Github user jinfengni commented on a diff in the pull request:
https://github.com/apache/drill/pull/371#discussion_r52563924
--- Diff:
exec/java-exec/src/main/java/org/apache/drill/exec/store/parquet/ParquetGroupScan.java
---
@@ -791,6 +799,43 @@ public FileGroupScan clone(FileSelection selection)
throws IOException {
}
@Override
+ public GroupScan applyLimit(long maxRecords) {
--- End diff --
I gave some thoughts about this optimization as well. Then, I realized that
until we have some performance measurement, it's not very clear which way we
want to have. For example, I'm not clear whether 1000 small parquet files is
better than 1 large parquet files. 1000 files might have big metadata overhead
than 1 large file (?). But 1000 small files might be better option, in case we
do want to parallelize the execution.
I'll add some comment saying further optimization could be done in terms of
how subset of files are chosen.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
