: So I have tested out the logic of stringField:[* TO *], with empty strings : in 7.5 and 8.8. They perform identically, returning empty string values, if : they exist in the index
I remember seeing something like this pop up before ... not sure where ... but i bet Houston's comment above is baed on a quick test against a StrField, and I bet clemensdev is talking about a query against a TextField. I think the issue is that when an Analyzer processes the input "" it produces no *terms* but the field still winds up with a norm value (based on term count of 0 fed to the similarity's norm() function) so the NormsFieldExistsQuery used in solr since SOLR-11746/8.5 matches a doc with a pre-analysis input of "", while the same doc won't match a PrefixQuery or RangeQuery that includes the literal term "" (because there are no terms to match the prefix/range) All of which begs the question: Should lucene be generating a norm when there are 0 terms? : documents with empty string values don't match. If I don't use that : processor chain, they do. : Would you check to see if there is a default updateRequestProcessorChain? : Those do not need to be specified in the requestHandler, so you won't see : them there.. : : Solr's default configSet comes with one, that includes the : RemoveBlankFieldUpdateProcessorFactory, by default unless you specify the : following java option: update.autoCreateFields=false : https://github.com/apache/solr/blob/main/solr/server/solr/configsets/_default/conf/solrconfig.xml#L1043 : : - Houston : : On Mon, May 24, 2021 at 1:19 PM clemens...@mysign.ch <clemens...@mysign.ch> : wrote: : : > Not explicitly, this being the requestHandler-section : : > <requestHandler name="/update" : > class="solr.UpdateRequestHandler"></requestHandler> : > : > : > : > On 2021/05/24 16:15:22, Houston Putman <h...@gmail.com> wrote: : > > So the backend implementation of the fieldName:[* TO *] query has : > changed> : > > as of 8.5 (SOLR-11746 <https://issues.apache.org/jira/browse/SOLR-11746 : > >).> : > > However I'm not sure that is resulting in the change here.> : > > : > > Are you using the RemoveBlankFieldUpdateProcessorFactory?> : > > : > > - Houston> : > > : > > On Mon, May 24, 2021 at 5:58 AM clemens...@mysign.ch <cl...@mysign.ch>> : > > wrote:> : > > : > > > I have a unit test that tests the "non-emptyness query". As of : > updating> : > > > from 7.5.5 to 8.8.2 this very test fails and returns a document that : > has an> : > > > empty string ( i.e. "" )> : > > > The "non-emptyness query" is:> : > > > fieldName:[* TO *]> : > > >> : > > > Has the default "behavior" of the above query changed?> : > > >> : > > > Does it possibly have anything to do with> : > > > RemoveBlankFieldUpdateProcessorFactory?> : > > >> : > > > Additional note: so far I haven't made any chages to the : > solrcofig.xml, so> : > > > I am still using the same configuration as for 7.5.5> : > > >> : > > > Regards> : > > > Clemens> : > > >> : > > : > : -Hoss http://www.lucidworks.com/