Re: [Json] Another problematic JSON Schema use-case

[Andrew Newton <[e-mail]>]

On Wed, May 25, 2016 at 8:26 PM, Carsten Bormann <[e-mail]> wrote:
> Andrew Newton wrote:
>> It's not too much to ask, but the problem may be with the JSON parser.
>> Many parsers ingest JSON and give back native data structures, which
>> is what makes JSON simple for most people. But a validator needs
>> context that a typical JSON parser does not provide.
>
> Can you give an example for that?
>

Given the following:

[
  { "foo" : "bar" }
]

Here is what happens if I use the built-in Ruby parser:

irb(main):017:0> s
=> "\n[\n  { \"foo\":\"bar\"}\n]\n"
irb(main):018:0> JSON.parse s
=> [{"foo"=>"bar"}]

If a validator wants to ensure that "foo" is "baz" and not "bar", its
lost the information to tell the user to look at line 2.

By contrast, if a validator were to use Jackson as a stream parser, it
would have access to JsonLocation:
http://fasterxml.github.io/jackson-core/javadoc/2.1.0/com/fasterxml/jackson/core/JsonLocation.html

> (I'm familiar with a problem like this from CBOR, where CDDL maybe
> sometimes wants to make serialization choices*), which are no longer
> visible at the data model level; I just haven't seen the need with JSON.)

I'm not sure we are talking about the same thing.

-andy