The splitVector function is currently serving 3 different purposes:
- Provide users a way to determine bounds for parallelizing data processing, through the splitVector command
- Determine the initial split points
- Determine the auto-splitter's split points
Objective of this ticket is to extract the logic currently used by the chunk splitter - basically copy the splitVector function and "peel" the code not reached during auto splitting - into a different function to be backported with minimum effort.
This is a precondition for
SERVER-55028 aimed to improve chunk splitter's split points selection policy without impacting the user-facing splitVector behavior.