ilm...@ilmari.org (Dagfinn Ilmari Mannsåker) writes:
> Tom Lane <t...@sss.pgh.pa.us> writes:
>
>> ilm...@ilmari.org (Dagfinn Ilmari =?utf-8?Q?Manns=C3=A5ker?=) writes:
>>> While playing around some more with the extension, I discoverered a few
>>> more issues:
>>> ...
>>> 4) SV_to_JsonbValue() throws an error for infinite NVs, but not NaNs
>>
>> The others sound like bugs, but that one's intentional, since type
>> numeric does have a concept of NaN. If you're arguing that we should
>> disallow that value in the context of jsonb, maybe so, but it'd likely
>> take changes in quite a few more places than here.
>
> The numeric type that's used internally to represent numbers in jsonb
> might have the concept of NaN, but JSON itself does not:
>
> Numeric values that cannot be represented in the grammar below (such
> as Infinity and NaN) are not permitted.
>
> - https://tools.ietf.org/html/rfc7159#section-6
[…]
> =# create or replace function jsonbnan() returns jsonb immutable language
> plperlu transform for type jsonb as '0+"NaN"';
> CREATE FUNCTION
[…]
> =# select jsonbnan()::json;
> ERROR: invalid input syntax for type json
> DETAIL: Token "NaN" is invalid.
> CONTEXT: JSON data, line 1: NaN
Also, it doesn't parse back in as jsonb either:
=# select jsonbnan()::text::json;
ERROR: invalid input syntax for type json
DETAIL: Token "NaN" is invalid.
CONTEXT: JSON data, line 1: NaN
And it's inconsistent with to_jsonb():
=# select to_jsonb('nan'::numeric);
┌──────────┐
│ to_jsonb │
├──────────┤
│ "NaN" │
└──────────┘
It would be highly weird if PL transforms (jsonb_plpython does the same
thing) let you create spec-violating jsonb values that don't round-trip
via jsonb_out/in.
- ilmari
--
"The surreality of the universe tends towards a maximum" -- Skud's Law
"Never formulate a law or axiom that you're not prepared to live with
the consequences of." -- Skud's Meta-Law
