search

cbor-wg / cddl

The draft leading up to RFC8610
https://tools.ietf.org/html/rfc8610
Other
22 stars 4 forks source link

Support for co-occurrence constraints #22

Open anweiss opened 6 years ago

anweiss commented 6 years ago

I don't currently see any support for co-occurrence constraints in the spec. Would be great to allow for values on the right-hand side of control operators to be cross-references to other map/group entry types in the same instance document.

cabo commented 6 years ago

While there are no co-occurrence constraints in CDDL, many actual use cases can be addressed by using the fact that a group is a grammar:

postal = {
  ( street: text,
    housenumber: text) //
  ( pobox: text .regexp "[0-9]+" )
}

Can you tell us more about the specific use cases you have in mind?

anweiss commented 6 years ago

Sure thing. So for example:

session = {
  timeout: uint,
}

other-session = {
  timeout: uint  .lt [session.timeout],
}

In this example, for a given instance document, I'd like for the value of the timeout entry in the other-session map to be less than the value of the timout entry in the session map.

anweiss commented 6 years ago

Ideally, would love to use CDDL to validate data relationships in JSON instance documents.

cabo commented 6 years ago

Right, this would require an XPath-like mechanism to crawl the tree to find that other item (which may be later in the serialization...). This is probably done better with a separate, predicate-based mechanism. It is interesting to see how this was not done for W3C schema, e.g.: https://www.w3.org/wiki/Value_arithmetic_required_-_attributes

On a more philosophical note, this moves the language way more into the context-sensitive realm. It is true that the key uniqueness constraint on maps already does this -- there is no way I know around that in a JSON-derived generic data model.

On a practical note, people are writing data validators in Python and not Schematron. Are we sure we can find a notation that is actually better than Python for this? If we do, this could be something for CDDL 1.1.

(When we did DSSSL, the precursor of XSLT/XSL, we opted to use Scheme as a programming language to write the transformations in. That had some amazing power [it was also where the tree-crawling that then went into XSLT was designed]. It also was sometimes hard to get right, and to understand what was specified.)

cabo commented 6 years ago

A related question would be: What is your favorite way to navigate a JSON tree? JSONpath?

anweiss commented 6 years ago

I've been using JSON Pointer and the I-D JSON Relative Pointer