KurtYoung commented on a change in pull request #9909: [FLINK-14381][table]
Partition field names should be got from CatalogTable instead of source/sink
URL: https://github.com/apache/flink/pull/9909#discussion_r338525048
##########
File path:
flink-table/flink-table-planner-blink/src/main/scala/org/apache/flink/table/planner/plan/rules/physical/batch/BatchExecSinkRule.scala
##########
@@ -45,28 +43,34 @@ class BatchExecSinkRule extends ConverterRule(
val newTrait = rel.getTraitSet.replace(FlinkConventions.BATCH_PHYSICAL)
var requiredTraitSet =
sinkNode.getInput.getTraitSet.replace(FlinkConventions.BATCH_PHYSICAL)
sinkNode.sink match {
- case partitionSink: PartitionableTableSink
- if partitionSink.getPartitionFieldNames != null &&
- partitionSink.getPartitionFieldNames.nonEmpty =>
- val partitionFields = partitionSink.getPartitionFieldNames
- val partitionIndices = partitionFields
- .map(partitionSink.getTableSchema.getFieldNames.indexOf(_))
- // validate
- partitionIndices.foreach { idx =>
- if (idx < 0) {
- throw new TableException(s"Partitionable sink ${sinkNode.sinkName}
field " +
- s"${partitionFields.get(idx)} must be in the schema.")
- }
+ case partitionSink: PartitionableTableSink =>
Review comment:
It doesn't feel right for current convert logic.
When output to a partitioned table, I think we should first see whether this
table is partitioned, through `CatalogTable.isPartitioned()`. If the table is
partitioned, then we have to check whether we have a `PartitionableTableSink`
to output data.
Current logic seems to be just opposite. We first check whether we have
partition sink capability, then we check whether the table itself is
partitioned.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
