Re: Does anyone ever use OPTIMIZER_DEBUG?

On Fri, 29 Sept 2023 at 10:59, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> We could also discuss keeping the "tracing" aspect of it, but
> replacing debug_print_rel with pprint(rel), which'd still allow
> removal of all the "DEBUG SUPPORT" stuff at the bottom of allpaths.c.
> That's pretty much all of the maintenance-requiring stuff in it.

To assist discussion, I've attached a patch for that.

I likely can't contribute much to that discussion due to being more of
an "attach a debugger" person rather than an "add printf statements"
person.

To eliminate a hurdle for anyone who wants to chip in, I've attached
the old and new debug output from the following query:

select * from pg_class where oid = 1234;

One observation is that the output is quite a bit larger with the
patched version and does not seem as useful if you wanted
OPTIMIZER_DEBUG to help you figure out why a given Path was chosen.

David