Minor regression with 0.18.5 around handling of Double.NaN


(Jan Fiedler) #1

I upgraded to 0.18.5 (from 0.18.4) and my test cases show a minor
regression around Double.NaN handling. My test indexes a couple of
documents with different Double values (null and NaN among them). The test
that turned read is a matchAllQuery with a missingFilter. With 0.18.4 this
test only identified documents that really have 'null' as double value.
With 0.18.5 this test also returns documents that have Double.NaN. Is this
behavior change intended (couldn't find a matching issue in the release
notes)?


(Shay Banon) #2

It might relate to a change in Lucene and handling NaN numeric values, I
don't know the specifics exactly, need to check it out.

On Mon, Dec 5, 2011 at 11:35 AM, Jan Fiedler fiedler.jan@gmail.com wrote:

I upgraded to 0.18.5 (from 0.18.4) and my test cases show a minor
regression around Double.NaN handling. My test indexes a couple of
documents with different Double values (null and NaN among them). The test
that turned read is a matchAllQuery with a missingFilter. With 0.18.4 this
test only identified documents that really have 'null' as double value.
With 0.18.5 this test also returns documents that have Double.NaN. Is this
behavior change intended (couldn't find a matching issue in the release
notes)?


(system) #3