[postgis-users] Vacuum analyse ERROR: Operation on twoGEOMETRIESwith different SRIDs

[Markus Schaber]

Hi, Mark,

Mark Cave-Ayland wrote:

> Having mixed SRID geometries in a column is, of course, completely
> non-sensical since the results of any spatial query could be invalid.

I don't think so, I can imagine several possible uses of mixing SRIDs in
the same column.

One could query by other columns, and get the resulted geometry. Or have
two columns, one containing the original geometry as received from
outside with whatever SRID they use, and the second one containing the
WGS84 representation.

Even a query that looks like "... WHERE transform(geom, 4326) &&
box2d(..." could make sense, and can even be made efficient using
functional indices.

And some SRID systems like the different Zones for Gauss-Krueger used in
europe are even compatible in the sense that they do not overlap when
used properly.

I do not know whether we really are in the position to tell users that
this is all nonsense.

We are in the position to tell them that this is currently not
supported, however :-)

Maybe the best short-term solution would be to forbid the generation of
indices on mixed SRID columns, but make ANALYZE stable enough to cope
with everything it stumbles over.

Markus