Re: BUG #14275: cursor's variable in pgsql doesn't respect scope

On Thu, Aug 4, 2016 at 10:10 AM, Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com>
wrote:

>
>
> 2016-08-04 15:03 GMT+02:00 klimych(at)tut(dot)by <klimych(at)tut(dot)by>:
>
>> Thank you!
>> Sorry, I should read the documentation more carefully (athough I didnt't
>> expect to find explanation of such behaviour in "Returning Cursors"
>> section, as well as I didn't belive such behaviour could be made on
>> purpose).
>> And thanks again for workaround! It seems to be the only way to use
>> cursors in pl/pgsql (the weirdest thing I've ever seen, I should say)
>>
>
> I agree so this is little bit strange - it looks like workaround of some
> historical limit of SPI. It is too late to change. But it has some
> advantage. Postgres can't to pass parameters by ref. With named cursors you
> can do it.
>
>
​The docs could maybe be improved, though it is documented and being
mis-informed simply results in an error and a question on the lists, so
expending get mental effort here isn't that appealing.

Improving the code would involve something like:

OPEN unbound_cursorvar [ [ NO ] SCROLL ] [ NAME system_name ] FOR query

Also, would a hint on the error message be too much to ask?

HINT: by default the global name of the cursor is equal to the variable
name to which it is assigned

David J.

​