"Diogo Biazus" <diogob@gmail.com> writes:
> The idea I've been discussing with Simon Riggs is to create a set of
> functions that can be called from within the database.
I'd question that at the very start. I don't see any strong reason to
do it that way, and as you admit further down it'd make it impossible to
use the viewer to work on extracting data from a failed cluster; which
is, at least in my mind, one of the primary use-cases for the thing.
I would suggest building the xlog-reader as a separate program, not part
of the backend. It would be useful to have options to (eg) translate
table OIDs to table names, which would require a connection to the
database, but this should be an *option* not an essential condition.
regards, tom lane