geserdugarov commented on code in PR #11501:
URL: https://github.com/apache/hudi/pull/11501#discussion_r1665804002
##########
hudi-common/src/main/java/org/apache/hudi/BaseHoodieTableFileIndex.java:
##########
@@ -360,13 +363,22 @@ protected HoodieTimeline getActiveTimeline() {
}
private Object[] parsePartitionColumnValues(String[] partitionColumns,
String partitionPath) {
- Object[] partitionColumnValues =
doParsePartitionColumnValues(partitionColumns, partitionPath);
- if (shouldListLazily && partitionColumnValues.length !=
partitionColumns.length) {
- throw new HoodieException("Failed to parse partition column values from
the partition-path:"
- + " likely non-encoded slashes being used in partition column's
values. You can try to"
- + " work this around by switching listing mode to eager");
+ HoodieTableConfig tableConfig = metaClient.getTableConfig();
+ Object[] partitionColumnValues;
+ if (null != tableConfig.getKeyGeneratorClassName()
+ &&
tableConfig.getKeyGeneratorClassName().equals(KeyGeneratorType.TIMESTAMP.getClassName())
+ &&
tableConfig.propsMap().get(TimestampKeyGeneratorConfig.TIMESTAMP_TYPE_FIELD.key()).matches("SCALAR|UNIX_TIMESTAMP|EPOCHMILLISECONDS"))
{
+ // For TIMESTAMP key generator when TYPE is SCALAR, UNIX_TIMESTAMP or
EPOCHMILLISECONDS,
+ // we couldn't reconstruct initial partition column values from
partition paths due to lost data after formatting in most cases
+ partitionColumnValues = new Object[partitionColumns.length];
Review Comment:
I could pass partition column values as `UTF8String`s instead of nulls if
implement the similar logic in
`SparkHoodieTableFileIndex::doParsePartitionColumnValues`. But I found out more
tricky problem with range filters on partition column.
For instance, we have `UNIX_TIMESTAMP` in column `ts`. And the table is also
partitioned by the same `ts` with
`hoodie.keygen.timebased.output.dateformat = "yyyy-MM-dd HH"`
Execution of query like:
`SELECT ... WHERE ts BETWEEN 1078016000 and 1718953003 ...`
will pass an empty result even with passed partition paths as `UTF8String`s,
because we need to deal with initial `UNIX_TIMESTAMP` values for this filtering.
Need to think about correct filtering in this case.
--
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]
