Github user jackylk commented on a diff in the pull request:
https://github.com/apache/incubator-carbondata/pull/642#discussion_r105578013
--- Diff:
core/src/main/java/org/apache/carbondata/core/datastore/columnar/BlockIndexerStorageForShort.java
---
@@ -192,12 +192,23 @@ private void
compressDataMyOwnWay(ColumnWithShortIndex[] indexes) {
}
map.add(start);
map.add(counter);
- this.keyBlock = convertToKeyArray(list);
- if (indexes.length == keyBlock.length) {
- dataIndexMap = new short[0];
- } else {
+ boolean useRle = (((list.size() + map.size()) * 100) / indexes.length)
> 70 ? false : true;
--- End diff --
Can we decide this in a more heuristic way? Like if we find there are more
than 5 pages not doing RLE, then do not pay the cost to try to compress it in
all future blocklets.
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
