JSON parsing fails: json-parse-error "too big integer"

Question

I'm trying to parse a JSON file which contains large numbers:

    {
      "uuid": "c73b470a-5efb-4d53-9ac6-c5f6487f227b",
      "description": "grains on square 64",
      "property": "square",
      "input": {
        "square": 64
      },
      "expected": 9223372036854775808
    },

I'm loading the file into a temp buffer and then I'm calling (json-parse-buffer), but the following error happens:

(json-parse-error "too big integer near '9223372036854775808'" "" 72 41 1992)

Supposedly Emacs supports integers of arbitrary size, but I guess the parser function doesn't support it? Is there any way to properly parse a JSON like this in Emacs?

My Emacs version:

GNU Emacs 28.2 (build 2, x86_64-pc-linux-gnu, GTK+ Version 3.24.20, cairo version 1.16.0) of 2023-03-20

Answer 1

Emacs supports bignums, but presumably libjansson does not (and json-parse-buffer is handing off to that C library to do the heavy lifting). There seems to be an upstream issue for that:
https://github.com/akheron/jansson/issues/69

The documentation for the current implementation is:
https://jansson.readthedocs.io/en/latest/conformance.html#numbers

From those docs, you are limited to either to long or long long integers, depending on the compiler which built the library. A long will be a minimum of 32-bits, and a long long will be a minimum of 64-bits.

I would expect that you're dealing with the latter, and my impression is that a long long tends to be no more than 64-bits, so that's (probably) a maximum of 18446744073709551615 unsigned, or 9223372036854775807 signed (and looking at your data, it's not a coincidence that your example failure is 1 higher than that number).

The older elisp-only (json-read) is happy reading that file, FYI.

(require 'json)
(url-handler-mode 1)
(json-read-file "https://raw.githubusercontent.com/exercism/problem-specifications/f771129746264394c20d3c59ba5053acb9de5f86/exercises/grains/canonical-data.json")

That even works in Emacs 26 (i.e. without bignum support), although in that case numbers larger than most-positive-fixnum are converted to floats.