Controll merging segments in Lucene

Question

I want to have a control in Lucene on the segment merging. To be more precise I have a per segment cache and each time there is a merge I want to have a "callback" and check which segment is going to be merged.

I had a look to the IndexWriter code and looks like its an internal operation of the index writer. I also dont want to hack the IndexWriter. I see an interface called MergeScheduler. Is the best idea to implement this class?

Do you have any idea how to do that? Thanks in advance!

Answer 1

Maybe you could subclass the MergePolicy you are using (the default one is TieredMergePolicy) and intercept the segments to be merged by overriding the findMerges method like this?

MergePolicy.MergeSpecification mergeSpecification = super.findMerges(segmentInfos);

// use information from mergeSpecifiation

return mergeSpecification;