Github user rdblue commented on the issue:
https://github.com/apache/spark/pull/21911
@jzhuge, I think it is confusing that this hint exposes the `shuffle`
boolean flag. The Spark API makes a clear distinction between `repartition` and
`coalesce` where `coalesce` means that Spark won't add a shuffle. Adding the
boolean here would allow users to coalesce with a shuffle (repartition) or
repartition without a shuffle (coalesce). I can't think of a good reason to do
that. The reason why there is a boolean `shuffle` param is to use the same plan
node internally, but I don't think that detail should be leaked into SQL hints.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
