hi,

This is regarding a statement in the 'Hibernate search in action' book -

According to my understanding, sorting is done on all the matching documents in Lucene before pagination is applied. Suppose have 1000 records in the db and 1000 corresponding documents in the Lucene index. Sorting by one of the fields is requested and also pagination to retrieve the first 10 records. Sorting will be done in Lucene for all the 1000 records before pagination is to fetch the 10 records. Then how can the above statement be correct? Or does it mean something else?

Thanks,
Seema

Good question.
I'm not sure how Lucene is doing this internally, but it is capable of sorting them without going throught the most expensive step which is extracting the term values from the index to produce the result.
Said otherwise, it's capable of doing a "blind sort" sorting the result without yet knowing what's going to be returned... you can experiment using projection (to disable entity loading from database) combined with sorting and pagination: you'll see it's faster when you have to return only 10 elements compared to 10000, even when both results are sorted the same way.
Make sure to have different orders of magnitude in the sizes of the testcases, as more stuff is happening behind the curtains like batch extraction.