Search before asking
Motivation
https://cwiki.apache.org/confluence/display/PAIMON/PIP-23+Introduce+bitmap+file+index
BitmapIndex tracks row positions of a specific value.
Let's say we have a primarykey table, a non-primary-key column: type. And we enable bitmap file index on the type column. For Spark, when DELETE FROM table WHERE type IN 'type1', it seems that it'll first scan parquet files to get FilePath and Position of rows, then shuffle by FilePath to update DV. I'm curious if we can directly use the bitmap index to update DV, skipping the file scan and shuffling.
Solution
No response
Anything else?
No response
Are you willing to submit a PR?