Hi all,
I index a large number of documents, nearly 800k of one specific class right now. My index is taking up 85M and I'd love to get this down if possible. There is a lot of redundancy in the data being indexed per document, ie each object being indexed has a many-to-one relationship to a child object, and there are only a few thousand of these child objects, but most of the data in each document is actually held in the child object. So my question is: is it possible to have in effect a many-to-one relationship in indexing somehow, to help eliminate this redundancy?

Hi,
the index world is not relational, so the quick answer is no.
However you shouldn't worry much about it because the content is not stored in the index, the strings are tokenized and each token is stored as a term. These terms are unique: so if each of your documents where written by the same author named "alcyon", "alcyon" is not going to take more disk space if it's repeated one of 1000 times. there are just some additional "pointers" from the document to the relevant term, but you'd have the same stuff in a relational world.

So in short even if that would be possible, it wouldn't make your index noticeabily smaller.

Ah gotcha, good to know. Thanks!