[GRASSLIST:139] Re: Considering replacing ESRI

[Paul Kelly]

On Wed, 21 May 2003, David Strauch wrote:

> On Wednesday, May 21, 2003, at 04:12  AM, Alexander Pucher wrote:
> > I still ask myself, what can ArcInfo do for me that GRASS can't...?
>
> Okay, how about orienting maps to other than true north (which I think
> is called "false northing")?
>
> I had limited experience with ArcInfo etc 10 years ago, but I remember
> that I could reorient maps, which was very useful.  I'm currently
> looking at a mountain range which trends NW-SE, and if I could orient
> my work that way I believe it would be both easier to look at and
> faster to work on, as there wouldn't be any processing time devoted to
> areas outside my area of interest (i.e. the SW and NE corners of my
> current region).

You could define a mask (r.mask) to exclude the corners you weren't
interested in. This would speed up processing for most GRASS
raster modules.

>
> I'm still a newbie to GRASS and have been hoping I would come across a
> way to do this, but lately I have begun to suspect that, as it appears,
> GRASS lacks this capacity.  Can someone prove me wrong?

I don't think it would be possible without a lot of work. In theory you
could have rotation information (centre of rotation co-ordinates and
an angle) stored along with the projection information, and the projection
programs modified to take account of the rotation when re-projecting
between different locations.

But I think the big problem would be in the region handling. As far
as I understand it GRASS only stores two northings (top & bottom) and two
eastings (right & left), which define the region. If the region wasn't
aligned to north/south and east/west the method in which raster cell
row/column co-ordinates, or pixels in the monitor are converted to
eastings and northings wouldn't work correctly. At present you can say,
e.g.
cell easting = (easting of left-hand edge of region) +
                         ( (east-west resolution) * (cell column no.) )
This wouldn't make sense if the region wasn't aligned with north/south and
east/west.

At a minimum the region would have to be defined in terms of the
co-ordinates of its four corners, and many other changes to calculations
would be needed. Possibly every module would have to be checked to see
that it didn't make assumptions about the alignment of the region. (I know
recent changes I made to s.surf.idw assume the easting and northing can
be calculated by multiplying the row/column number by the resolution and
adding it to the value for the region edge.)

Of course if you wanted to rotate the map and not keep the geo-referencing
it would be fairly simple to implement. But I'm assuming Arc/Info does
something more advanced than that (I've never used it).

Paul