Re: [Cbor] CDDL 2.0 import and paths

[Carsten Bormann <[e-mail]>]

On 2023-02-06, at 15:21, Michael Richardson <[e-mail]> wrote:
> 
> 
> Carsten Bormann <[e-mail]> wrote:
>> ;# include rfc9052
>> ;;; include all the rules from rfc9052.cddl as well as the ones
>> imported/included there, no questions asked
> 
> So that's recursive.

Well, cddlc will break its little mind quickly if it actually gets recursive…

>> ;# import rfc9052
>> ;;; include only those rules from rfc9052.cddl that actually are
>> undefined references in the current CDDL file, treating rfc9052 as a
>> “library”.
> 
> So, that's not recursive, and if there were dangling references in
> rf9062.cddl, I'd have to take care of it?

Actually, this also runs a transitive closure *within the imported CDDL file*, and it acts on directives inside that (but not adding anything nested based on undefined references from outside).  I added a nested import directive to rfc9053.cddl, if you want to play with this.

But the intention is that a recursive import would actually stop (haven’t set up a test case for that yet).

> A question about CDDL_INCLUDE_PATH, does it always start at the beginning, or
> does it start at the place in the path where the current file was found?

The path always starts at the beginning.
So you can override/reach into dependency chains (on purpose or inadvertently).

>> I plan to add something like
> 
>> ;# import rfc9052 as COSE
> 
>> which would rename the names in the imported rules to start with COSE(*).
>> (This is of course interesting because of other imported names, as well as prelude names…)
> 
> A good idea.
> I'll update my use of cddlc later this month when I return to that work.
> 
>> (*) I’m not yet sure whether the auto-import mechanism should add glue
>> rules automatically, so you don’t have to call your referencing names
>> COSE.COSE_Encrypt0 while still getting overly generic names such as
>> `label` or `values` prefixed as `COSE.label` and `COSE.values`.  This
>> could be limited to specific names listed in a “import … from” part of
>> the directive.  I need to see more examples to make a useful
>> suggestion.
> 
> A good question, I think prefixing is okay.

Thanks for the feedback!

Grüße, Carsten