Hello, We decided to upgrade solr to use Synonym graph filter, and this filter requires sow to be set to false. But after setting sow to false we’ve started to see some unexpected behaviour and after digging down the problem we’ve found out that mm results in wrong behaviour when setting sow to false and mm.autoRelax doesn’t work properly.
The tests were as follows: solr setup: Solr version: 6.6.0 and 6.6.1 Index: id,value doc1,“This is the first sentence” doc2,“This is the second sentence” doc3,“This is the third one” doc4,“This is the last one” Field type: <fieldType name="test_field" class="solr.TextField" positionIncrementGap="100"> <analyzer> <tokenizer class="solr.StandardTokenizerFactory"/> <filter class="solr.StopFilterFactory" ignoreCase="true" words="stopwords.txt" /> </analyzer> </fieldType> stopwords.txt file: This Is The random_word Tests: test 1: Query string: This one mm value: 2 mm.autoRelax: TRUE Expected result: doc1 Actual result: doc1 test 2: Query string: This one mm value: 2 mm.autoRelax: FALSE Expected result: nothing Actual result: doc1 test 3: Query string: This one mm value: 1 mm.autoRelax: TRUE Expected result: doc1 Actual result: doc1 test 4: Query string: This one mm value: 1 mm.autoRelax: FALSE Expected result: nothing Actual result: doc1 As we can see in the tests, mm.autoRelax isn’t working properly. It behaves like it’s always true in the previous cases. This behaviour was present in both solr 6.6.0 and 6.6.1 What can we do to keep using SynonymGraphFilter without introducing the above issues with autoRelax and mm?