[gdal-dev] Encoding of keys and values returned by GDALGetMetadata?

[Even Rouault]

Sean,

> Little is said at
> https://www.gdal.org/gdal_8h.html#a25929a66e99d92c2788485ecf13e2e82. Are
> these values UTF-8 (as was the intent in
> https://trac.osgeo.org/gdal/wiki/rfc5_unicode) or does it depend on the
> format driver?

The key/value mechanism itself just assumes NUL-terminated strings. The 
general aim of GDAL should be to have UTF-8 everywhere, but the reality is 
that not all drivers enforce this, so you might easily end up with 'random' 
encodings.

> In https://github.com/mapbox/rasterio/issues/1663 a rasterio
> user is encountering HDF5 metadata items that can't be decoded using
> Python's UTF-8 codec and I've love to be able to sort this out.

OK, looking at the details of this particular case, this is an inappropriate 
handling by the HDF5 driver of some types of attributes.
Fixed by
https://github.com/OSGeo/gdal/commit/9640098fc79b6ba6bb135f8715fa3e7dd2705691
which I've backported to 2.4 as well

Even

-- 
Spatialys - Geospatial professional services
http://www.spatialys.com