Re: pg_dump and premature optimizations for objects not to be dumped

I wrote:
> BTW, the bulk of the changes in this patch are just there to pass down
> the DumpOptions struct to where it's needed, following some pre-existing
> work that passed that around as a separate pointer. I wonder why things
> were done that way, rather than adding a DumpOptions pointer in the
> Archive struct? The vast majority of the functions I had to touch already
> had an Archive* parameter, and would not have needed to be touched if it
> were done the other way. I'm tempted to push a preliminary patch that
> adds such a field and gets rid of DumpOptions as a separate parameter
> to anything that also takes an Archive* pointer.

Specifically, something like the attached, which is bulky but mostly
mechanical. This undoes a lot of API changes introduced by 0eea8047bf
to pass DumpOptions as a separate argument to assorted functions.
I think this is a good idea because (a) those changes were pretty random
and incomplete, as evidenced by the fact that I had to do a bunch more of
them in my upthread patch; (b) this significantly reduces the delta
between 9.5/HEAD and pre-9.5 code in pg_dump, which will no doubt ease
future back-patching. So unless I hear objections, I plan to push this
pretty soon.

regards, tom lane