Custom Serializer formater, use of included log formatter

[ch0wm3in]

Hi,

Consider the following

def serialize(record):
    """Serialize the JSON log.
    """
    log = {}
    log["time"] = record["time"].timestamp()
    log["level"] = record["level"].name
    log["message"] = record["message"]
    log["func_string"] = f"{record['name']}:{record['function']}:{record['line']}"
    log["correlation_id"] = call_uuid
    log["total_runtime"] = record["elapsed"].seconds


    if record["exception"] is not None:
        error_data = {
            "error": {
                "stack": "".join(
                    traceback.format_exception(
                        record["exception"].type,
                        record["exception"].value,
                        record["exception"].traceback,
                    )
                ),
                "kind": getattr(record["exception"].type, "__name__", "None"),
                "message": str(record["exception"].value),
            },
        }
        log.update(error_data)
    
    return json.dumps(log, ensure_ascii=False)

How would i go about using the time format of "{time:YYYY-MM-DDTHH:mm:ss.SSSSSSSSSZ}" in a json serializer for the log["time"]? I see that record["time"] a datetime object, python datetime does not support nanoseconds afaik, but above works fine with stderr sink. Is there a way to set default time format globally across all sinks using the time formatter that's included in the libary?

[Delgan]

Hi.

Loguru does not support nanosecond neither.
When you use "SSSSSSSSS" in the format, it's certainly parsed as "SSSSSS" concatenated with "SSS". Basically, microseconds joined with milliseconds, that give the impression of being nanoseconds.

Maybe should I detect this pattern and raise an exception?

Anyway, if you need nanoseconds in your JSON serializer, you can use a patch function:

def add_nanoseconds(record):
    record["extra"]["nanoseconds"] = time.time_ns()

logger.configure(patcher=add_nanoseconds)

Then you can access record["extra"]["nanoseconds"] each time a message is logged.

[ch0wm3in]

Oh okay, yeah i would like an exception or at least a notion of that it's not actually nanoseconds 😁

Thanks for the supplied, i ended up accepting using rfc3339micro instead of nano with pendulum doing the following in my serializer:

import pendulum

def serialize(record):
    """Serialize the JSON log.
    """
    dt = pendulum.instance(record["time"])
    log = {}
    log["log_time"] = dt.format("YYYY-MM-DDTHH:mm:ss.SSSSSSZ")
    log["level"] = record["level"].name
    log["message"] = record["message"]
    log["func_string"] = f"{record['name']}:{record['function']}:{record['line']}"
    log["correlation_id"] = call_uuid
    log["total_runtime"] = record["elapsed"].seconds
    log["script_name"] = record["extra"].get("script_name", "")


    if record["exception"] is not None:
        error_data = {
            "error": {
                "stack": "".join(
                    traceback.format_exception(
                        record["exception"].type,
                        record["exception"].value,
                        record["exception"].traceback,
                    )
                ),
                "kind": getattr(record["exception"].type, "__name__", "None"),
                "message": str(record["exception"].value),
            },
        }
        log.update(error_data)

[Delgan]

Indeed, using pendulum would have been my suggestion, since Loguru implements the same formatting rules. 👍

I keep this ticket open and I'll try to detect pattern with more than 6 successive "S" and raise an exception in that case.