Re: [Json] Schemas & so on

[Peter Cordell <[e-mail]>]

On 03/05/2016 14:12, John Cowan wrote:
> However, I think that there is a counterexample to the "no JSON syntax"
> claim, which is Examplotron.  ...
>
> I had hoped, when I started to read the JCR document, that JCR was a
> language of this type.  But while valid JSON instances are for the most
> part also valid JCR, an instance does not match other structurally similar
> instances: instead, it matches only itself, making it entirely useless.
>

I disagree that JCR is "entirely useless".  I think it's a great 
strength that you can start with an actual JSON example, but then 
generalise it via annotation to be more embracing.  I think Examplotron 
is neat, but it makes lots of assumptions about what you're trying to 
express.  It essentially guesses what you mean by a construct.  This may 
work for a little while, but soon runs out of steam.  There's a reason 
why people say you should be explicit about what you want.

Plus, there is the issue of practical syntax.  XML offers namespaces so 
you can do things like <bar eg:occurs="*"> to differentiate between one 
or more and zero or more (or optional etc).  Lacking even comments it's 
hard to see how a JSON based syntax could do this is an elegant way.  (I 
can imagine some ugly hacks involving unlikely member names, but I'd 
really rather not.)

And the ability to explicitly state that you want a particular string 
(for example), rather than having that being inferred as any string, 
opens up the use of JCR as part of a test vector suite, which could be 
very useful as part of implementation validation.

So, very useful indeed IMO.

Pete.
-- 
---------------------------------------------------------------------
Pete Cordell
Codalogic Ltd
Read & write XML in C++, http://www.xml2cpp.com
---------------------------------------------------------------------