[Cbor] Co-occurrence constraints (Re: CDDL optionality and choices and optionality in CBOR and CDDL)

[Carsten Bormann <[e-mail]>]

On 13. Apr 2024, at 02:49, Rohan Mahy <[e-mail]> wrote:
> 
> Order = [
>   food = [ * food_items ], 
>   ( includes_alcohol: true, drinks = [ * drink_items ] ) / ( includes_alcohol: false, to_go: bool )
> ]

Almost (modulo some syntax issues).
The constructs in the parentheses are CDDL groups, so you would need a group choice:

> Order = [
>   food: [ * food_item ], 
>   ( includes_alcohol: true, drinks: [ * drink_item ] //
>     includes_alcohol: false, to_go: bool )
> ]

(Note the low operator precedence of group choices.
I also used the singular form for the items, because each item is just one item.)

This approach can solve a certain subset of the use cases that we call “co-occurrence constraints”; a more general solution usually requires some predicate language that in turn requires a navigation mechanism such as JSONPath [2]. 

BTW:

> bool = true / false

You don’t have to say this, because it already is defined [almost] that way in the CDDL prelude [1].
(Because of the “almost”, you actually cannot say it this way.)

Grüße, Carsten

[1]: https://www.rfc-editor.org/rfc/rfc8610.html#appendix-D
[2]: https://www.rfc-editor.org/rfc/rfc9535.html