Upgrade to lucene-6.5.0-snapshot-f919485.#23087
Conversation
| @@ -172,6 +172,7 @@ public void testPhrasePrefix() throws ExecutionException, InterruptedException { | |||
| assertSearchHits(searchResponse, "1", "2", "3", "7", "8"); | |||
| } | |||
|
|
|||
There was a problem hiding this comment.
Regarding the failing tests, this is due to a change in Lucene 6.5.0. In 6.4 Queries with multi word synonyms create a GraphQuery with all possible paths. So if you have "foo, foo bar" and a query like "foo boo" it will produce two boolean queries packed in a GraphQuery:
foo boo OR foo bar boo
Since we have all paths we can apply the cutoff frequency to each but in 6.5 the query parser will now create a query:
(+foo OR (+foo +bar)) AND boo
So the multi synonyms is now treated as a single term. This is better in terms of performance but it breaks some tests because the logic has changed. We cannot apply cutoff frequency on multi terms synonyms anymore (they are treated as a single term). Same problem with minimum_should_match.
IMO we should just rewrite this test and remove the multi terms synonym.
There was a problem hiding this comment.
However I'll leave it for a follow-up PR if you don't mind.
There was a problem hiding this comment.
Sure, I can take over when you merged this PR. Thanks
| @@ -0,0 +1,101 @@ | |||
| /* | |||
There was a problem hiding this comment.
I think this deserves a separate PR, WDYT ? We need to document how this works and make sure that we add all the warnings regarding the restrictions of using this filter in conjunction with others.
There was a problem hiding this comment.
I can do that. I added it because the test that checks whether all analysis components are exposed failed otherwise.
There was a problem hiding this comment.
Ok I created #23104 to track the inclusion of this new filter. We can add the documentation and tests in a follow-up
* master: Fix alias HEAD requests Upgrade to lucene-6.5.0-snapshot-f919485. (elastic#23087) Add BulkProcessor methods with XContentType parameter (elastic#23078)
This is an in-progress PR as I have a failure in MatchQueryIT that looks related to recent changes around graph token streams. If someone could help me dig it, I would appreciate.