[CARBONDATA-3515] Limit local dictionary size to 10% of allowed blocklet size

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.6.1
• Assignee: Kumar Vishal
• Reporter: Ajantha Bhat
• Created: 10/Sep/19 05:16
• Updated: 09/Oct/19 14:55
• Resolved: 12/Sep/19 07:32

Description

problem: currently local dictionary max size is 2GB, because of this for varchar columns or long string columns, local dictionary can be of 2GB size. so, as local dictionary is stored in blocklet. blocklet size will exceed 2 GB, even though configured maximum blocklet size is 64MB. some places inter overflow happens during casting.

 

solution: limit the local dictionary size to 10% of maximum allowed blocklet size

Issue Links

• links to: GitHub Pull Request #3380