[QGIS-trac] Re: [Quantum GIS] #2444: Discrepancy between gdalinfo and QGIS raster metadata

[Quantum GIS]

#2444: Discrepancy between gdalinfo and QGIS raster metadata
-----------------------------------------------------+----------------------
        Reporter:  alobo                             |         Owner:  nobody       
            Type:  bug                               |        Status:  new          
        Priority:  major: does not work as expected  |     Milestone:  Version 1.5.0
       Component:  Rasters                           |       Version:  HEAD         
      Resolution:                                    |      Keywords:               
Platform_version:                                    |      Platform:  Debian       
        Must_fix:  No                                |   Status_info:  1            
-----------------------------------------------------+----------------------
Comment (by rbivand):

 Replying to [comment:5 rbivand]:
 > Replying to [comment:2 mmassing]:
 > > To summarize this more concisely, I think this is a bug in R (which
 fails to handle general
 > > coordinate systems), and probably can be worked around by reprojecting
 the problematic file to a north-up orientation.
 >
 >
 > No, in no way. The R rgdal package interfaces GDAL, and the GDAL driver
 sees the file in exactly the same way. If the origin is NW, but the
 resolution increment is positive, as here, any application using the
 information GDAL has will behave in the same way. No jumping through hoops
 here, the problem must be in QGIS and/or the plugin writing the file with
 inappropriate parameters. I'm not planning to make any changes in the
 R/GDAL interface to suit a very odd case. If the data were correctly saved
 and described, GDAL would see them correctly. Maybe the GDAL driver might
 accommodate oddities, but at out end, we can't make adjustments for one
 application.


 In addition, it appears that QGIS is doing things in
 src/core/raster/qgsrasterlayer.cpp that perhaps use geoTransform values in
 interpretative ways, like switching Ymin and Ymax from those declared in
 the file. It would be better to declare them right initially, wouldn't it?

 I cannot locate the offending interpolation tool or plugin to examine its
 source code. Could someone tell me where it is?

 My understanding is that QGIS knows that the raster is saved the wrong way
 up, but uses VRT to display it correctly. In my opinion it simply should
 be stored the right way up, which would remove the need for jumping
 through hoops. I can consider adding a warning to rgdal raster reading to
 indicate that flip() should be used in the imported raster object, but
 there are limits to what is sensible.

-- 
Ticket URL: <https://trac.osgeo.org/qgis/ticket/2444#comment:6>
Quantum GIS <http://qgis.org>
Quantum GIS is an Open Source GIS viewer/editor supporting OGR, PostGIS, and GRASS formats