ADQL evolution: OFFSET?

[Markus Demleitner]

Hi,

On Thu, Jan 21, 2016 at 03:25:05PM +0000, Mark Taylor wrote:
> On Thu, 21 Jan 2016, Walter Landry wrote:
> 
> > Markus Demleitner <msdemlei at ari.uni-heidelberg.de> wrote:
> > > The result would be, for ADQL,
> > > 
> > > <order_by_clause> ::= 
> > >   ORDER BY <sort_specification_list> <result offset clause>
> > > <result offset clause> ::=
> > >   OFFSET <unsigned integer>
> > > 
> > > -- which is a relevant simplification, in particular restricting
> > > ourselves to unsigned integers, but I propose that's a good idea.
> > > Dropping the ROW[S], however, makes this not even a proper subset of
> > > SQL 2008.  So, if anyone stands up and says "let's keep the ROW[S]",
> > > I'll have no choice but to concur.
> > 
> > Let's keep the ROW[S].  We should not make it gratuitously different
> > from SQL.
> 
> Yeah.  Instead of the reek of COBOL, try thinking of it as retro chic.

That certainly makes it more palatable, yes.  So, it would be

  <order_by_clause> ::= 
    ORDER BY <sort_specification_list> [ <result offset clause> ]
  <result offset clause> ::=
    OFFSET <unsigned integer> (ROW|ROWS)

(note: original proposal was missing the square brackets).

Dave -- do you maintain the ADQL grammar?  Where would that go?

Cheers,

        Markus