John McKown wrote
>> insert into sales values
>> (tstzrange('2014-1-1', '2014-1-2')),
>> (tstzrange('2014-1-2', '2014-1-3')),
>> (tstzrange('2014-1-2', '2014-1-4')),
>> (tstzrange('2014-1-5', '2014-1-6'));
>>
>> -- want back:
>> -- tstzrange('2014-1-1', '2014-1-4')
>> -- tstzrange('2014-1-6', '2014-1-6')
>>
I presume the second output row should be [5,6)...
And why are you using a timestamp range when your data are dates?
My first thought is to explode the ranges into distinct dates, order them
inside a window, use lag(...) to find breaks,p and assign groups, the for
each group take the min and max of the group and form a new range. Not sure
exactly what the SQL looks like - especially the range explosion - but
should technically work even though performance may suck. Probably want to
use lateral and generate_series(...) if you are on a more recent version.
David J.
--
View this message in context: http://postgresql.1045698.n5.nabble.com/Finding-date-intersections-tp5824102p5824194.html
Sent from the PostgreSQL - general mailing list archive at Nabble.com.