Re: [SQL] Order by - Mailing list pgsql-sql

Herouth Maoz wrote:

>
> At 13:45 +0300 on 02/06/1999, Jan Wieck wrote:
>
>
> >     That's all right and I'm not going to allow this just for the
> >     ORDER BY.  It will be enabled  as  kind  of  a  side  effect.
> >     Because   view's   can   have   a   GROUP   BY   clause,  the
> >     rewriter/planner  must  be  able  to  produce   plans   where
> >     different  left-/righttrees have their own grouping. Grouping
> >     requires  a  sort  and  thus  sorting  of  subplans  must  be
> >     supported.
> >
> >     Why should we deny ORDER BY on views explicitly?
>
> The SQL92 standard, if I read it right, defines the view by a "query
> expression", which in its turn relies on "query specification", which in
> turn relies on "table expression", which does not include ORDER BY.
>
> In short, ORDER BY is not part of the query syntax used in the definition
> of views, because order is meaningless in views.

    Right now, the parser accepts ORDER BY in CREATE VIEW because
    the grammar includes the full SelectStmt rule for the  view's
    query  expression.   The LIMIT, DISTINCT and ORDER BY clauses
    are rejected later in rewriteDefine.c.

>
> The basic idea is that we should get the same results if we use standard
> SQL whether we run on PostgreSQL or on Oracle. So, one may write an
> application which selects the data from the view, and relies on the order
> in the view for something (like grouping). The same program won't work in
> any other database system, because you simply can't make that assumption
> without adding 'ORDER BY' to your actual query.

    GROUP BY must be  possible  in  a  view's  query  expression.
    Without  that  views  with aggregate columns could never work
    properly.

    If I now read your above statements right, "query expression"
    ->  "query specification" -> "table expression" also forbid's
    GROUP BY, because it cannot occur in  a  table  expression  -
    right?  That  means  that  a  fully SQL92 compliant DB cannot
    support aggregates in view definitions?

    Oh what a perfect world - I'm stressing my brain  to  find  a
    way  how  they  could work and the real solution is to forbid
    them :-)


Jan

--

#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me.                                  #
#======================================== jwieck@debis.com (Jan Wieck) #