“Gibberish” false positive when parsing Unicode identifiers in objects about demjson HOT 4 OPEN

Not to mention that "gibberish" is never an appropriate way to describe texts because they're in non-Latin script.

from demjson.

I think the sanity check would do the correct thing if it was only used in strict mode, however there is a good reason to remove it there too

Indeed, ECMA-404 (since the 1st edition) and RFC 8259 both allow any value to appear at the top level of a JSON document, not just Arrays and Objects as specified in RFC 4627. This makes "あ" valid JSON text, yet it fails the sanity check.

from demjson.

That sanity check does indeed seem rather useless (any encoding errors will hopefully be caught at a later point anyways) and wrong in at least two ways: It appears to rely on the RFC's statement that "the first two characters of a JSON text will always be ASCII characters" but ignores that A) when the RFC refers to "JSON text", it means a JSON object or array, not a simple value like a string and B) non-strict mode allows arbitrary unicode in unquoted identifiers.

The point of that statement in the RFC is not to allow more-or-less meaningless sanity checks but to show a way to detect the encoding (UTF-8/16/32) based on only the first two characters, and only when the input is a JSON object or array.

I think the sanity check would do the correct thing if it was only used in strict mode, however there is a good reason to remove it there too: It completely fails its stated goal of "rais[ing] a suitably descriptive error rather than an obscure syntax error later on", since the error message indicates improper unicode encoding while the actual problem is just as likely to actually be a syntax error (e.g. trying to decode something like {あ:2} in strict mode).

from demjson.

Oh yeah, you are right. It took me a bit to look up the spec and write that up, and by the time I wrote that last part I had already forgotten that the sanity check is also used on values like strings.

EDIT: Although, the check does allow anything if the first character is a quote I believe, so it should not fail. Might have misread that part of the code though.

from demjson.