feat: moved json_dumps call to internal helper

[mj0nez]

As suggested, the call of json.dumps was moved to an internal helper in the StruclogFormatter, which now can be overwritten by inheriting classes.

I was unsure how and if you would like to add a recipe to the docs, but this would be a minimal configuration:

# log configuration
from collections.abc import Callable
from typing import Any

import orjson
import structlog

from ecs_logging import StructlogFormatter as EcsStruclogFormatter
from ecs_logging import _utils

def get_orjson_serializer() -> Callable[[dict[str, Any], bool], bytes]:

    def serializer(data: dict[str, Any], sort: bool) -> bytes:
        return orjson.dumps(
            data,
            option=orjson.OPT_SORT_KEYS if sort else None,
            default=_utils._json_dumps_fallback,
        )

    return serializer

class BinaryEcsFormatter(EcsStruclogFormatter):
    def _json_dumps(self, value: dict[str, Any]) -> bytes:
        # Ensure that the first three fields are '@timestamp',
        # 'log.level', and 'message' per ECS spec
        ordered_fields = {}
        try:
            ordered_fields["@timestamp"] = value.pop("@timestamp")
        except KeyError:
            pass

        # log.level can either be nested or not nested so we have to try both
        try:
            ordered_fields["log.level"] = value["log"].pop("level")
            if not value["log"]:  # Remove the 'log' dictionary if it's now empty
                value.pop("log", None)
        except KeyError:
            try:
                ordered_fields["log.level"] = value.pop("log.level")
            except KeyError:
                pass
        try:
            ordered_fields["message"] = value.pop("message")
        except KeyError:
            pass

        serializer = get_orjson_serializer()

        # Because we want to use 'sorted_keys=True' we manually build
        # the first three keys and then build the rest with the serializer
        if ordered_fields:

            ordered_json = serializer(ordered_fields, sort=False)
            if value:
                return ordered_json[:-1] + b"," + serializer(value, True)[1:]
            else:
                return ordered_json
        # If there are no fields with ordering requirements we
        # pass everything into the serializer
        else:
            return serializer(value, sort=True)

structlog.configure(
    processors=[
        structlog.processors.TimeStamper(key="@timestamp", fmt="iso", utc=True),
        BinaryEcsFormatter(),
    ],
    logger_factory=structlog.BytesLoggerFactory(),
)

# your script
logger = structlog.get_logger()

if __name__ == "__main__":
    logger.info("Hi there!")

Should produce a correct output like:

{"@timestamp":"2024-05-15T21:03:08.796637Z","log.level":"info","message":"Hi there!","ecs":{"version":"1.6.0"}}

Resolves #112

[cla-checker-service[bot]]

💚 CLA has been signed

[xrmx]

run docs-build

[mj0nez]

Hey, is there something I need to do to trigger the docs build?

[xrmx]

run docs-build