Hi Matt,
To verify I understand correctly, are this your settings? :
- one "MAIN" index containing all the data:
used for search;
never does addDocument();
- Several side "INC" indexes:
addDocument() here for new/modified documents;
never searched;
- at some point all INC indexes are "merged" to MAIN index:
for each INC: {
(a) delete INC's corresponding docs from MAIN;
(b) MAIN.addIndexes({INC});
}
It would be interesting to know why you selected this setting - what are
the benefits from not addDocument()ing on MAIN? For instance, are the INC
indexes maintained on separate drives, and you get better performance this
way?
But ignoring the "why" and looking at the "how" - if you are using
IndexReader for deletion in (a), then yes, you would have to close the
IndexWriter, because otherwise the IndexReader would fail to delete -
although called "Reader", it has to get write privilege to the index (using
Lucene's internal locking).
As of Lucene 2.1 though, IndexWriter supports delete document(s) by term,
which I assume is what you are doing, so you could use it, and avoid the
second IndexReader on MAIN. If each INC index have different documents (by
your definition / application_doc_id) then you could first delete all the
documents and then add all the indexes in one call to addIndexes (or
addIndexesNoOptimize).
Hope this helps,
Doron
--------------------------------
"DECAFFMEYER MATHIEU" <[EMAIL PROTECTED]> wrote on 13/03/2007
06:51:35:
> Hi,
> I need to merge several indexes (I call them incremental index) with
> my main index.
> Each incremental index can contain the same url's of the main index,
> that's why I have a list of url's to update, that I will delete from
> the main index before merging with an incremental index.
> I have also a list of urls to delete,
> Then after these operations I can merge the indexes.
> I am not sure about the operations Open / Close on the Index Writer,
> I may need some help ...
>
> for i=0 ; i < list_increment_indexes.size() ; ++i
> { // BEGIN WHILE
> 1. OPEN A SECOND READER ON THE MAININDEX
> (for not interrupting user queries on
> the main indexreader on the main index)
> 2. DELETE DOCS TO BE DELETED
> DELETE DOCS TO BE UPDATED
> 3. CLOSE THE SECOND READER ON THE MAIN INDEX ==> FLUSH DELETIONS !!
> 4. RE-OPEN A SECOND READER ON THE MAIN INDEX FOR MERGING
> 5. OPEN A READER ON THE INCREMENTAL INDEX
> 6. OPEN AN INDEXWRITER
> TO MERGE THE READER ON THE INCREMENTAL INDEX
> WITH THE READER ON THE MAIN INDEX
> 7. CLOSE THE INDEXWRITER
> 8. CLOSE THE SECOND READER ON THE MAIN INDEX
> 9. CLOSE THE READER ON THE INCREMENTAL INDEX
> } // END WHILE !
> 10. CLOSE THE INDEXSEARCHER
> 11. CLOSE THE MAIN INDEXREADER ON THE MAIN INDEX
> 12. OPEN A NEW READER ON THE INDEX
>
> I am not sure if I can open the idnexwriter *before* the loop and
> close it *outside*,
> So I do it in the loop for each incremental index for now on ...
>
> __________________________________
> Matt
>
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
