Github user osma commented on the pull request:
https://github.com/apache/jena/pull/53#issuecomment-98649858
> Indeed, both of them will be dropped (however, can be solved with multi
language proposal)
I'm curious, how would the multi language proposal help with this problem?
> I don't think that storing the original literal in non-debug mode is a
good way to follow. It will result in a huge index and will affect performance.
> Instead of that, storing a smaller identifier generated from a hash of a
key (like uri+literal+language) will surely be a better solution.
I agree. A hash with sufficient uniqueness guarantees could work here. I
think it should incorporate all parts of the quad (subject, predicate, literal
value + language, named graph). This way, the entries in the Lucene index would
correspond exactly to quads. When a quad gets deleted, the corresponding entry
can be uniquely identified via the hash and deleted from the Lucene index as
well.
But I'm not sure how this would interact with the recent work done by Chris
Dollin (@ehedgehog) on cross field conjunctive queries (pull request #42 and
[JENA-686](https://issues.apache.org/jira/browse/JENA-686)). In that work, the
Lucene index is modified so that entries ("documents") in the Lucene index do
not correspond to triples/quads, but instead there is one entry per subject
resource (which may incorporate information from several triples/quads),
allowing more efficient conjunctive queries on the Lucene index level. How
would you deal with deletions in such a scenario?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
