Re: [Json] Call for real-world examples of how parsers deal with duplicate keys

[Stephan Beal <[e-mail]>]

On Thu, Jun 6, 2013 at 7:41 PM, Paul Hoffman <[e-mail]> wrote:

> Greetings again. Knowing what current parsers do with duplicate keys might
> be useful to this discussion. Of course, some people will exclaim "but
> that's wrong!" to some of what we see, but seeing it may be useful
> nonetheless.
>
> I propose the following JSON text for the tests:
>
> {"a":1,"a":2}
>

My C and C++ implementations would result in {"a":2}. That was the only
implementation which made intuitive sense to me (and is faster than
erroring on dupes in the C++ code, which of course uses std::map as the
basis, and avoids me having to search that map to check for a dupe). While
library-level coding effort would not be notably more to error on dupes, it
does add another error case/code for clients to check for, and it just
didn't make sense to me to deal with this (what i consider to be a) corner
case. This approach also, IMO, conforms to the "principal of least
surprise," which for me means "last one wins.'

-- 
----- stephan beal
http://wanderinghorse.net/home/stephan/
http://gplus.to/sgbeal