[postgis-users] Input versus output geometry

[strk]

On Thu, Sep 02, 2010 at 02:24:30AM -0700, Alex Zepeda wrote:
> strk wrote:
> 
> >Are you sure your constraint checked your input in the original table ?
> 
> Yup.  I went over this with the postgres guys on postgres-bugs and managed
> to distill something of a test case as far as the constraints being checked.
> The queries I posted (courtesy of Tom Lane) produced repeatable results on
> (at least) two independent machines.
> 
> # SELECT
> dropbbox(ST_expand(setsrid(makepoint(-122.50367,37.74189),4326), 0.00004)) 
> &&
> dropbbox(ST_expand(setsrid(makepoint(-122.50376,37.74185),4326), 0.00004)),
> ST_expand(setsrid(makepoint(-122.50367,37.74189),4326), 0.00004) &&
> ST_expand(setsrid(makepoint(-122.50376,37.74185),4326), 0.00004);
>  ?column? | ?column?
> ----------+----------
>  t        | f

Very interesting.
Also this:
SELECT
box2d(
   dropbbox(
   ST_expand(setsrid(makepoint(-122.50376,37.74185),4326), 0.00004)
   )
),
box2d(
   ST_expand(setsrid(makepoint(-122.50376,37.74185),4326), 0.00004)
);

box2d | BOX(-122.503807067871 37.7418098449707,-122.503715515137 37.7418937683105)
box2d | BOX(-122.503799438477 37.7418098449707,-122.503723144531 37.7418899536133)


Uhm. It's definitely a bug in LWGEOM_expand.
Please file it and someone will take care of it (possibly me, but
not sure I'll find the time shortly)

--strk;

  ()   Free GIS & Flash consultant/developer
  /\   http://strk.keybit.net/services.html