Github user xubo245 commented on a diff in the pull request:
https://github.com/apache/carbondata/pull/2372#discussion_r196633489
--- Diff:
hadoop/src/main/java/org/apache/carbondata/hadoop/CarbonRecordReader.java ---
@@ -80,7 +80,7 @@ public void initialize(InputSplit inputSplit,
TaskAttemptContext context)
}
// It should use the exists tableBlockInfos if tableBlockInfos of
queryModel is not empty
// otherwise the prune is no use before this method
- if (!queryModel.isFG()) {
+ if (queryModel.getTableBlockInfos().isEmpty()) {
--- End diff --
If use (queryModel.getTableBlockInfos().isEmpty()), when the prune result
is empty by FG in search mode, it will use the original TableBlockInfos and
execute again, it mean the FG no use in this scenario. So we cannot change it
to queryModel.getTableBlockInfos().isEmpty()
---
