Github user sounakr commented on a diff in the pull request:
https://github.com/apache/carbondata/pull/2257#discussion_r185403178
--- Diff: hadoop/src/main/java/org/apache/carbondata/hadoop/api/CarbonFileInputFormat.java ---
@@ -126,13 +132,27 @@ protected CarbonTable getOrCreateCarbonTable(Configuration configuration) throws
FilterResolverIntf filterInterface = carbonTable.resolveFilter(filter, tableProvider);
- String segmentDir = CarbonTablePath.getSegmentPath(identifier.getTablePath(), "null");
+ String segmentDir = null;
+ if (carbonTable.isTransactionalTable()) {
+ segmentDir = CarbonTablePath.getSegmentPath(identifier.getTablePath(), "null");
+ } else {
+ segmentDir = identifier.getTablePath();
+ }
FileFactory.FileType fileType = FileFactory.getFileType(segmentDir);
if (FileFactory.isFileExist(segmentDir, fileType)) {
// if external table Segments are found, add it to the List
List<Segment> externalTableSegments = new ArrayList<Segment>();
- Segment seg = new Segment("null", null, readCommittedScope);
- externalTableSegments.add(seg);
+ Segment seg;
+ if (carbonTable.isTransactionalTable()) {
+ seg = new Segment("null", null, readCommittedScope);
--- End diff --
Prior Non Transactional implementation, SDK used to write into the Segment Path instead of Table Path. i.e. inside /Fact/Part0/Sement_null. It used to deliberately pass the segment as "null". Those codes to read and write into the segment path is still existing and they make it an external transactional table and the code flow goes through CarbonInputFormat.
---
Locked
