Selon Ole Nielsen <ole.moller.niel...@gmail.com>: > Hi Even > > Thanks again for all your help - all tests now pass using either gdal 1.6 or > 1.8. Just to summarise, there were three issues as I see them: > > > 1. Bounding boxes for raster data are computed differently (but better) > in newer versions
Specific to PixelIsPoint in the GTiff driver. > 2. While v1.6 of ogre would allow an single number of type numpy.ndarray > this is no longer allow. Casting it to a float solves the problem. I wasn't aware it could work before. There's perhaps now an ambiguity on which method to use, due to the addition of new possibilities for SetField() (just a bling guess). > 3. New versions of ogre will not allow attribute names of length longer > than 10 (at least when using the ESRI driver). Formerly, I think it > silently > truncated, now it is up to the user. That's at least what I did. I've a looked quickly a bit at your code and I've the feeling that you truncate yourself, instead of relying on OGR. The Shapefile driver does more than truncating. It also ensures uniqueness of the truncated names. For example, if you try creating "a_very_long_name" and "a_very_long_name_again", they will be respectively truncated to "a_very_lon" and "a_very_l_1". A safer solution to get the field name that got to the file would be to : field_defn = ogr.FieldDefn("a_very_long_name", ogr.OFTString) lyr.CreateField(field_defn) layer_defn = lyr.GetLayerDefn() last_field_idx = layer_defn.GetFieldCount() - 1 real_field_name = layer_defn.GetFieldDefn(last_field_idx).GetNameRef() _______________________________________________ gdal-dev mailing list gdal-dev@lists.osgeo.org http://lists.osgeo.org/mailman/listinfo/gdal-dev