We are using grouping in a distributed environment, and we have noticed a discrepancy:
On a single core with a group.limit > 1 and group.main=true, setting rows=10 will return 10 documents. A distributed setup with the same parameters will return 10 groups. We plan to open a jira ticket and submit a fix, but there is the question of which way to fix it. In the case where group.main is not set, the group.limit applies to the number of groups for both single and multi core cases, so that approach would be consistent. However, it seems to us that a user requesting the group.main results format will likely expect the group.limit to apply to the number of documents. A discussion held around an older fix a couple of years ago supports this view. (https://issues.apache.org/jira/browse/SOLR-2063) Unless there is a good case for the first approach, we plan to go with the second; I wanted to put this out to see if we're overlooking something - or if this was implemented in the way for some reason - feedback? Monica Skidmore Search Application Services Engineering Lead CareerBuilder.com
