re. http://grass.itc.it/pipermail/grass5/2005-January/017029.html
Currently r.in.gdal imports unprojected data in the negative, i.e. 0,0
is the top right of the image.
(unlike r.in.tiff|png in GRASS 5.4 which makes 0,0 bottom left)
This can be "fixed" with r.region by the user after import.
some comments from the source code:
/* TODO: most unreferenced formats are read in with negative
* coordinates - desired??
*/
[...]
/* use negative XY coordinates per default for unprojected data
but not for all formats... (MN: I don't like it at all) */
/* for hDriver names see gdal/frmts/gdalallregister.cpp */
if ( l1bdriver ||
( strcmp(GDALGetDriverShortName(hDriver),"GTiff") ||
strcmp(GDALGetDriverShortName(hDriver),"JPEG") ) == 0 )
{
/* e.g. L1B - NOAA/AVHRR data must be treated differently */
/* define positive xy coordinate system to avoid GCPs confusion */
cellhd.north = cellhd.rows;
cellhd.south = 0.0;
[...]
else
{
/* for all other unprojected data ... */
/* define negative xy coordinate system to avoid GCPs confusion */
cellhd.north = 0.0;
cellhd.south = (-1) * cellhd.rows;
I don't understand the logic or intention here: who's supposed to be
getting set to negative and why?
FWIW the above strcmp() part of the test will always fail.
i.e. strcmp() returns 0 if the strings match. It will never match both a
GeoTIFF and a JPEG, so one side or the other of the "OR" will be always
be non-zero and the ==0 part will always be false. So it really reduces
to if(l1bdriver) {make positive} else {make negative}.
(l1bdriver is AVHRR; http://www.gdal.org/frmt_l1b.html
Will this data ever be without GCPs & so touch this part of the code?!)
[And PNG should be doing the same thing as TIFF & JPEG.]
So what is the intent of the test?
Are GeoTIFFs and JPEGs supposed to be ending up positive or negative?
(positive please)
Who's getting what confused with GCPs? Does that situation still exist?
I prefer to make everything positive as it's less confusing.
gdalinfo reports this on an unprojected TIFF:
Driver: GTiff/GeoTIFF
Size is 1309, 4920
...
Corner Coordinates:
Upper Left ( 0.0, 0.0)
Lower Left ( 0.0, 4920.0)
Upper Right ( 1309.0, 0.0)
Lower Right ( 1309.0, 4920.0)
Center ( 654.5, 2460.0)
So in GDAL 0,0 is top left (image coordinates (or satellite scanlines)
as opposed to cartesian coordinates)
GRASS XY uses cartesian coordinates so 0,0 should be bottom left if we
want to keep map units positive. (external support for idea: this is the
reason for having false eastings and northings)
So is this the "GCP confusion" the code speaks of? Solution is to make
it neither? As we are GRASS, matching GRASS XY conventions seems
like logical thing to do.
I would like to use i.point's $GROUP/POINTS file as a template for
setting GCPs in a GeoTIFF directly with gdal_translate or geotifcp,
but it seems like I'll have to fuddle the i.points results around
slightly to do so? (flip y-axis) If everything is positive it is
easy enough to write a script to form a "gdal_translate -gcp"
command string from i.points output.
It seems it is trivial to make everything positive in the r.in.gdal
code; can anyone justify a reason not to?
Aka any reason I should not just go ahead and do this in 6.1-cvs?
thanks,
Hamish