Hello,
as posted some time ago I'm working on a native, versioned XML-DBMS [1].
I'd like to provide a full text index and I recently read about
customized Codecs which can be plugged in. Usually data (for instance
XML nodes) are stored on RecordPages. I'm still not sure if it is
possible and makes sense to implement PostingsFormat and possibly Directory.
What I want to achieve is to be able to use my infrastructure for
transaction-safe versioning. That is I need some kind of record for the
different types (I think fields, terms, documents and term positions)
with a simple record-ID to retrieve the record from disk and which kind
the record is. Furthermore all I need is a serialization/deserialization
mechanism for each record type. Probably I can simply reuse the default
serialization/deserialization routine. I'm furthermore not sure if it
would be nice to provide a B+-tree implementation which always clusters
for instance the fields, the terms, then the documents and the term
positions. I don't know what index structure Lucene uses per default,
but I think it must be something which is performant with any kind of
disks (reading/writing blocks of data).
Any hints and suggestions would be nice.
kind regards,
Johannes
[1] https://github.com/JohannesLichtenberger/sirix
---------------------------------------------------------------------
To unsubscribe, e-mail: java-user-unsubscr...@lucene.apache.org
For additional commands, e-mail: java-user-h...@lucene.apache.org
