Hi,
We modified Lucene 2.4.1 sources to add a MergePolicy option to insert empty documents in places where a document was deleted during the merge phase. The motivation was to retain the same document locations (docID) when there are deletions, so an external tracking database can persist the docIDs and not be concerned about the IDs becoming invalid upon a merge. This option in combination with a "no-optimize" allows an index to be tolerant of a small number of deletions without requiring another parallel immutable ID space. While it is possible to create a field called ID within each document which is immutable, large retrievals require a relatively expensive search of the IDs, which an immutable ID space will avoid. This also helps in constructing a filter bit map to a search, where the bit map was created using other business logic. Our implementation has added a few unit tests to confirm its correct operation. Is there an interest in pursuing this as a useful capability? Thanks, Venkat Rangan Clearwell Systems Inc. (650) 526 0639 http://www.clearwellsystems.com <http://www.clearwellsystems.com/> - Delivering Intelligent eDiscovery
