Once you have played with some examples, at some point, you will probably be curious about the possibility to view results, grouped by a certain field value, in a similar way to what is possible with SQL.
From Solr's point of view, this functionality is intended to be used in two scenarios:
- Field collapsing: It is used when we want to collapse a group of results into a certain amount of entries, aggregating them by a common field value. This is a typical case for search engines or internal engines on large datasets, and it is generally useful to provide only the top results for some value choices to give an idea on what the users should expect with some link to the complete series of results. This is somewhat similar (from a different point of view) to what we have done with the
wordcloudsprototype; so, this approach will be useful when you have lots of shared values, and it's not complex to use some HTML or JavaScript to produce the query for the complete results. - Result grouping: In this case, we are grouping documents with a common field value and show only at the top of the results. This is widely used in the e-commerce context for providing the most viewed or bought products. On a news or OPAC site, it would show the most recent or more interesting resource.
You are certainly not missing some analogies with facets, but the main difference here is that we are playing with actual data, not with counts of matching documents. Another very interesting aspect that you won't miss is relative to what kind of relevancy function is used to produce the top elements for a group or even the group itself. (We could even construct a group for documents that return the same value when applying a certain function.) Going deeper with these functions can force us to use vectors and similar types but, at the moment, you are probably more interested in having an idea of which parameters can be used while grouping results. Refer to the following link for more info:
http://wiki.apache.org/solr/FieldCollapsingsolr
There is another related resource at the LucidWorks documentation site: