Re: JSON validation behavior

"David G. Johnston" <david(dot)g(dot)johnston(at)gmail(dot)com> writes:
> The following does seem buggy though:
> select json '{ "a": "null \u0000 escape"}' -> 'a' as fails;

Perhaps, but AFAICS it's entirely accidental that any variant of this
doesn't fail. Although this manages not to fail:

regression=# select json '{ "a": "null \u0000 escape"}' ;
json
------------------------------
{ "a": "null \u0000 escape"}
(1 row)

this does:

regression=# select jsonb '{ "a": "null \u0000 escape"}' ;
ERROR: unsupported Unicode escape sequence

The error message is actually being thrown in the JSON parser, and it
only doesn't get thrown if the parser knows that it's just error-checking
the data and not producing any converted output (cf the
"if (lex->strval != NULL)" block starting at json.c:832). It actually
seems to me that this behavior is a bug, in that there are a bunch of
error checks there (not only this one) that are skipped in the allegedly
"error checking only" path.

But anyway, making it work as suggested here would take a substantial
amount of refactoring, and it would not (I think) work anyway for jsonb,
so it doesn't quite seem worth a lot of work. I could get behind fixing
it to always throw the error, but that's not what Sergei was hoping for.

regards, tom lane