Thanks Jack and Shawn. I checked these Jira tickets, but I am not sure if the slowness of MatchAllDocsQuery is also caused by the removal of fieldcache. Can someone please explain a little bit?
Thanks, Wei On Fri, Nov 6, 2015 at 7:15 AM, Shawn Heisey <apa...@elyograg.org> wrote: > On 11/5/2015 10:25 PM, Jack Krupansky wrote: > > I vaguely recall some discussion concerning removal of the field cache in > > Lucene. > > The FieldCache wasn't exactly *removed* ... it's more like it was > renamed, improved, and sort of hidden in a miscellaneous package. Some > things still require this functionality, so they use the hidden class > instead, which was changed to use the DocValues API. > > https://issues.apache.org/jira/browse/LUCENE-5666 > > I am not qualified to discuss LUCENE-5666 beyond what I wrote in the > paragraph above, and it's possible that some of what I said is wrong > because I do not really understand the APIs involved. > > The change has caused problems for Solr. End result from Solr's > perspective: Certain things which used to work perfectly fine (mostly > facets and grouping) in Solr 4.x have one of two problems in 5.x: > Either they don't work at all, or performance has gone way down. Some > of these problems are documented in Jira. These are the issues I know > about: > > https://issues.apache.org/jira/browse/SOLR-8088 > https://issues.apache.org/jira/browse/SOLR-7495 > https://issues.apache.org/jira/browse/SOLR-8096 > > For fields where adding docValues is a viable option (most field types > other than solr.TextField), adding docValues and reindexing is very > likely to solve those problems. > > Sometimes adding docValues won't work, either because the field type > doesn't allow it, or because it's the indexed terms that are needed, not > the original field value. For those situations, there is currently no > solution. > > Thanks, > Shawn > >