TAPNotes & Mistake in ADQL grammar ?

[Markus Demleitner]

Dear DAL,

On Wed, Oct 01, 2014 at 06:47:25PM +0200, Grégory Mantelet wrote:
>     1/ In section 2.2.5 of TAPNotes, the given grammar contains an
> item called "<query primary>" which is referenced nowhere in the
> given new rules and in the current ADQL grammar. I would say it is a
> copy-paste error, because the content of this rule is exactly the
> same as the new "<query_expression>". So, let's say it is a very

As <query primary> is not referenced anywhere, it is indeed 
superfluous in this grammar.  Going back to sql1992, I notice I
really messed up the rule above, fortunately in a way so gross it'd
have been obvious in implementation right away:

     <non_join_query_term> ::=
              [...] <query term> INTERSECT [ ALL ]

is of course missing its second operand.  In sql1992, this second
operand is the <query_primary>.  As Grégeory correctly observes, in
ADQL <query_primary> and <query_expression> have the same expansions.
So, I'm fixing the above rule to 

              [...] <query term> INTERSECT [ ALL ] <query_expression>

in SVN.

>     2/ However, while looking a such rule in the current grammar
> (ADQL 2.0 standard document), I discovered that it was valid,
> according to the given BNF, to have a query limited to a table name
> or a table join. Indeed, it is possible due to the following rules:
> 
>  <query_expression> ::=  <query_specification>
>                                          | <joined_table>
> 
> So, a query can be either a "SELECT .... FROM ..." or just a table
> name (or a join between 2 tables), according to the following rules:

Here, I disagree with Grégory -- unless I'm completely confused here,
the root production of ADQL is query_specification, not
query_expression, no?

Cheers,

         Markus