Hi Uwe, Apologies for the late reply.
I suggested to use this reader to some customers, but they were using Solr > or Elasticsearch and it's not easy to implement it there. And they didn't > want to pay the expensive Uwe. > 😜 > Oh, they should know better - Mr Uwe is worth his price tag for technical merits but also his eloquence and samurai-sword-sharp code smell observation skills! > How do you handle deletes. Because the main issue with those readers is > that you can't update documents without also updating the main reader > (although it's a fake update). > It's not a general search/ document storage system - we allow updates but they're not real time and reindexing creates both new segments and second-tier segments. It gets even more complicated because there are multiple living Lucene commits (custom deletion policy). Long story but works just fine! > If this is used, have you thought of a SynchronizedMergePolicy that just > applies the same merges in the secondary index? > I don't think it's applicable to us (as above), but it's an interesting idea! D. >
