Truncation of significant digits in floating point numbers in json formatting

[bogatikov]

When a JSON body contains a float value like 0.40000000000000002, the log records it as 0.4.
This occurs when the content type is application/json and the request is processed by JacksonJsonFieldBodyFilter.

Content-Type: application/json

{
  "amount": 0.40000000000000002
}

Log output:

{
  "origin": "remote",
  "type": "request",
  "correlation": "9c6416830b522967",
  "protocol": "HTTP/1.1",
  "remote": "/127.0.0.1:53879",
  "method": "POST",
  "uri": "http://localhost:8080/helloworld",
  "host": "localhost",
  "path": "/helloworld",
  "scheme": "http",
  "port": "8080",
  "headers": {
    ....
    "Content-Type": [
      "application/json"
    ],
    ....
  },
  "body": {
    "amount": 0.4
  }
}

The original float value is truncated, making the logs inconsistent with the actual request. This can lead to confusion when analyzing logs, as the exact input is not preserved

Expected Behavior

The log should retain and display the original float value exactly as it appeared in the request body, without altering the data.

Actual Behavior

Float values are truncated to 16 significant digits when logged

Possible Fix

The issue stems from how Jackson processes and copies JSON tokens. When Jackson encounters a float value, it parses it as a double, resulting in a loss of precision:

generator.copyCurrentEvent(parser);

Starting from Jackson 2.15+, the following method can be used to handle float values more accurately:

    public void copyCurrentEventExact(JsonParser p) throws IOException

This method treats float values as BigDecimal, preserving their full precision.

Note:
• This change may impact performance.
• It is recommended to make this behavior configurable, allowing users to choose between exact precision or standard copying for performance reasons.

Steps to Reproduce

1. Send a request with a JSON body containing a float value with more than 16 significant digits.

Context

The current behavior results in misleading log entries when float values are involved, potentially affecting debugging and auditing processes

Your Environment

• Framework: Spring WebFlux
• Logging Library: Logbook 3.10.0
• Spring Versions: 3.0.4 / 3.4.0
• Jackson Versions: 2.14, 2.15, 2.18