Thread: [DOCS] DECLARE syntax is different from the 2016 SQL standard

The following documentation comment has been logged on the website:

Page: https://www.postgresql.org/docs/9.1/static/sql-declare.html
Description:

Hi,

The 2016 standard uses a different syntax to declare a cursor:

<declare cursor> ::=
  DECLARE <cursor name> <cursor properties>
      FOR <cursor specification>

Notice that it does not require or permit the "CURSOR" keyword before the
"FOR" keyword.

I don't know why this was changed and unfortunately fixing this would break
existing implementations, but there should be a note about this under the
SQL Standard section.

This was discovered with: https://github.com/elliotchance/sqltest

BNF for 2016 can be found in the same project:
https://github.com/elliotchance/sqltest/blob/master/standards/2016/bnf.txt#L2371-L2372

elliotchance@gmail.com writes:
> The 2016 standard uses a different syntax to declare a cursor:

> <declare cursor> ::=
>   DECLARE <cursor name> <cursor properties>
>       FOR <cursor specification>

> Notice that it does not require or permit the "CURSOR" keyword before the
> "FOR" keyword.

For some reason, the spec puts the CURSOR keyword into <cursor
properties>, but it's definitely there:

<cursor properties> ::=
  [ <cursor sensitivity> ] [ <cursor scrollability> ] CURSOR
      [ <cursor holdability> ]
      [ <cursor returnability> ]

            regards, tom lane

On 14 March 2017 at 1:28:37 PM, Tom Lane (tgl@sss.pgh.pa.us) wrote:

elliotchance@gmail.com writes:
> The 2016 standard uses a different syntax to declare a cursor:

> <declare cursor> ::=
> DECLARE <cursor name> <cursor properties>
> FOR <cursor specification>

> Notice that it does not require or permit the &quot;CURSOR&quot; keyword before the
> &quot;FOR&quot; keyword.

For some reason, the spec puts the CURSOR keyword into <cursor
properties>, but it's definitely there:

<cursor properties> ::=
[ <cursor sensitivity> ] [ <cursor scrollability> ] CURSOR
[ <cursor holdability> ]
[ <cursor returnability> ]

regards, tom lane

Elliot Chance <elliotchance@gmail.com> writes:
> Although... it has bought up another possible bug with docs. The "Report an
> error" box (which I used to send the original message) seems to not be
> escaping HTML correctly... as seen in the previous message.

Yeah, our web team is aware of that ...

            regards, tom lane