[postgis-devel] UnknownSRID value: 0 or -1 ?

[Sandro Santilli]

On Mon, Dec 12, 2011 at 04:23:01PM +0100, Sandro Santilli wrote:
> On Fri, Dec 09, 2011 at 10:03:52AM -0800, dustymugs wrote:
> > On 12/09/2011 09:35 AM, Sandro Santilli wrote:
> > >I've started work on testing dump/restore procedure for getting to 2.0
> > >and realized the data w/out a proper SRID won't come back due to
> > >ensure_srid_geom checks on tables insisting of the SRID being -1
> > >rather than 0.
> > >
> > >There's no much we can do to fix this, unless we're willing to
> > >drop the constraints from the user dump.
> > >
> > >For this reason I think it would be safer to get our unknown
> > >srid back to -1 before it's too late (srid=0 getting in dumps
> > >and making things even harder).
> > >
> > >If somebody has smarter ideas I'm happy to listen to them.
> > >
> > 
> > I have no smarter ideas but I prefer 0 without any real reasons.  I
> > don't get the purpose of having a SRID check constraint if the
> > constraint is enforcing the unknown SRID.
> 
> Well, it's not something we have control on, it's stuff you have
> in your database backup, unless you did something about it _before_
> dropping it.
> 
> The only way I can think which could fix this would be changing ST_Srid
> to return a SpatialRefSys object rather than an integer and make it
> so an integer <= 0 casts to the _same_ SpatialRefSys object
> (the "unknown srs" object). Mind you: SQL/MM does have such type...

I got to think of another way and implemented it.
As of r8366 the utils/new_postgis_restore.pl will _rewrite_
those constraints to ensure there's an st_ prefix and that
checks for srid==-1  become checks for srid==0.
It worked perfectly for me.

I also ported the logic about operators from the old version
of that script and am now pretty happy about the core.

Raster and Topology may still need some love but the core seems fine.

--strk;

  ,------o-. 
  |   __/  |    Thank you for PostGIS-2.0 Topology !
  |  / 2.0 |    http://www.pledgebank.com/postgistopology
  `-o------'