mflog 0.0.2

opinionated python (structured) logging library built on structlog

What is it ?

It is an opinionated python (structured) logging library built on structlog for the MetWork Framework (but it can be used in any context).

• Coverage

Structured logging means that you don’t write hard-to-parse and hard-to-keep-consistent prose in your logs but that you log events that happen in a context instead.

• https://www.structlog.org/en/stable/why.html

Example:

from mflog import get_logger

# Get a logger
log = get_logger("foo.bar")

# Bind some attributes to the logger depending on the context
log = log.bind(user="john")
log = log.bind(user_id=123)

# [...]

# Log something
log.warning("user logged in", happy=True, another_key=42)

On stderr, you will get:

2019-01-28T07:52:42.903067Z  [WARNING] (foo.bar#7343) user logged in {another_key=42 happy=True user=john user_id=123}

On json output file, you will get:

{
    "timestamp": "2019-01-28T08:16:40.047710Z",
    "level": "warning",
    "name": "foo.bar",
    "pid": 29317,
    "event": "user logged in",
    "another_key": 42,
    "happy": true,
    "user": "john",
    "user_id": 123
}

(opinionated) Choices and Features

• we use main ideas from structlog library
• we log [DEBUG] and [INFO] messages on stdout (in a human friendly way)
• we log [WARNING], [ERROR] and [CRITICAL] on stderr (in a human friendly way)
• (and optionally) we log all messages (worse than a minimal configurable level) in a configurable file in JSON (for easy automatic parsing)
• (and optionally) we send all messages (worse than a minimal configurable level) to an UDP syslog server (in JSON or in plain text)
• we can configure a global minimal level to ignore all messages below
• we reconfigure automatically python standard logging library to use mflog
• Unicode and Bytes messages are supported (in Python2 and Python3)
• good support for exceptions (with backtraces)
• override easily minimal levels (for patterns of logger names) programmatically or with plain text configuration files

How to use ?

A mflog logger can be used as a standard logging logger.

For example:

# Import
from mflog import get_logger

# Get a logger
x = get_logger("foo.bar")

# Usage
x.warning("basic message")
x.critical("message with templates: %i, %s", 2, "foo")
x.debug("message with key/values", foo=True, bar="string")

try:
    1/0
except Exception:
    x.exception("we catched an exception with automatic traceback")

x = log.bind(context1="foo")
x = log.bind(context2="bar")
x.info("this is a contexted message", extra_var=123)

How to configure ?

In python

import mflog

# Configure
mflog.set_config(minimal_level="DEBUG", json_minimal_level="WARNING",
                 json_file="/foo/bar/my_output.json")

# Get a logger
x = mflog.get_logger("foo.bar")

# [...]

With environment variables

$ export MFLOG_MINIMAL_LEVEL="DEBUG"
$ export MFLOG_JSON_MINIMAL_LEVEL="WARNING"
$ export MFLOG_JSON_FILE="/foo/bar/my_output.json"

$ python

>>> import mflog
>>>
>>> # Get a logger
>>> x = mflog.get_logger("foo.bar")
>>>
>>> # [...]

Note

When you get a mflog logger, if default configuration is applied automatically if not set manually before.

How to override minimal level for a specific logger

If you have a "noisy" specific logger, you can override its minimal log level.

The idea is to configure this in a file like this:

# lines beginning with # are comments

# this line say 'foo.bar' logger will have a minimal level of WARNING
foo.bar => WARNING

# this line say 'foo.*' loggers will have a minimal level of DEBUG
# (see python fnmatch for accepted wildcards)
foo.* => DEBUG

# The first match wins

Then, you can use

# yes we use a list here because you can use several files
# (the first match wins)
mflog.configure([...], override_files=["/full/path/to/your/override.conf"])

or

# if you want to provide multiple files, use ';' as a separator
export MFLOG_MINIMAL_LEVEL_OVERRIDE_FILES=/full/path/to/your/override.conf

Link with standard python logging library

When you get a mflog logger or when you call set_config() function, the standard python logging library is reconfigured to use mflog.

Example:

import logging
import mflog

# standard use of logging library
x = logging.getLogger("standard.logger")
print("<output of the standard logging library>")
x.warning("foo bar")
print("</output of the standard logging library>")

# we set the mflog configuration
mflog.set_config()

# now logging library use mflog
print()
print("<output of the standard logging library through mflog>")
x.warning("foo bar")
print("</output of the standard logging library through mflog>")

Output:

<output of the standard logging library>
foo bar
</output of the standard logging library>

<output of the standard logging library through mflog>
2019-01-29T09:32:37.093240Z  [WARNING] (standard.logger#15809) foo bar
</output of the standard logging library through mflog>

mflog loggers API

.debug(message, *args, **kwargs)

Log the given message as [DEBUG].

• *args can be used for placeholders (to format the given message)
• **kwargs can be used for key/values (log context).

Examples:

from mflog import get_logger

x = get_logger('my.logger')
x.debug("my debug message with placeholders: %s and %i", "foo", 123,
        key1="value1, key2=True, key5=123)

.info(message, *args, **kwargs)

Same as .debug but with [INFO] severity level.

.warning(message, *args, **kwargs)

Same as .debug but with [WARNING] severity level.

.error(message, *args, **kwargs)

Same as .debug but with [ERROR] severity level.

.critical(message, *args, **kwargs)

Same as .debug but with [CRITICAL] severity level.

.exception(message, *args, **kwargs)

Same as .error (so with [ERROR] severity level) but we automatically add the current stacktrace in the message through special key/values.

.bind(**new_values)

Return a new logger with **new_values added to the existing ones (see examples at the beginning).

.unbind(*keys)

Return a new logger with *keys removed from the context. It raises KeyError if the key is not part of the context.

.try_unbind(*keys)

Like .unbind but best effort: missing keys are ignored.

FAQ

If I want to use mflog inside my library ?

If you write a library and if you want to use mflog, use mflog normally. You just should avoid to call set_config() inside your library.

Do you have "thread local context mode" ?

This mode is explained here.

You have to understand what you are doing.

If you want to use it, just add thread_local_context=True to your set_config() call. And you can use .new(**new_values) on mflog loggers to clear context and binds some initial values.

Can I globally add an extra context to each log line ?

If you add extra_context_func=your_python_func to your set_config() call, and if your_python_func returns a dict of key/values as strings when called with no argument, these key/values will be added to your log context.

Another way to do that without even calling set_config() is to define an environment variable called MFLOG_EXTRA_CONTEXT_FUNC containing the full path to your python func.

Full example:

# in shell
export MFLOG_EXTRA_CONTEXT_FUNC="mflog.unittests.extra_context"

then, in your python interpreter:

>>> from mflog import get_logger
>>> get_logger("foo").info("bar")
2019-04-11T07:32:53.517260Z     [INFO] (foo#15379) bar {extra_context_key1=extra_context_value1 extra_context_key2=extra_context_value2}

Here is the code of mflog.unittests.extra_context:

def extra_context():
    return {"extra_context_key1": "extra_context_value1",
            "extra_context_key2": "extra_context_value2"}

Can I filter some context keys in stdout/stderr output (but keep them in json output) ?

Yes, add json_only_keys=["key1", "key2"] to your set_config() call or use MFLOG_JSON_ONLY_KEYS=key1,key2 environment variable.

What about if I don't want to redirect standard python logging to mflog ?

You can add standard_logging_redirect=False in your set_config() call of set MFLOG_STANDARD_LOGGING_REDIRECT=0 environment variable.

Can I silent a specific noisy logger?

You can use override_files feature to do that or you can also use the mflog.add_override function.

For example:

import mflog

# for all mylogger.* loggers (fnmatch pattern), the minimal level is CRITICAL
mflog.add_override("mylogger.*", CRITICAL)

# Not very interesting but this call will be ignored
mflog.get_logger("mylogger.foo").warning("foo")

How can I use syslog logging?

You can configure it with these keyword arguments during set_config() call:

• syslog_minimal_level: WARNING, CRITICAL...
• syslog_address: null (no syslog (defaut)), 127.0.0.1:514 (send packets to 127.0.0.1:514), /dev/log (unix socket)...
• syslog_format: msg_only (default) or json

or with corresponding env vars:

• MFLOG_SYSLOG_MINIMAL_LEVEL
• MFLOG_SYSLOG_ADDRESS
• MFLOG_SYSLOG_FORMAT