[gdal-dev] AREA_OR_POINT vs GDALMD_AREA_OR_POINT

[Joaquim Luis]

On 29-05-2014 19:26, Even Rouault wrote:
>
> Joaquim,
>
> I'm confused by what you are actually trying to do.
>

OK Even, I'll try to explain better. As we all know here, GDAL reports 
all region limits as pixel corner coordinates (AREA_OR_POINT=Area for 
GeoTiffs). GMT on the other hand keeps track of the original 
registration (we call it grid or node registration (cell center) and 
pixel registration (cell corner)). And we can toggle between one and 
another either by just adjusting the grid limits or by re-interpolation.

When GMT reads grids via GDAL it only knows that what the latter tells 
it, but not the data's original info. A decision was made (mea culpa) 
that all rasters of char type (images) are treated in GMT as 'Area' 
(pixel) and the others (grids) as 'Point'. This allow as, for example, 
that a netcdf grid created and read by GMT internal code or via GDAL  
sow the same limits (most of times). It's a pity that the original 
registration type is not preserved in GDAL so we could respect it. But 
GeoTIffs when they use the AREA_OR_POINT key store that info, so in that 
case I try to respect it. So, answering to your question, that is what I 
am trying to do

> GDALMD_AREA_OR_POINT is a #define that evaluates to "AREA_OR_POINT". 
> The string "GDALMD_AREA_OR_POINT" however is nowhere used in GDAL.
>

So let me repeat one of my previous questions, why was this ignored?

     GDALSetMetadataItem(hDstDS, "AREA_OR_POINT", "Point", NULL);

> Regarding the MEM driver, GDALMD_AREA_OR_POINT is just a random 
> metadata item.
>
> Some drivers will set GDALMD_AREA_OR_POINT to the GDALMD_AOP_POINT = 
> "Point" value.
>
> And the GTiff driver uses it to determine if it must apply a 
> half-pixel shift between the GDAL convention and the GeoTIFF convention.
>
> Even
>
> > Hi,
>
> >
>
> > I made GMT respect as far as possible the AREA_OR_POINT GeoTiff key
>
> > for read/write, but ...
>
> >
>
> > The following was silently ignored (GMT uses the MEM driver)
>
> >
>
> > GDALSetMetadataItem(hDstDS, "AREA_OR_POINT", "Point", NULL);
>
> >
>
> > so digging here and there I found about GDALMD_AREA_OR_POINT, and in
>
> > fact this one is now taken into consideration
>
> >
>
> > GDALSetMetadataItem(hDstDS, "GDALMD_AREA_OR_POINT", "Point", NULL);
>
> >
>
> >
>
> > However, this implies that on reading I have now to test for both
>
> >
>
> > GDALGetMetadataItem(hDataset, "AREA_OR_POINT", NULL)
>
> >
>
> > ---- AND ----
>
> >
>
> > GDALGetMetadataItem(hDataset, "GDALMD_AREA_OR_POINT", NULL)
>
> >
>
> > This means that other codes that pay attention to the AREA vs POINT
>
> > thing will have to do both tests as well and if they do only the first
>
> > one, GeoTiffs created by GDAL will have the problem ignored, and quite
>
> > likely an error in coordinates will follow.
>
> >
>
> > So the question is, why this name duplication (AREA_OR_POINT vs
>
> > GDALMD_AREA_OR_POINT) and why GDALSetMetadataItem() ignores the
>
> > AREA_OR_POINT key?
>
> >
>
> > Thanks
>
> > Joaquim
>
> >
>
> > ----------------------------------------------------------------
>
> > This message was sent using IMP, the Internet Messaging Program.
>
> >
>
> > _______________________________________________
>
> > gdal-dev mailing list
>
> > gdal-dev at lists.osgeo.org
>
> > http://lists.osgeo.org/mailman/listinfo/gdal-dev
>
> -- 
>
> Geospatial professional services
>
> http://even.rouault.free.fr/services.html
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/gdal-dev/attachments/20140529/93f3020b/attachment.html>