Github user jkbradley commented on the issue:
https://github.com/apache/spark/pull/20442
@huaxingao Thanks for this follow-up! I realized that
https://github.com/apache/spark/pull/19715 introduced a breaking change which
we missed in Spark 2.3 QA: In Spark 2.2, a user could set inputCol but not set
outputCol (since outputCol has a default value). The new check causes such
user code to start failing in Spark 2.3.
Since you're already working on this follow-up, would you mind adding a
unit test which checks this? (Setting inputCol but not outputCol, and making
sure that works.)
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
