Tom Lane wrote:
>
> Bernard Frankpitt <frankpit@pop.dn.net> wrote (a couple weeks ago):
> > When I was altering the xfunc.sgml page I came across this:
>
> " " "
> >
> That's talking about builtin functions, ie functions implemented by
> statically-linked routines in the standard backend.
>
> regards, tom lane
Oh, That explains it. It didn't occur to me that people adding custom
functionality to the backend wouldn't use a dynamically linked
interface. In fact it occured to me that it might be a good idea to
convert some of `poor relations' the stuff like gist, and perhaps rtrees
to dynamically linked modules in /contrib. They might then provide
better examples of how to develop and link major extensions into the
backend in a relatively painless way. Also an exercise like that would
really provide a good opportunity to define and document the backend
code interfaces between the executor and access methods, and between
access methods and the low-level database functionality (buffer
management, tuple time-validation etc.). Once I finish my dissertation,
I was sort of planning to start chipping away at some documentation for
the code internals.
To me, the extensibility features and open design of PostgreSQL are its
most exciting features, and I think that a good set of documents on the
internal functionality and interfaces would be rewarded in the long term
by innovative features and unusual applications from developers in a
wide variety of fields.
Bernie
