Re: [Cbor] eliding CBOR diagnostic notation binary contents

[Michael Richardson <[e-mail]>]

Carsten Bormann <[e-mail]> wrote:
    >> Signed PGP part
    >>
    >> Carsten Bormann <[e-mail]> wrote:
    >>>>> Actually processing elisions should be a special flag.  That should
    >>>>> address your concern.
    >>>>
    >>>> Yes, that works for me.
    >>
    >>> In cbor-diagnostic, this flag is now “-ah”.
    >>
    >> Cool.
    >>
    >>> $ gem install cbor-elision
    >>
    >>> $ echo "h'...4711...'" | diag2pretty.rb -ah
    >>
    >> The direction I really want to automate is cbor2pretty, when generating examples.

    > Are you sure about cbor2pretty (annotated hexdump) vs. cbor2diag?

Uhm.  Both probably.  I'm often confused about which one is which :-)
I find myself using pretty often because it is easy to load them up as
reference values for unit test cases.

    > A hexdump is much harder to have elisions that you can read in again:
    > The length information implicitly conveyed by the syntactic structure
    > of diagnostic mode is missing.
    > So you don’t know how many bytes the elision is supposed to cover.
    > (Unless you start parsing the annotations…)

Agreed, but for showing some things the hexdump is necessary, but it is not
useful to show long sequences like signatures or embedded payloads.

    >> There are some ways in which I can see doing this:
    >> 1) some option which just elides any bstr longer than X.
    >> 2) some option which elides based upon some other (content-relevant) tag.
    >> 3) some option which takes some kind of CBOR-PATH to specify what to elide.
    >>
    >> I think that "CBOR-PATH" is not yet a thing, but I think I've heard you talk
    >> about how we can essentially use JSONPATH to do this?

    > Also, it might be useful to mix in a CDDL spec and build a path based on the parse tree.

yes, this would certainly be interesting: ".elidable" or some such attribute.

    > CBOR-Path should be able to handle all of these, but it still needs to
    > be invented (after JSONPath is done, presumably).

I am blissfully ignorant of JSONPath situation.

What do we do in the meantime?  I think that generating (diag format)
examples is the killer-app here.  Maybe even to the point where kramdown
should have some "```` foobar" for doing this.
I guess the CDDL matching part maybe.  I suppose this also validates the CBOR
against the CDDL along the way, which is a win, except when it's a pain.


--
Michael Richardson <[e-mail]>   . o O ( IPv6 IøT consulting )
           Sandelman Software Works Inc, Ottawa and Worldwide