[LUCENE-10371] Make IndexRearranger able to arrange segment into a determined order

• Type: Improvement
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 9.1
• Lucene Fields: New
• Assignee: Unassigned
• Reporter: Patrick Zhai
• Created: 11/Jan/22 05:56
• Updated: 28/Aug/22 16:34
• Resolved: 15/Feb/22 21:56

Description

Previously when I tried to make change to luceneutil to let it use IndexRearranger for a faster deterministic index construction, I found that even each segment contains the same documents set, the order of segments will impact the estimated hit number (using BMW): https://markmail.org/message/zl6zsqvbg7nwfq6w

At that time the discussion tend to tolerant the small hit count difference to resolve the issue, after some time when I discuss this issue again with mikemccand , we thought it might also be a good idea to just add ability of rearranging the segments order to IndexRearranger, so that we can ensure each time the rearranged index is truly the same.

Issue Links

• links to: GitHub Pull Request #630
• links to: GitHub Pull Request #683