aokolnychyi commented on code in PR #4652:
URL: https://github.com/apache/iceberg/pull/4652#discussion_r890336099
##########
spark/v3.2/spark/src/main/java/org/apache/iceberg/spark/actions/BaseDeleteOrphanFilesSparkAction.java:
##########
@@ -361,4 +403,90 @@ static PathFilter forSpecs(Map<Integer, PartitionSpec>
specs) {
return partitionNames.isEmpty() ? HiddenPathFilter.get() : new
PartitionAwareHiddenPathFilter(partitionNames);
}
}
+
+ private static Dataset<Row> sanitizeAndSplitPaths(Dataset<Row> filesDF) {
+ return filesDF.mapPartitions((MapPartitionsFunction<Row, Row>) input ->
Review Comment:
I actually think we should offer methods that return `Dataset<String>` for
valid paths in the base class. That way, we can remove one round of unnecessary
conversion to and from internal Spark representation. Right now, we have a
dataset of string, which we convert to a dataset of rows and immediately
convert back to access strings. Spark won't optimize this, unfortunately.
The idea of using a bean to carry the metadata sounds promising. I'd give it
a try.
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
