Practically all sparse counters require knowledge of the counter distribution. This is derived by iterating the Solr doc->ordinal structures and collecting statistics in a temporary large counter structure (int[#unique values]). In the special case of a fully optimized segment, this information can be derived on a streaming basis, thus eliminating the need of the temporary large stat collector.
Practically all sparse counters require knowledge of the counter distribution. This is derived by iterating the Solr doc->ordinal structures and collecting statistics in a temporary large counter structure (int[#unique values]). In the special case of a fully optimized segment, this information can be derived on a streaming basis, thus eliminating the need of the temporary large stat collector.