[postgis-devel] ST_Rescale issues again

[Sandro Santilli]

On Thu, Aug 28, 2014 at 09:57:21AM +0200, Sandro Santilli wrote:
> On Thu, Aug 28, 2014 at 09:56:20AM +0200, Sandro Santilli wrote:
> > On Thu, Aug 28, 2014 at 09:53:46AM +0200, Sandro Santilli wrote:
> > > I keep getting suprised by ST_Rescale/ST_Resample/etc outputs.
> > > What do you think about this.
> > > 
> > > 
> > > Original table tiles are fully aligned: 10x10 sized, all with scale 1,-1:
> > > 
> > >  =# select distinct st_scalex(r), st_scaley(r), st_width(r), st_height(r)
> > >  => from ( select r from res1 ) f;
> > >   st_scalex | st_scaley | st_width | st_height
> > >  -----------+-----------+----------+-----------
> > >           1 |        -1 |       10 |        10
> > >  (1 row)
> > > 
> > > Passing each of them through ST_Rescale with values 2.0, -2.0 result in
> > > some tiles getting the new scale requested and some not:
> > > 
> > >  =# select distinct st_scalex(r), st_scaley(r), st_width(r), st_height(r)
> > >  => from ( select ST_Rescale(r,2.0,-2.0) r from res1 ) f;
> > >   st_scalex | st_scaley | st_width | st_height
> > >  -----------+-----------+----------+-----------
> > >           2 |        -2 |        5 |         5
> > >           1 |        -1 |        5 |         5
> > >  (2 rows)
> > > 
> > > Why is that ?
> > 
> > I shall add that I get this warning 4 times:
> > 
> >  NOTICE:  Raster has default geotransform. Adjusting metadata for use of GDAL Warp API
> >  CONTEXT:  SQL function "st_rescale" statement 1
> > 
> > And that the deviant tile is a single one over a total of 648 tiles:
> > 
> >   count | st_scalex | st_scaley | st_width | st_height
> >  -------+-----------+-----------+----------+-----------
> >     647 |         2 |        -2 |        5 |         5
> >       1 |         1 |        -1 |        5 |         5
> >  (2 rows)
> > 
> > Should I file a ticket ?
> 
> Also, GDAL is "GDAL 1.11.0, released 2014/04/16" 

It fails in both 2.1 branch and trunk.

The offending tile has extent (0 -10, 10 0), scale (1,-1) and size (10,10).
Out of ST_Rescale(2.0, -2.0) it gets an extent of (0, -5, 5 0), scale(1,-1)
and size of (5,5).

Ain't that a (pretty big) bug ?

--strk;