Hi Péter,
There has been a fair amount of discussion on this topic, and there are some
pending things being looked at with regards inclusive/exclusive combinations
etc.
For today, the easiest way to get the results you want to use a granular
time that allows you to specify a non-overlapping period:
For example, using milliseconds:
[2010-04-2710:00:00.001 TO 2010-04-2711:00:00.000]
[2010-04-2711:00:00.001 TO 2010-04-2712:00:00.000]
[2010-04-2713:00:00.001 TO 2010-04-2713:00:00.000]
You can use whatever date format you want, so long as the stored field has
sufficient granularity to define a non-overlapping period.
You could equally just do it with seconds:
[2010-04-27-10:00:01 TO 2010-04-27-11:00:00]
[2010-04-27-11:00:01 TO 2010-04-27-12:00:00]
[2010-04-27-12:00:01 TO 2010-04-27-13:00:00]
Peter
2010/4/28 Király Péter <pkir...@tesuji.eu>
> Grant Ingersoll said:
>
>> You should be able to do inclusive/exclusive ranges using the query parser
>> by
>> mixing matching brackets [] and braces {}.
>> See
>> http://lucene.apache.org/java/2_9_1/queryparsersyntax.html#Range%20Searches
>>
>
> Hi Grant,
>
> Thanks for your answer, but my problem is not that how to query, but the
> content
> of facet list. The list's numbers are only true if we use [date1 TO date2]
> in query.
> If we use {date1 TO date2} in query, the number of results will be wrong.
> What is the wrong with that? I should display dates in facet list as
> following:
>
> publication date
> 1000-1100 (3)
> 1100-1200 (3)
> 1200-1300 (12)
> ...
> as you can see here, the dates are duplicated, the 'to' date wil be the
> next 'from' date.
>
> However it would be more elegant to display something like that:
>
> publication date
> 1000-1099 (2)
> 1100-1199 (1)
> 1200-1299 (14)
>
> It would involve, that
> <lst name="facet_dates">
> <lst name="date_df">
>
> <int name="1000-01-01T00:00:00Z">3</int>
> <int name="1100-01-01T00:00:00Z">3</int>
> <int name="1200-01-01T00:00:00Z">12</int>
> ..
> should be understand, that between 1000-01-01T00:00:00Z
> and 1100-01-01T00:00:00Z minus 1 sec (or millisec) there are 3 results.
>
> Grant, from your answer it is clear, that currently no way for the date
> facets to
> behave like this, but I guess, that it may be an alternative solution.
>
> Péter
>
