Re: pgsql/src/backend/nodes (copyfuncs.c outfuncs.c print.c)

Tom Lane wrote:

> Hiroshi Inoue <Inoue(at)tpf(dot)co(dot)jp> writes:
> > How would be the behavior of the following command sequence ?
>
> > begin;
> > declare myc cursor for select * from t1 limit 1;
> > fetch in myc;
> > fetch in myc;
>
> > Could the last fetch return a row ?
>
> As the code now stands, the second fetch would return nothing.
> I think this is clearly what any reasonable person would expect
> given the LIMIT 1 clause.
>

Different from ordinary select statements we could
gain the same result in case of cursors.

begin;
declare myc cursor for select * from t1;
fetch in myc;

For exaple,

begin;
declare myc cursor for select * from t1 limit all;
fetch 20 in myc; (the first page)
...(interaction)
fetch 20 in myc; (the next page)
..(interaction)
fetch backward 20 in myc; (the previous page)
...

What I expect here is to get rows of each page in
an average response time not the total throughput
of db operation.

Regards, Hiroshi Inoue

>
> LIMIT ALL is a different story, because there's no semantic difference
> between writing LIMIT ALL and writing no limit clause at all. We have
> the option to create a distinction for planning purposes, however.
> Question is do we need one?
>
> regards, tom lane