Re: Improving EXPLAIN's display of SubPlan nodes

Dean Rasheed <dean(dot)a(dot)rasheed(at)gmail(dot)com> writes:
> The get_rule_expr() code could perhaps be simplified a bit, getting
> rid of the show_subplan_name variable and moving the recursive calls
> to get_rule_expr() to after the switch statement -- if testexpr is
> non-NULL, print it, else print the subplan name probably works for all
> subplan types.

Oooh, good idea. The symmetry wasn't apparent when we started, but
it's there now, and the code does look nicer this way.

> The "colN" notation has grown on me, especially when you look at
> examples like those in partition_prune.out with a mix of Param types.

OK, I've left it like that in the attached v5, but I'm still open
to other opinions.

>> The undecorated reference to (SubPlan 1) is fairly confusing, since
>> it doesn't correspond to anything that will actually get output.
>> I suggest that perhaps instead this should read
>> Output: (SubPlan 1).col1, (SubPlan 1).col2, IGNORE(SubPlan 1), i.tableoid, i.ctid
>> or
>> Output: (SubPlan 1).col1, (SubPlan 1).col2, RESET(SubPlan 1), i.tableoid, i.ctid

> I think "RESET()" or "RESCAN()" or something like that is better than
> "INGORE()", because it indicates that it is actually doing something.
> I don't really have a better idea. Perhaps not all uppercase though,
> since that seems to go against the rest of the EXPLAIN output.

Hm. I used "rescan(SubPlan)" in the attached, but it still looks
a bit odd to my eye.

I did some more work on the documentation too, to show the difference
between hashed and not-hashed subplans. I feel like we're pretty
close here, with the possible exception of how to show MULTIEXPR.

regards, tom lane