lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Uwe Schindler" <>
Subject RE: Enumerating NumericField using TermEnum?
Date Sun, 13 Sep 2009 20:51:59 GMT
> > If we do not need a new RC fort hat I can do it tomorrow! I am not yet
> sure
> > what to write: I tend to say: "Use NumericField, but with infinite
> > precisionStep for low-cardinality fields - and you get the old TermEnum
> > value list as before (with conversion through NumericUtils)". In
> general,
> > users should use NumericField for numbers, but use a appropinquate
> precStep,
> > so infinite if no faster RangeQueries are possible because of low
> > cardinality.
> Okay, good point. That makes sense - how about an example of low card,
> just for grounding? In the hundreds? Under 10,000? Only 10's?

The example of Phil is good: 24 for a hour fiel dis good. I would say,
everything upto 100 is low cardinality. It does not hurt to use small
precSteps, but you loose the possibility to enumerate the terms easily. The
typical example of a drop down list filled by a list terms in the web
interface is a typical example for low cardinality. E.g. credit card
expiration,... These are no real numeric values (although "hour" is a
number), they are used as list of preselected terms (and the list is
intelligently filled by the index). I have a lot of these, but they are
mostly country names, project names, and so on (or better said: facets). So:
low cardinality lists. If you use numbers this way, you can handle them as
simple text terms (and you will not use RangeQueries on them).

> Also, do you mean to use Integer.Max_VALUE as infinite?

Yes, sorry.

> My personal opinion is that we can make javadoc changes for the final
> without doing an RC, as long as no code/build/scipts at all is touched.
> Not sure how others feel though.

I just wanted to ask for confirmation.


To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message