Re: [PROPOSAL] Temporal query processing with range types

2017-02-27 11:21 GMT+01:00 Peter Moser <pitiz29a(at)gmail(dot)com>:
> 2017-02-24 21:25 GMT+01:00 Jim Nasby <Jim(dot)Nasby(at)bluetreble(dot)com>:
>> It's not that hard to construct a range type on-the-fly from 2 columns, so
>> (without having looked at the patch or really followed the thread) I would
>> think the answer is yes.

Thank you for your suggestion.

> We discussed and decided to remove the syntax for separate columns.

Please find attached the new patch with "range-type-only" syntax. It
is around 400 lines of
code shorter than its predecessor.

Now the syntax is as follows:

SELECT * FROM ( r ALIGN s ON q WITH (r.time, s.time) ) c;
SELECT * FROM ( r NORMALIZE s ON q WITH (r.time, s.time) ) c;
SELECT * FROM ( r NORMALIZE s USING(atts) WITH (r.time, s.time) ) c;

...where r and s are relations, c an alias, q any boolean expression,
atts a column name list, and r.time/s.time range typed columns.

This means that the syntax with four columns (i.e., scalar time point
start/end for relations r and s)
inside the WITH-clause is no longer supported.

Best regards,
Anton, Michael, Johann, Peter