You could get rid of all these elif switches by splitting the typify function, and removing jsonify.

Wrong name, as it's reading from JSON, instead of producing JSON.

Can be shorter:

functions = {
    SNAPSHOT: self.snapshot,
    REVISION: self.revision,
    # ...
}
return functions[pid_.object_type](pid_)

(and it would check the type is known)

same

What about generating a list of branches instead? That's what callers of this function will have to do one way or an other, anyway.

yield from r.iter_content(chunk_size=None, decode_unicode=False)

I don't get what you mean about typify here.

If you want to split into different typify_foo functions and a dictionary based dispatched, I disagree, as some will be oneliners and then we'll have to catch the KeyError in there to reraise a ValueError. All in all, the code will be longer.

If it's something else, can you explain?

jsonify() was actually producing json, with .json() from the response object, and returning it as Python data.
But I got rid of it anyway as it was overkill to have a function just for that.

I haven't done dictionary based dispatching for the iter() case, because yield is syntactically invalid within lambda, so I'd have to use def-s instead for a bunch of oneliners, and that would make the code longer. (I've added the fall through else to check for errors which was missing though.

Good idea, done.

I note that this is now inconsistent with directory, which does return the parent directory id in each entry, but it's an inconsistency that is derived from the API itself, that should probably be fixed there with 2.0 anyway.

You don't need dictionary dispatch, because all lines calling typify have the value of obj_type defined statically; so they might as well call typify_* functions directly.