[GitHub] [carbondata] QiangCai commented on pull request #3988: [CARBONDATA-4037] Improve the table status and segment file writing

2021-01-19 Thread GitBox
QiangCai commented on pull request #3988: URL: https://github.com/apache/carbondata/pull/3988#issuecomment-763288420 @ajantha-bhat collect all index files from executor write task into driver side in memory, then write carbonindexmerge once. it can avoid writing index file in executor t

[GitHub] [carbondata] QiangCai commented on pull request #3988: [CARBONDATA-4037] Improve the table status and segment file writing

2021-01-18 Thread GitBox
QiangCai commented on pull request #3988: URL: https://github.com/apache/carbondata/pull/3988#issuecomment-762090072 how about skip carbonindex files step to generate carbonindexmerge files directly on driver side? This is a

[GitHub] [carbondata] QiangCai commented on pull request #3988: [CARBONDATA-4037] Improve the table status and segment file writing

2020-10-29 Thread GitBox
QiangCai commented on pull request #3988: URL: https://github.com/apache/carbondata/pull/3988#issuecomment-719134288 In the description, after it implements point 2, it should immediately delete the index file. So point 1 is no needed, right? --