Support parsing numbers representable by int64_t but not by number as BigInts in returned messages of StructuredIrStreams.

[junhaoliao]

Bug

1. The nlohmann/json library, which clp-ffi uses to serialize kv-pair log events into JSON, acknowledges there are implementations whose number types can store up only to 2^53-1 : https://json.nlohmann.me/features/types/number_handling/#number-interoperability; however, it does serialize numbers out of that range: https://github.com/nlohmann/json/blob/4a602df34e71afca2242b34c743d4cc2486cd071/tests/src/unit-serialization.cpp#L168-L183
2. When numbers in such ranges are parsed by JSON.parse(), loss of precision is expected. Below sample code illustrates the problem:

   const myNumStr = "9007199254740993"; // 2^53 + 1
   console.log(JSON.parse(myNumStr)); // 9007199254740992

3. Therefore, it is expected that numbers in that range, when returned as part of the messages by clp-ffi-js, cannot be accurately represented in the formatted messages displayed in the log-viewer.

Potential solutions

1. Override the serialization / deserialization behaviours for big numbers:

   const myNum = 1234567890123456789n
   
   BigInt.prototype.toJSON = function () {
       return JSON.rawJSON(this.toString());
   };
   
   const jsonStr = JSON.stringify(myNum)
   
   console.log(jsonStr)
   
   console.log(JSON.parse(jsonStr)) // loss of precision: 1234567890123456800
   const isNumberRepresentable = value => {
     const num = Number(value);
     return (
       false === isNaN(num) &&
       Number.isInteger(num) &&
       num <= Number.MAX_SAFE_INTEGER &&
       num >= Number.MIN_SAFE_INTEGER
     );
   };
   
   console.log(JSON.parse(jsonStr, (_, __, context) => {
       if (false === isNumberRepresentable(context.source)) {
           return BigInt(context.source)
       }
       return value;
   })) // 1234567890123456789n
   

2. Instead of JSON, return the decoded structured log event in a byte-string format with another serialization method. e.g., MessagePack.

yscope-log-viewer version

#85

Environment

The sample code was run with Node.js 22 which uses the V8 engine which is also used by any Chromium-based browsers such as Google Chrome and Microsoft Edge.

Reproduction steps

1. Run above sample code with Node.js 22 and observe the loss of precision in the console output.