> The title and long_description are almost the same. The > only difference is that I've taken out the > "solr.SnowballPorterFilterFactory" filter, so that the > titles are not stemmed. I actually as part of the test > before writing to this list I tried creating the type that > has only the standard tokenizer defined - and the results > were the same.
Hmm. That difference might causing it. Because parsed query will different for title and long_description field due to stemmer. I don't know if the query is re-parsed using long_description's analyzer during highlighting. What happens if you query this: http://localhost:8983/solr/select?q=long_description:terminator^0 title:terminator&fl=sku,title,long_description&hl=true&hl.fl=long_description with and without &hl.usePhraseHighlighter=true&hl.requireFieldMatch=true > In regards to the caching... I cleaned out the index and > now I it's containing just those 2 documents (london and > terminator). I can see the changes straight away if I change > and re-index - so I think that should be fine. It's really > puzzling that it works for one doc but not for the other. I mean solr's Http caching. When testing from browser I usually disable it with <httpCaching never304="true"> in solrconfig.xml.