Re: [netmod] definition of "fully expanded YANG"

[Robert Varga <[e-mail]>]

On 17/07/2019 15:54, Juergen Schoenwaelder wrote:
> On Wed, Jul 17, 2019 at 01:18:57PM +0000, Michael Rehder wrote:
>> Yes, inlining of everything, including augments. What you'd see in the tree representation.
>> For example augments makes it particularly impossible for a consumer of the YANG schema to process the schema since the final result.
>>
>> Some use cases:
>> - translating to another schema language
>> - translating to a UI dialog for data entry
>> - an accurate diff on versions
>>
>> Model-driven-design means using the "computable specification" for all sorts of things, increasing the ROI of the schema definition.
> 
> It sounds to me like you are interested in what compiler people call
> the abstract syntax tree, which is quite a different thing than the
> YANG module you start with. The easiest and most robust way to do
> translations is likely to simply expand pyang (or any other compiler
> of choice).

Depends on how you define the AST (and how many different ASTs are in
your compiler), as .yang file has an AST itself, which you then
cross-reference between modules/submodules, and you arrive at some IR,
from which you derive the effective model with whatever indices/details
your use case requires.

This is certainly possible with OpenDaylight tooling, i.e. arrive at
https://github.com/opendaylight/yangtools/blob/master/yang/yang-model-api/src/main/java/org/opendaylight/yangtools/yang/model/api/EffectiveModelContext.java,
which gives you a per-module tree of statements with all RFC7950
semantic processing applied. It also contains some generally-useful
indices (like schema tree/data tree traversal) as well as references to
original statement declaration (i.e. as it appeared in .yang file, if
applicable).

Regards,
Robert