[postgis-users] Incorrect results from queries specifying an SRID

[Paul Ramsey]

Actually, this is a slightly user-hostile annoyance I ran across last 
night while writing the PostGIS workshop for MUM (I hope everyone is 
coming, I expect to see all 450 of you there (and expect all of you to 
buy me a beer (the workshop might be a little disjointed))).

Since box3d's cannot hold an SRID, if you ever get a cast into box3d, 
you lose your SRID, which results in ugly SQL like this:

WHERE the_geom && setsrid(expand(geometryfromtext('POINT(100000 
100000)',42102),100),42102)

(I recognize that I can shrink the lenght of it a bit by not initially 
declaring an SRID, it will get stripped later, after all.) To the user 
it is not immediately obvious why the more straightforward SQL does not 
work:

WHERE the_geom && expand(geometryfromtext('POINT(100000 
100000)',42102),100)

ERROR:  Operation on two GEOMETRIES with different SRIDs

You make a geometry, and expand it, seems like a nice obvious piece of 
SQL, but the error message you get says your SRIDs mismatch. But you 
set an SRID, it's right there in the SQL! Confusion reigns.

On Thursday, June 3, 2004, at 01:36 AM, Mark Cave-Ayland wrote:
>
> Hmmm..... I can't see why this one doesn't work. All I can think of is
> that the BOX3D cast causes the SRID not be set or set incorrectly. I'll
> try and have a look at this in the near future, however I'm not sure
> exactly when I'll be able to get around to it.

      Paul Ramsey
      Refractions Research
      Email: pramsey at refractions.net
      Phone: (250) 885-0632