Add support to custom `JSONEncoder`

[thiromi]

Is your feature request related to a problem? Please describe.
My company is moving out from structlog to start using this package, as it is intended to have better integration with google products. We're using StructuredLogHandler to send the log records to google cloud logging. But something structlog used to do out-of-the-box was the ability to automatically convert the extra data we provide to the logger to be sent to the stream, and therefore, we need to check all the calls we're making to logger.info(), debug(), warning(), etc. to check if the data type is supported by JSONEncoder, otherwise the call to hadler.format() will raise TypeError, stating that Object of type PosixPath is not JSON serializable (this is one example of error we're getting, because the default encoder cannot encode pathlib.Path instances) and the log record will not be pushed to gcloud logging
Describe the solution you'd like
I'd like to be able to specify a custom class to encode the message to JSON, so the formatting is handled by the encoder, not the caller of logger.info() and its siblings
Describe alternatives you've considered
An alternative would be having a filter that prepares the data to be used by the handler, I'm even wrote a filter to handle that:

class JsonableConverterFilter(logging.Filter):
 """Prepare log record to be JSON encoded

 :class:`google.cloud.logging_v2.handlers.StructuredLogHandler` encodes directly to default JSON encoder without
 the possibility to set up a custom encoder, so we need to hydrate the log record to not break when emitting
 (because broken logs break silently)

 The ideal solution would be to have a LogFormatter, but as GCloud's log handler formats to JSON inside the handler,
 we need to circumvent this issue with a :class:`logging.Filter`
 """

 def filter(self, record: logging.LogRecord) -> bool:
 fields: Dict[str, Any] = getattr(record, "json_fields", {})

 for key, value in fields.items():
 fields[key] = JsonableConverterFilter.prepare_field(value)

 return True

 @classmethod
 def prepare_field(cls, value: Any) -> TypeJSON:
 """Converts the provided type into a plain JSON type"""
 # basic types
 if isinstance(value, (int, str, float, bool)) or value is None:
 return value

 try:
 iterator = iter(value)
 except TypeError:
 # probably a custom class
 if not hasattr(value, "__dict__"):
 return str(value)

 iterator = iter(value.__dict__)

 # hashmaps
 if hasattr(value, "keys") and hasattr(value, "values"):
 return {key: cls.prepare_field(val) for key, val in value.items()}

 # sequences
 return [cls.prepare_field(item) for item in iterator]

but it seems to me I'm deviating the purpose of log filters by doing that
Additional context
n/a for now

[daniel-sanche]

So to be clear, you are creating JSON logs with dictionary values using something like this:

logging.info("message", extra={"json_fields":{"a": pathlib.Path("test")})

But are having issues because some items in json_fields aren't automatically serializable, so you'd like to be able to pass in a custom JSONEncoder object as part of the library initialization. Is that correct?

[thiromi]

Hi there, thank you for your reply. Yes, that's exactly my use case

…

On Tue, Oct 25, 2022, 23:37 Daniel Sanche ***@***.***> wrote: So to be clear, you are creating JSON logs with dictionary values using something like this: logging.info("message", extra={"json_fields":{"a": pathlib.Path("test")}) But are having issues because some items in json_fields aren't automatically serializable, so you'd like to be able to pass in a custom JSONEncoder object as part of the library initialization. Is that correct? — Reply to this email directly, view it on GitHub <#656 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAAUKXZBD2F25FRO5VH7O43WFBHKVANCNFSM6AAAAAARN3HXOA> . You are receiving this because you authored the thread.Message ID: ***@***.***>