I think we're being unnecessarily restrictive here as we're don't actually do anything with the encoding except keep it around for other systems to use. We occasionally translate fixed length string types to a numpy dtype, in which case we would raise if the encoding doesn't make sense in numpy land.
I think we're being unnecessarily restrictive here as we're don't actually do anything with the encoding except keep it around for other systems to use. We occasionally translate fixed length string types to a numpy dtype, in which case we would raise if the encoding doesn't make sense in numpy land.