Standard Library Logging#

Ideally, structlog should be able to be used as a drop-in replacement for standard library’s logging by wrapping it. In other words, you should be able to replace your call to logging.getLogger by a call to structlog.get_logger and things should keep working as before (if structlog is configured right, see Suggested Configurations below).

If you run into incompatibilities, it is a bug so please take the time to report it! If you’re a heavy logging user, your help to ensure a better compatibility would be highly appreciated!

Just Enough logging#

If you want to use structlog with logging, you should have at least a fleeting understanding on how the standard library operates because structlog will not do any magic things in the background for you. Most importantly you have to configure the logging system additionally to configuring structlog.

Usually it is enough to use:

import logging
import sys

logging.basicConfig(
    format="%(message)s",
    stream=sys.stdout,
    level=logging.INFO,
)

This will send all log messages with the log level logging.INFO and above (that means that e.g. logging.debug calls are ignored) to standard out without any special formatting by the standard library.

If you require more complex behavior, please refer to the standard library’s logging documentation.

Concrete Bound Logger#

To make structlog’s behavior less magicy, it ships with a standard library-specific wrapper class that has an explicit API instead of improvising: structlog.stdlib.BoundLogger. It behaves exactly like the generic structlog.BoundLogger except:

  • it’s slightly faster due to less overhead,

  • has an explicit API that mirrors the log methods of standard library’s logging.Logger,

  • it has correct type hints,

  • hence causing less cryptic error messages if you get method names wrong.


If you’re using static types (e.g. with Mypy) you also may want to use structlog.stdlib.get_logger() that has the appropriate type hints if you’re using structlog.stdlib.BoundLogger. Please note though, that it will neither configure nor verify your configuration. It will call structlog.get_logger() just like if you would’ve called it – the only difference are the type hints.

asyncio#

For asyncio applications, you may not want your whole application to block while your processor chain is formatting your log entries. For that use case structlog comes with structlog.stdlib.AsyncBoundLogger that will do all processing in a thread pool executor.

This means an increased computational cost per log entry but your application will never block because of logging.

To use it, configure structlog to use AsyncBoundLogger as wrapper_class.

Processors#

structlog comes with a few standard library-specific processors:

render_to_log_kwargs:

Renders the event dictionary into keyword arguments for logging.log that attaches everything except the event field to the extra argument. This is useful if you want to render your log entries entirely within logging.

filter_by_level:

Checks the log entry’s log level against the configuration of standard library’s logging. Log entries below the threshold get silently dropped. Put it at the beginning of your processing chain to avoid expensive operations from happening in the first place.

add_logger_name:

Adds the name of the logger to the event dictionary under the key logger.

ExtraAdder:

Add extra attributes of logging.LogRecord objects to the event dictionary.

This processor can be used for adding data passed in the extra parameter of the logging module’s log methods to the event dictionary.

add_log_level():

Adds the log level to the event dictionary under the key level.

add_log_level_number:

Adds the log level number to the event dictionary under the key level_number. Log level numbers map to the log level names. The Python stdlib uses them for filtering logic. This adds the same numbers so users can leverage similar filtering. Compare:

level in ("warning", "error", "critical")
level_number >= 30

The mapping of names to numbers is in structlog.stdlib._NAME_TO_LEVEL.

PositionalArgumentsFormatter:

This processes and formats positional arguments (if any) passed to log methods in the same way the logging module would do, e.g. logger.info("Hello, %s", name).

structlog also comes with ProcessorFormatter which is a logging.Formatter that enables you to format non-structlog log entries using structlog renderers and multiplex structlog’s output with different renderers (see below for an example).

Suggested Configurations#

Note

We do appreciate that fully integrating structlog with standard library’s logging is fiddly when done for the first time.

This is the price of flexibility and unfortunately – given the different needs of our users – we can’t make it any simpler without compromising someone’s use-cases. However, once it is set up, you can rely on not having to ever touch it again.

Depending where you’d like to do your formatting, you can take one of three approaches:

Rendering Within structlog#

This is the simplest approach where structlog does all the heavy lifting and passes a fully-formatted string to logging. Chances are, this is all you need.

flowchart TD %%{ init: {'theme': 'neutral'} }%% User structlog stdlib[Standard Library\ne.g. logging.StreamHandler] User --> |"structlog.get_logger().info('foo')"| structlog User --> |"logging.getLogger().info('foo')"| stdlib structlog --> |"logging.getLogger().info(#quot;{'event': 'foo'}#quot;)"| stdlib ==> Output Output

A basic configuration to output structured logs in JSON format looks like this:

import structlog

structlog.configure(
    processors=[
        # If log level is too low, abort pipeline and throw away log entry.
        structlog.stdlib.filter_by_level,
        # Add the name of the logger to event dict.
        structlog.stdlib.add_logger_name,
        # Add log level to event dict.
        structlog.stdlib.add_log_level,
        # Perform %-style formatting.
        structlog.stdlib.PositionalArgumentsFormatter(),
        # Add a timestamp in ISO 8601 format.
        structlog.processors.TimeStamper(fmt="iso"),
        # If the "stack_info" key in the event dict is true, remove it and
        # render the current stack trace in the "stack" key.
        structlog.processors.StackInfoRenderer(),
        # If the "exc_info" key in the event dict is either true or a
        # sys.exc_info() tuple, remove "exc_info" and render the exception
        # with traceback into the "exception" key.
        structlog.processors.format_exc_info,
        # If some value is in bytes, decode it to a unicode str.
        structlog.processors.UnicodeDecoder(),
        # Add callsite parameters.
        structlog.processors.CallsiteParameterAdder(
            {
                structlog.processors.CallsiteParameter.FILENAME,
                structlog.processors.CallsiteParameter.FUNC_NAME,
                structlog.processors.CallsiteParameter.LINENO,
            }
        ),
        # Render the final event dict as JSON.
        structlog.processors.JSONRenderer()
    ],
    # `wrapper_class` is the bound logger that you get back from
    # get_logger(). This one imitates the API of `logging.Logger`.
    wrapper_class=structlog.stdlib.BoundLogger,
    # `logger_factory` is used to create wrapped loggers that are used for
    # OUTPUT. This one returns a `logging.Logger`. The final value (a JSON
    # string) from the final processor (`JSONRenderer`) will be passed to
    # the method of the same name as that you've called on the bound logger.
    logger_factory=structlog.stdlib.LoggerFactory(),
    # Effectively freeze configuration after creating the first bound
    # logger.
    cache_logger_on_first_use=True,
)

To make your program behave like a proper 12 factor app that outputs only JSON to stdout, configure the logging module like this:

import logging
import sys

logging.basicConfig(
    format="%(message)s",
    stream=sys.stdout,
    level=logging.INFO,
)

In this case only your own logs are formatted as JSON:

>>> structlog.get_logger("test").warning("hello")
{"event": "hello", "logger": "test", "level": "warning", "timestamp": "2017-03-06T07:39:09.518720Z"}

>>> logging.getLogger("test").warning("hello")
hello

Rendering Using logging-based Formatters#

You can choose to use structlog only for building the event dictionary and leave all formatting – additionally to the output – to the standard library.

flowchart TD %%{ init: {'theme': 'neutral'} }%% User structlog stdlib[Standard Library\ne.g. logging.StreamHandler] User --> |"structlog.get_logger().info('foo', bar=42)"| structlog User --> |"logging.getLogger().info('foo')"| stdlib structlog --> |"logging.getLogger().info('foo', extra={"bar": 42})"| stdlib ==> Output Output
import structlog

structlog.configure(
    processors=[
        structlog.stdlib.filter_by_level,
        structlog.stdlib.add_logger_name,
        structlog.stdlib.add_log_level,
        structlog.stdlib.PositionalArgumentsFormatter(),
        structlog.processors.StackInfoRenderer(),
        structlog.processors.format_exc_info,
        structlog.processors.UnicodeDecoder(),
        # Transform event dict into `logging.Logger` method arguments.
        # "event" becomes "msg" and the rest is passed as a dict in
        # "extra". IMPORTANT: This means that the standard library MUST
        # render "extra" for the context to appear in log entries! See
        # warning below.
        structlog.stdlib.render_to_log_kwargs,
    ],
    logger_factory=structlog.stdlib.LoggerFactory(),
    wrapper_class=structlog.stdlib.BoundLogger,
    cache_logger_on_first_use=True,
)

Now you have the event dict available within each log record. If you want all your log entries (i.e. also those not from your app/structlog) to be formatted as JSON, you can use the python-json-logger library:

import logging
import sys

from pythonjsonlogger import jsonlogger

handler = logging.StreamHandler(sys.stdout)
handler.setFormatter(jsonlogger.JsonFormatter())
root_logger = logging.getLogger()
root_logger.addHandler(handler)

Now both structlog and logging will emit JSON logs:

>>> structlog.get_logger("test").warning("hello")
{"message": "hello", "logger": "test", "level": "warning"}

>>> logging.getLogger("test").warning("hello")
{"message": "hello"}

Warning

With this approach, it’s the standard library logging formatter’s duty to do something useful with the event dict. In the above example that’s jsonlogger.JsonFormatter.

Keep this in mind if you only get the event name without any context, and exceptions are ostensibly swallowed.

Rendering Using structlog-based Formatters Within logging#

Finally, the most ambitious approach. Here, you use structlog’s ProcessorFormatter as a logging.Formatter for both logging as well as structlog log entries.

Consequently, the output is the duty of the standard library too.

flowchart TD %%{ init: {'theme': 'neutral'} }%% User structlog structlog2[structlog] stdlib["Standard Library"] User --> |"structlog.get_logger().info(#quot;foo#quot;, bar=42)"| structlog User --> |"logging.getLogger().info(#quot;foo#quot;)"| stdlib structlog --> |"logging.getLogger().info(event_dict, {#quot;extra#quot;: {#quot;_logger#quot;: logger, #quot;_name#quot;: name})"| stdlib stdlib --> |"structlog.stdlib.ProcessorFormatter.format(logging.Record)"| structlog2 structlog2 --> |"Returns a string that is passed into logging handlers.\nThis flow is controlled by the logging configuration."| stdlib2 stdlib2[Standard Library\ne.g. logging.StreamHandler] ==> Output

ProcessorFormatter has two parts to its API:

  1. On the structlog side, the processor chain must be configured to end with structlog.stdlib.ProcessorFormatter.wrap_for_formatter as the renderer. It converts the processed event dictionary into something that ProcessorFormatter understands.

  2. On the logging side, you must configure ProcessorFormatter as your formatter of choice. logging then calls ProcessorFormatter’s format() method.

    For that, ProcessorFormatter wraps a processor chain that is responsible for rendering your log entries to strings.

Thus, the simplest possible configuration looks like the following:

import logging
import structlog

structlog.configure(
    processors=[
        # Prepare event dict for `ProcessorFormatter`.
        structlog.stdlib.ProcessorFormatter.wrap_for_formatter,
    ],
    logger_factory=structlog.stdlib.LoggerFactory(),
)

formatter = structlog.stdlib.ProcessorFormatter(
    processors=[structlog.dev.ConsoleRenderer()],
)

handler = logging.StreamHandler()
# Use OUR `ProcessorFormatter` to format all `logging` entries.
handler.setFormatter(formatter)
root_logger = logging.getLogger()
root_logger.addHandler(handler)
root_logger.setLevel(logging.INFO)

which will allow both of these to work in other modules:

>>> import logging
>>> import structlog

>>> logging.getLogger("stdlog").info("woo")
woo      _from_structlog=False _record=<LogRecord:...>
>>> structlog.get_logger("structlog").info("amazing", events="oh yes")
amazing  _from_structlog=True _record=<LogRecord:...> events=oh yes

Of course, you probably want timestamps and log levels in your output. The ProcessorFormatter has a foreign_pre_chain argument which is responsible for adding properties to events from the standard library – i.e. that do not originate from a structlog logger – and which should in general match the processors argument to structlog.configure so you get a consistent output.

_from_structlog and _record allow your processors to determine whether the log entry is coming from structlog, and to extract information from logging.LogRecords and add them to the event dictionary. However, you probably don’t want to have them in your log files, thus we’ve added the ProcessorFormatter.remove_processors_meta processor to do so conveniently.

For example, to add timestamps, log levels, and traceback handling to your logs without _from_structlog and _record noise you should do:

timestamper = structlog.processors.TimeStamper(fmt="%Y-%m-%d %H:%M:%S")
shared_processors = [
    structlog.stdlib.add_log_level,
    timestamper,
]

structlog.configure(
    processors=shared_processors + [
        structlog.stdlib.ProcessorFormatter.wrap_for_formatter,
    ],
    logger_factory=structlog.stdlib.LoggerFactory(),
    cache_logger_on_first_use=True,
)

formatter = structlog.stdlib.ProcessorFormatter(
    # These run ONLY on `logging` entries that do NOT originate within
    # structlog.
    foreign_pre_chain=shared_processors,
    # These run on ALL entries after the pre_chain is done.
    processors=[
       # Remove _record & _from_structlog.
       structlog.stdlib.ProcessorFormatter.remove_processors_meta,
       structlog.dev.ConsoleRenderer(),
     ],
)

which (given the same logging.* calls as in the previous example) will result in:

>>> logging.getLogger("stdlog").info("woo")
2021-11-15 11:41:47 [info     ] woo
>>> structlog.get_logger("structlog").info("amazing", events="oh yes")
2021-11-15 11:41:47 [info     ] amazing    events=oh yes

This allows you to set up some sophisticated logging configurations. For example, to use the standard library’s logging.config.dictConfig to log colored logs to the console and plain logs to a file you could do:

import logging.config
import structlog

timestamper = structlog.processors.TimeStamper(fmt="%Y-%m-%d %H:%M:%S")
pre_chain = [
    # Add the log level and a timestamp to the event_dict if the log entry
    # is not from structlog.
    structlog.stdlib.add_log_level,
    # Add extra attributes of LogRecord objects to the event dictionary
    # so that values passed in the extra parameter of log methods pass
    # through to log output.
    structlog.stdlib.ExtraAdder(),
    timestamper,
]

def extract_from_record(_, __, event_dict):
    """
    Extract thread and process names and add them to the event dict.
    """
    record = event_dict["_record"]
    event_dict["thread_name"] = record.threadName
    event_dict["process_name"] = record.processName

    return event_dict

logging.config.dictConfig({
        "version": 1,
        "disable_existing_loggers": False,
        "formatters": {
            "plain": {
                "()": structlog.stdlib.ProcessorFormatter,
                "processors": [
                   structlog.stdlib.ProcessorFormatter.remove_processors_meta,
                   structlog.dev.ConsoleRenderer(colors=False),
                ],
                "foreign_pre_chain": pre_chain,
            },
            "colored": {
                "()": structlog.stdlib.ProcessorFormatter,
                "processors": [
                   extract_from_record,
                   structlog.stdlib.ProcessorFormatter.remove_processors_meta,
                   structlog.dev.ConsoleRenderer(colors=True),
                ],
                "foreign_pre_chain": pre_chain,
            },
        },
        "handlers": {
            "default": {
                "level": "DEBUG",
                "class": "logging.StreamHandler",
                "formatter": "colored",
            },
            "file": {
                "level": "DEBUG",
                "class": "logging.handlers.WatchedFileHandler",
                "filename": "test.log",
                "formatter": "plain",
            },
        },
        "loggers": {
            "": {
                "handlers": ["default", "file"],
                "level": "DEBUG",
                "propagate": True,
            },
        }
})
structlog.configure(
    processors=[
        structlog.stdlib.add_log_level,
        structlog.stdlib.PositionalArgumentsFormatter(),
        timestamper,
        structlog.processors.StackInfoRenderer(),
        structlog.processors.format_exc_info,
        structlog.stdlib.ProcessorFormatter.wrap_for_formatter,
    ],
    logger_factory=structlog.stdlib.LoggerFactory(),
    wrapper_class=structlog.stdlib.BoundLogger,
    cache_logger_on_first_use=True,
)

This defines two formatters: one plain and one colored. Both are run for each log entry. Log entries that do not originate from structlog, are additionally pre-processed using a cached timestamper and add_log_level().

Additionally, for both logging and structlog – but only for the colorful logger – we also extract some data from logging.LogRecord:

>>> logging.getLogger().warning("bar")
2021-11-15 13:26:52 [warning  ] bar    process_name=MainProcess thread_name=MainThread

>>> structlog.get_logger("structlog").warning("foo", x=42)
2021-11-15 13:26:52 [warning  ] foo    process_name=MainProcess thread_name=MainThread x=42

>>> pathlib.Path("test.log").read_text()
2021-11-15 13:26:52 [warning  ] bar
2021-11-15 13:26:52 [warning  ] foo    x=42

(Sadly, you have to imagine the colors in the first two outputs.)

If you leave foreign_pre_chain as None, formatting will be left to logging. Meaning: you can define a format for ProcessorFormatter too!