Why 3 types of number? about json HOT 4 CLOSED

Json doesn't define what a number means, so we err here on flexibility in order to avoid losing precision. If we only had a single f64 for our numbers, it could only precisely represent the integers up to -253 to 253. This protects us from, say, someone using a random u64 and an id.

One downside with our approach is that serde doesn't yet provide a good model for arbitrarily sized integers, so we cut things off at 64 bytes. If I recall correctly, this is the same behavior as RapidJSON and the various java parsers so I figured it wouldn't be a significant impact.

from json.

Ok, thanks! 😊

from json.

I am not convinced by the justification. I filed #57 to follow up after #18 is resolved.

from json.

This was fixed in #188 / serde_json 0.9.0. Now there is one serde_json::Value::Number variant.

from json.