Justin Pryzby <pryzby@telsasoft.com> writes:
> I thought that would be desirable, although I don't see any better way of
> getting there than this.
Hm, but a lot of the \d commands involve more than one underlying query,
as well as a bunch of postprocessing. I doubt that the approach you seem
to be using here can handle such cases.
I think you have also stomped all over the semantics of
query-buffer-related commands that are executed in the vicinity of a \d.
Up to now, \d didn't change the query buffer. That has its uses, eg
select somecol
\dt some* -- how's that table spelled again?
from sometable;
We could maybe think about making it work by moving the \watch repetition
up a level, so that exec_command() as a whole would be repeated ... but
I wonder what people would think repetition of other commands such as \i,
\e, \r, \if, etc etc should mean.
On the whole I think this is not a can of worms I want to open. There's
a clear distinction right now between plain SQL and backslash commands,
and this is going to fuzz that in ways that are hard to predict the
consequences of.
regards, tom lane