RangeQuery without lower term and inclusive=false skips blank fields [LUCENE-38] about lucene HOT 5 CLOSED

Dejan Nenov (migrated from JIRA)

Added additional tests, using "null" as the lower term in the range query. The tests are commented to indicate how they should be modified to behave once this LUCENE-38 is fixed.

from lucene.

Mark Miller (@markrmiller) (migrated from JIRA)

Does this need to be 'fixed' ? RangeQuery now uses the semantics from ConstantScoreRangeQuery, which decided that open ended sides of a range must be inclusive (and are converted as such if not). Is that acceptable and we close this bug? Or jump a hoop or two for this rather niche case?

from lucene.

Otis Gospodnetic (@otisg) (migrated from JIRA)

This thing is 6+ years old and I don't recall this being mentioned on the list in the last half a decade. I'll leave you the Won't Fix pleasure, Mark.

from lucene.

Michael McCandless (@mikemccand) (migrated from JIRA)

Actually, this should have already worked, because RangeTermEnum forces includeLower to be true when lowerTermText is null.

But indeed the test still fails, so I dug into a bit and I think the test is faulty. The test expects the empty string doc ("") to be returned as a result, but the problem is the empty string doc when analyzed does not produce an empty string Token. So I modified the test (attached) to use an analyzer that emits empty string token, and then the test passes as expected.

I'll commit shortly.

from lucene.

Michael McCandless (@mikemccand) (migrated from JIRA)

Committed revision 713696.

from lucene.