Re: pg_restore ignore error patch

Dear Andrew,

> > I transfer some data from one server to the other. On such occasion, I
> > use my laptop to connect with server X, I download the data, then I
> > restore them to server Y. The versions on my laptop and both servers
> > are likely to be different. My laptop is likely to have some
> > development version, and the servers may be in 7.3 or 7.4.
>
> What problems have you encountered other than dollar quoting?

If you're interested to know, there were some complaints about
the new "default_with_oids" setting, as far as I can remember.

> In the most general case, ISTM you would need to teach pg_dump how to
> degrade gracefully, and a m:1 sources to targets relationship suddenly
> becomes m:m. I'm not convinced it is worth the trouble.

I don't know what is worth the trouble. I'm just reporting a "feature"
that make the tool not work properly for me. I've fixed part of
the problem and submitted a patch, and I just noted in the mail that
there was another small problem that I did not fix.

I agree that fixing this would mean redesigning the tool in depth.
That's why I did not fixed it in the patch I submitted.

My opinion would be that pg_dump should just dump the data in some raw
format, and that all syntax re-generation issues should be dealt with by
pg_restore (INSERT vs COPY, quoting, and so on). But the tool was
not resigned this way in the beginning.

--
Fabien Coelho - coelho@cri.ensmp.fr