Manish Gupta created CARBONDATA-2039:
----------------------------------------
Summary: Add relative blocklet id during initialization in the blocklet data map
Key: CARBONDATA-2039
URL:
https://issues.apache.org/jira/browse/CARBONDATA-2039 Project: CarbonData
Issue Type: Bug
Reporter: Manish Gupta
Assignee: Manish Gupta
h5. Add relative blocklet id during initialization in the blocklet data map.
Problem
Currently while loading the blocklets in data map all the blocklets are stored in the unsafe manager on array indexes for all the blocklets in one segment. So lets say if 1 segment has 7 task and each task has 10 part files with 3 blocklets each, total number of blocklets in the segment would be 210. Unsafe store will store all these blocklets in one array with start index as 0 and end index as 210.
Due to this while filling the blocklet information the blocklet id taken is the array index which can be any number from 0 to 210. This is leading to loss in actual mapping of blocklet with respect to carbondata file.
Solution
Add the relative blocklet id during loading of blocklets in the unsafe store (relative blocklet id is the id of blocklet in the carbondata file)
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
Locked
