set the same precedent for SMJ/SHJ, so nothing regressed here. Worth
adding one negative case so the reject half is pinned too?
One way it could look, in case it helps:
```scala
val df = checkSparkAnswer("""
|SELECT /*+ BROADCAST(r) */ l.id
|FROM bhj_left l LEFT JOIN bhj_right r ON l.id = r.id AND l.lv < r.rv
|""".stripMargin)
assert(
stripAQEPlan(df.queryExecution.executedPlan)
.collectFirst { case _: NativeBroadcastJoinExec => true }
.isEmpty)
```
--
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]
