lyd_validate_all and empty tree

[choppsv1]

I have a question about empty trees. In FRR we initialize a new tree by calling, dnode = NULL; lyd_validate_all(&dnode, NOSTATE); This is nice that it gets us a dnode (and more apparently) to use as a tree anchor.

I was replicating this behavior with the python libyang bindings, and it got me to wondering what exactly was being put in these dnodes. I only loaded a single module (frr-logging.yang) (and ietf-syslog-types for it's identities), and the root node it returns is a DNode named logging. What's interesting is that in the module itself the root is container logging {} so it seems like the DNode being returned is special in some way (as it's not a DContainer and the container is non-presence container at that).

Is this documented anywhere? I'd like to better understand what this non-container node for a container node is and what I can expect from it (API guarantee-wise).

BTW, I actually need an supported way to do the above, currently I have to use _libyang.ffi etc to invoke lyd_validate_all() directly as the only python binding for validate call is currently a method of DNode