[GRASS-dev] Rast3d_location2coord inconsistent behavior

[Anna Petrášová]

On Sun, Jul 20, 2014 at 12:04 PM, Sören Gebbert <
soerengebbert at googlemail.com> wrote:

> 2014-07-20 6:10 GMT+02:00 Anna Petrášová <kratochanna at gmail.com>:
> >
> >
> >
> > On Sat, Jul 19, 2014 at 6:30 PM, Sören Gebbert
> > <soerengebbert at googlemail.com> wrote:
> >>
> >> Hi Anna,
> >>
> >> 2014-07-19 19:04 GMT+02:00 Anna Petrášová <kratochanna at gmail.com>:
> >> > Hi,
> >> >
> >> > I have a question about Rast3d_location2coord [1]. It is casting the
> >> > col,
> >> > row, depth values to integer like this:
> >> >
> >> > (int) col
> >> >
> >> > instead of using
> >> >
> >> > (int)floor(col)
> >> >
> >> > as a result, col = 0.1 => col = 0 and col = -0.1 => col = 0, which
> >> > doesn't
> >> > seem correct to me. Any opinions?
> >>
> >> I totally agree that there should be a function that returns the row,
> >> col and depth as double values, so the developer can decide how to
> >> handle these values. Otherwise there would be no convenient way to
> >> determine where coordinates are located in the voxel.
> >>
> >> But i am not sure if the  integer casting is a bug, since the
> >> Rast_easting_to_col() /Rast_northing_to_row() docs suggest integer
> >> casting as conversion method. There are algorithms in GRASS that use
> >> only integer casting and algorithms that make use of rounding with
> >> floor() before casting to integer.
> >>
> >> The problem arise in case of negative column,row or depth values when
> >> casting to integer without rounding to an integer smaller than the
> >> double number. So the question is are there any chances to receive
> >> negative cols,rows and depths?
> >>
> >> It does not seems do to me for col and row, since in case of negative
> >> coordinates will west and south always be smaller than east and north,
> >> resulting in always positive cols and rows. And the same should be
> >> true for depths, unless the top is smaller than bottom.
> >
> >
> > Yes, but you still can get negative col, for example when looking for the
> > neighboring voxels of a voxel on the edge of the 3D raster. Then, the
> > inconsistency causes problem.
>
> Good point, i have updated the Rast3d_location2coord() function to use
> floor.
>

Thanks, I hope it won't be too much overhead comparing to int.

Anna


> Best regards
> Soeren
>
> > Thanks a lot for implementing Rast3d_location2coord_double.
> >
> > Anna
> >
> >>
> >> Best regards
> >> Soeren
> >>
> >> >
> >> > Thanks, Anna
> >> >
> >> > BTW, similar function Rast_northing_to_row from raster library returns
> >> > double.
> >> >
> >> >
> >> > [1] http://grass.osgeo.org/programming7/region_8c_source.html#l00291
> >> >
> >> > _______________________________________________
> >> > grass-dev mailing list
> >> > grass-dev at lists.osgeo.org
> >> > http://lists.osgeo.org/mailman/listinfo/grass-dev
> >
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/grass-dev/attachments/20140720/73e2f180/attachment.html>