consider removing the notion of "data version"
Ladicek opened this issue · comments
IndexReader and its related classes expose 2 notions of version: index version (version of the persistent format) and data version (version of information stored in the persistent format). The original idea was (probably) that these 2 would increment in parallel when new data were added to the persistent format, but the data version would stay intact if only the persistent format changed and no new information is added (i.e., persistent format refactoring).
I don't think this has ever been used. More importantly, IndexReaderV2 still reports data version as 4, even though the index version was incremented multiple times.
Done in #210.