Question: Comparison with other Python-JSON packages

[MartinThoma]

I'm currently comparing various Python packages for handling JSON (link: cpython 3.8.6 json, simplejson, ujson, simdjson, orjson, rapidjson). I'm trying to understand their differences. I would be super happy if you could help me with that by answering some questions:

1. Is there any difference in features between the JSON packages?
2. I compared reading/dumping a 2.3MB GeoJSON and a 631KB Twitter JSON. Is there any other thing you think I should compare for benchmarking? Do you have internal benchmarks? (I've seen the README and I think I will add a JSON with many floats :-) )
3. Are you in contact with the other Python JSON package developers? Do you maybe share the way you benchmark or test cases?
4. Are you in contact with JSON package developers from other languages?
5. Are there other packages / articles for comparison I should have a look at? (I've seen your comparison with nujson)

[hugovk]

Hello! UltraJSON didn't have any releases for a long time, 2016-2020. It's now been revived but is probably best considered to be in maintenance mode, although PRs to add new features will definitely be considered. Recent additions have been precompiled wheels for many platforms (#219, #421).

1. Yes. UltraJSON aims to be on par with the stdlib json library, and for example no longer has generic serialisation of objects/iterables (see 2.0: Removed generic serialization of objects/iterables #374) like some other JSON libraries may do.
2. The benchmarking on the README is also done on the CI for all merges to master. It runs this script https://github.com/ultrajson/ultrajson/blob/master/tests/benchmark.py and you can see the most recent results here: https://github.com/ultrajson/ultrajson/runs/1138506789. Additions welcome!
3. I'm personally not in contact, and the UltraJSON benchmarking and tests are shared in this repo.
4. The benchmarking is done against nujson, orjson and simplejson.

[MartinThoma]

THank you for your answer, I appreciate it 🤗