Improve handling of `merge_objs` with `splitted`.
Encodings built with Encoding.splitted
can never be provided to Encoding.merge_objs
even when the json variant is indeed an object. IIUC there is no reason to have such a limitation: an encoding built with merge_objs
cannot fail when used with the binary reader/writer, it can only fails when used with Json reader/writer.
In fact, it seems that the field is_obj
(or is_tup
) in the case Encoding.Splitted
is always set to false. Is there a reason not to use the function is_obj
from json-data-encoding
? Take aside the fact that the function is currently not exported...
Here is what I have in mind to remove such a limitation:
Edited by G.-B. Fefe