Full reindex needed if termVectors added to fields in schema?

[Eric Kilby <ki...@stylefeeder.com>]

hi,

I've successfully added fields to my schema.xml before, and been able to
incrementally keep indexing documents with just the new ones picking up the
fields.  This appears to be similar to the case of not including certain
fields in certain documents, as the other documents simply don't have them
until they're added.

I'm looking into testing a MoreLikeThis implementation, and have read on
here that termVectors are needed to make it run acceptably.  I'd like to
rebuild my index, but that will take some time given the number of documents
involved, and I'd like to keep incremental updates running at the same time. 
The constraint is on the database side not the SOLR indexing side, so
improvements to indexing performance aren't my main concern here.  

So, my question is whether adding termVectors="true" to a couple of schema
fields will work similarly to adding new fields, where the updated documents
will get the vectors added and the others won't get them but will continue
to work, allowing me to rebuild "in the background" while not breaking
anything in my existing incremental update/release cycle.

I appreciate your help.

Eric Kilby

-- 
View this message in context: http://www.nabble.com/Full-reindex-needed-if-termVectors-added-to-fields-in-schema--tp21081315p21081315.html
Sent from the Solr - User mailing list archive at Nabble.com.

Re: Full reindex needed if termVectors added to fields in schema?

[Chris Hostetter <ho...@fucit.org>]

: So, my question is whether adding termVectors="true" to a couple of schema
: fields will work similarly to adding new fields, where the updated documents
: will get the vectors added and the others won't get them but will continue
: to work, allowing me to rebuild "in the background" while not breaking
: anything in my existing incremental update/release cycle.

I beielve it should work ... if the document you are generating a MLT 
query for was indexed before the termVectors="true" option was added, the 
MLT class should use it's fallback behavior of reanalizing the sored text 
... if it was reindexed after that, then the termVectors will be used ... 
i think.



-Hoss