Re: how robust are custom dumps?

On Wed, 2012-04-25 at 10:40 +0200, Willy-Bas Loos wrote:
> On Wed, Apr 25, 2012 at 9:51 AM, Magnus Hagander <magnus(at)hagander(dot)net>wrote:
>
> > We used to have a bug/lackoffeature in pg_dump at the 2GB boundary as
> > well, IIRC, specifically on Win32. Maybe you were hit by that one..
>
> Yes, possibly. I didn't even know how to make a compressed plain dump, but
> that doesn't really plea my case :/
>
>
> > > i do have one suggestion.
> > > pg_restore only gives a user this feedback, when he makes this
> > > mistake:"pg_restore: [archiver] input file does not appear to be a valid
> > > archive".
> > >
> > > Would it be feasible for pg_restore to detect that it is a different
> > pg_dump
> > > format and inform the user about it?
> >
> >
> The main one you'd want to detect is plain I think - and I don't know
> > if we can reliably detect that. It could be just a generic textfile,
> > after all - how would we know the difference?
> >
>
>
>
> Well, on linux you could make pg_dump run /usr/bin/file on the file to see
> what kind it is. If it is gzipped, suggest that it might be a gzipped plain
> dump, if it is plain text, suggest that it might be a plain dump (etc,
> also bzip2). That's all.
> You don't have to be sure that it is valid, just say a bit more than "does
> not appear to be a valid archive". Help a user in a bad situation.
>
> Only, i know that postgres runs on many platforms, so you probably can't
> run /usr/bin/file on all of those (or might not be installed on linux
> machine). So it probably should be part of pg_restore itself.
>

pg_restore will do so for plain backups on 9.2:

$ pg_dump b1 > b1.dump
$ pg_restore -d b2 b1.dump
pg_restore: [archiver] input file appears to be a text format dump.
Please use psql.

--
Guillaume
http://blog.guillaume.lelarge.info
http://www.dalibo.com