Filter by publisher seems to be broken

[jthomasmock]

Howdy! I am noticing that this is closed this as done:
#1447
However, I don’t see it working on the web, even for the exact search term example with publisher:vscode. Is it not really deployed yet?

Per @netomi the change is present in the current deployment, however the implementation seems to have a flaw.

It uses the term (https://www.elastic.co/guide/en/elasticsearch/reference/7.17/query-dsl-term-query.html) query for namespaces

However, elastic search transforms the namespace name during analysis

e.g. ms-vscode becomes ms and vscode

Thus the search for publisher:vscode will match for ms-vscode as well

this is a bug and should be fixed.

the fix will most likely be to change

boolQuery.must(QueryBuilders.term(builder -> builder.field("namespace").value(options.namespace()).caseInsensitive(true)));
##> to
boolQuery.must(QueryBuilders.term(builder -> builder.field("namespace.raw").value(options.namespace()).caseInsensitive(true)));

[VijayabaskarR-06]

The publisher: filter uses an Elasticsearch term query on an analyzed field, which causes namespaces like ms-vscode to be tokenized into ms and vscode. As a result, publisher:vscode incorrectly matches ms-vscode.

A likely fix would be to query a non-analyzed field (e.g. publisher.keyword) or adjust the mapping so publisher/namespace is indexed as a keyword for exact matching.

I’d be happy to work on this and submit a fix with a regression test if that sounds reasonable. Could you please assign this issue to me?

[netomi]

@VijayabaskarR-06 happy to review a PR, the fix should be rather straight-forward as you mention.

[VijayabaskarR-06]

Thanks for confirming! I’ll work on this fix by switching the query to a non-analyzed field and adding a regression test to ensure exact publisher matching.

Could you please assign the issue to me?