feat: Support Spark explode outer #11954
Conversation
facebook-github-bot
added
the
CLA Signed
✅ Deploy Preview for meta-velox ready!
To edit notification comments on pull requests, go to your Netlify site configuration. |
|
@pedroerp @rui-mo @jinchengchenghh @PHILO-HE Can you please help review this? |
velox/exec/tests/UnnestTest.cpp
Outdated
| // is_outer = false | ||
| auto op1 = PlanBuilder().values({vector}).unnest({"c0"}, {"c1"}, std::nullopt, false /* is_outer */).planNode(); | ||
| auto params1 = makeCursorParameters(op1); | ||
| auto expected1 = makeRowVector({ |
There was a problem hiding this comment.
nit: can we have 'op', 'params', 'expected' and reuse them in the other test cases?
There was a problem hiding this comment.
nit: can we have 'op', 'params', 'expected' and reuse them in the other test cases?
just follow the style of other test cases in UnnestTest. Maybe another PR to refactor it?
velox/exec/tests/UnnestTest.cpp
Outdated
|
|
||
| createDuckDbTable({vector}); | ||
|
|
||
| // is_outer = false |
There was a problem hiding this comment.
Thanks for your contribution. Overall looks good. % minors.
I don't think we need to use isOuter as template, isOuter is not a too expensive cost, It will make the change not precise. What do you think? @rui-mo
| if (rawOrdinalityIsNull_[row]) { | ||
| ordinalityVector->setNull(index, true); | ||
| } | ||
| } | ||
| std::iota(rawOrdinality, rawOrdinality + size, start + 1); | ||
| rawOrdinality += size; |
There was a problem hiding this comment.
Then we don't need to move rawOrdinality, just use rawOrdinality + index in L263.
There was a problem hiding this comment.
you mean change it to std::iota(rawOrdinality + index, rawOrdinality + index + size, start + 1); right?
Support the explode/posexplode with outer semantic. Unlike explode/posexplode, if the array/map is null or empty then null is produced.
Spark doc: https://spark.apache.org/docs/latest/api/python/reference/pyspark.sql/api/pyspark.sql.functions.explode_outer.html