Should any of the field specific searches be exact? about warclight HOT 4 CLOSED

The problem is that for user-issued keyword search you want the field to be tokenized into bits and normalised quite a bit. That makes it hard (read: I don't know how) to ensure that your query is a full match. If we have document A "foo/bar.zoo?query=Hamburger+City", a user-query oriented indexing would result in ["foo", "bar", "zoo", "query", "hamburger", "city"] or something like that. When you make a machine-oriented search (click on a facet or such) for "foo/bar.zoo?query=Hamburger", it will also match document A, as Solr does not care about the extra token "city" in the document when matching.

The easy answer is to increase the index size and make a parallel version of the field where the indexing is verbatim. StrField works well for that and has the added bonus that docValues works for that field. Do note that this really means verbatim searches: "foo/bar.zoo?query=hamburger+city" will not match "foo/bar.zoo?query=Hamburger+City". That is normally not a problem if the field is intended for machine-queries.

This method is used in my latest pull-request to webarchive-discovery-code for the fields url_norm and url_search, where one field is an indexed & docValues StrField and the other is a heavily tokenized & normalised TextField.

from warclight.

URL as exact and rest as fuzzy makes sense to me!

from warclight.

@tokee what's the preferred method to do an exact match? Something like Arclight has setup here? Or, something else?

from warclight.

Resolved with 3e034d2

from warclight.