Re: [Json] Proposed minimal change for duplicate names in objects

[Tatu Saloranta <[e-mail]>]

On Sun, Jul 7, 2013 at 11:43 AM, John Cowan <[e-mail]> wrote:

> Carsten Bormann scripsit:
>
> > In many JSON designs, the receiver knows the full set of keys ahead
> > of time.  In this case, you are better off with conditionally interning
> > it (i.e., if I already have the symbol, return it, otherwise keep the
> > string around, now we know that it didn't map to a known key).
>
> I don't know any way to do that in Java, because there seems to be no
> way to ask if a string is interned except by interning it.
>

This can be done if caller has a way to indicate set of expected possible
names, which may be the case for data-binding. This also means that actual
intern()ing can be done ahead of time, and potentially decode straight from
byte sequences to intern()ed Strings. Jackson does something similar,
although it goes via general symbol table as reuse coupled with
multi-level, varying vocabularies becomes rather hairy thing to manage
(multi-level symbol tables).

Anyway... this works for subset of cases; and many users use dynamically
generated names, open-ended domain.

-+ Tatu +-