# Copyright 2011-2019, Damian Johnson and The Tor Project
# See LICENSE for licensing information
"""
Functions to aid library logging. The default logging
:data:`~stem.util.log.Runlevel` is usually NOTICE and above.
**Stem users are more than welcome to listen for stem events, but these
functions are not being vended to our users. They may change in the future, use
them at your own risk.**
**Module Overview:**
::
get_logger - provides the stem's Logger instance
logging_level - converts a runlevel to its logging number
escape - escapes special characters in a message in preparation for logging
log - logs a message at the given runlevel
log_once - logs a message, deduplicating if it has already been logged
trace - logs a message at the TRACE runlevel
debug - logs a message at the DEBUG runlevel
info - logs a message at the INFO runlevel
notice - logs a message at the NOTICE runlevel
warn - logs a message at the WARN runlevel
error - logs a message at the ERROR runlevel
LogBuffer - Buffers logged events so they can be iterated over.
|- is_empty - checks if there's events in our buffer
+- __iter__ - iterates over and removes the buffered events
log_to_stdout - reports further logged events to stdout
.. data:: Runlevel (enum)
Enumeration for logging runlevels.
========== ===========
Runlevel Description
========== ===========
**ERROR** critical issue occurred, the user needs to be notified
**WARN** non-critical issue occurred that the user should be aware of
**NOTICE** information that is helpful to the user
**INFO** high level library activity
**DEBUG** low level library activity
**TRACE** request/reply logging
========== ===========
"""
import logging
import stem.prereq
import stem.util.enum
import stem.util.str_tools
# Logging runlevels. These are *very* commonly used so including shorter
# aliases (so they can be referenced as log.DEBUG, log.WARN, etc).
Runlevel = stem.util.enum.UppercaseEnum('TRACE', 'DEBUG', 'INFO', 'NOTICE', 'WARN', 'ERROR')
TRACE, DEBUG, INFO, NOTICE, WARN, ERR = list(Runlevel)
# mapping of runlevels to the logger module's values, TRACE and DEBUG aren't
# built into the module
LOG_VALUES = {
Runlevel.TRACE: logging.DEBUG - 5,
Runlevel.DEBUG: logging.DEBUG,
Runlevel.INFO: logging.INFO,
Runlevel.NOTICE: logging.INFO + 5,
Runlevel.WARN: logging.WARN,
Runlevel.ERROR: logging.ERROR,
}
logging.addLevelName(LOG_VALUES[TRACE], 'TRACE')
logging.addLevelName(LOG_VALUES[NOTICE], 'NOTICE')
LOGGER = logging.getLogger('stem')
LOGGER.setLevel(LOG_VALUES[TRACE])
FORMATTER = logging.Formatter(
fmt = '%(asctime)s [%(levelname)s] %(message)s',
datefmt = '%m/%d/%Y %H:%M:%S',
)
# There's some messages that we don't want to log more than once. This set has
# the messages IDs that we've logged which fall into this category.
DEDUPLICATION_MESSAGE_IDS = set()
# Adds a default nullhandler for the stem logger, suppressing the 'No handlers
# could be found for logger "stem"' warning as per...
# http://docs.python.org/release/3.1.3/library/logging.html#configuring-logging-for-a-library
class _NullHandler(logging.Handler):
def __init__(self):
logging.Handler.__init__(self, level = logging.FATAL + 5) # disable logging
def emit(self, record):
pass
if not LOGGER.handlers:
LOGGER.addHandler(_NullHandler())
[docs]def get_logger():
"""
Provides the stem logger.
:returns: **logging.Logger** for stem
"""
return LOGGER
[docs]def logging_level(runlevel):
"""
Translates a runlevel into the value expected by the logging module.
:param stem.util.log.Runlevel runlevel: runlevel to be returned, no logging if **None**
"""
if runlevel:
return LOG_VALUES[runlevel]
else:
return logging.FATAL + 5
[docs]def is_tracing():
"""
Checks if we're logging at the trace runlevel.
.. versionadded:: 1.6.0
:returns: **True** if we're logging at the trace runlevel and **False** otherwise
"""
for handler in get_logger().handlers:
if handler.level <= logging_level(TRACE):
return True
return False
[docs]def escape(message):
"""
Escapes specific sequences for logging (newlines, tabs, carriage returns). If
the input is **bytes** then this converts it to **unicode** under python 3.x.
:param str message: string to be escaped
:returns: str that is escaped
"""
if stem.prereq.is_python_3():
message = stem.util.str_tools._to_unicode(message)
for pattern, replacement in (('\n', '\\n'), ('\r', '\\r'), ('\t', '\\t')):
message = message.replace(pattern, replacement)
return message
[docs]def log(runlevel, message):
"""
Logs a message at the given runlevel.
:param stem.util.log.Runlevel runlevel: runlevel to log the message at, logging is skipped if **None**
:param str message: message to be logged
"""
if runlevel:
LOGGER.log(LOG_VALUES[runlevel], message)
[docs]def log_once(message_id, runlevel, message):
"""
Logs a message at the given runlevel. If a message with this ID has already
been logged then this is a no-op.
:param str message_id: unique message identifier to deduplicate on
:param stem.util.log.Runlevel runlevel: runlevel to log the message at, logging is skipped if **None**
:param str message: message to be logged
:returns: **True** if we log the message, **False** otherwise
"""
if not runlevel or message_id in DEDUPLICATION_MESSAGE_IDS:
return False
else:
DEDUPLICATION_MESSAGE_IDS.add(message_id)
log(runlevel, message)
# shorter aliases for logging at a runlevel
[docs]def trace(message):
log(Runlevel.TRACE, message)
[docs]def debug(message):
log(Runlevel.DEBUG, message)
[docs]def info(message):
log(Runlevel.INFO, message)
[docs]def notice(message):
log(Runlevel.NOTICE, message)
[docs]def warn(message):
log(Runlevel.WARN, message)
[docs]def error(message):
log(Runlevel.ERROR, message)
[docs]class LogBuffer(logging.Handler):
"""
Basic log handler that listens for stem events and stores them so they can be
read later. Log entries are cleared as they are read.
.. versionchanged:: 1.4.0
Added the yield_records argument.
.. deprecated:: 1.8.0
This will be dropped in Stem 2.x. Use python's logging.BufferingHandler instead.
"""
def __init__(self, runlevel, yield_records = False):
# TODO: At least in python 2.6 logging.Handler has a bug in that it doesn't
# extend object, causing our super() call to fail. When we drop python 2.6
# support we should switch back to using super() instead.
#
# super(LogBuffer, self).__init__(level = logging_level(runlevel))
logging.Handler.__init__(self, level = logging_level(runlevel))
self.formatter = FORMATTER
self._buffer = []
self._yield_records = yield_records
[docs] def is_empty(self):
return not bool(self._buffer)
def __iter__(self):
while self._buffer:
record = self._buffer.pop(0)
yield record if self._yield_records else self.formatter.format(record)
[docs] def emit(self, record):
self._buffer.append(record)
class _StdoutLogger(logging.Handler):
def __init__(self, runlevel):
logging.Handler.__init__(self, level = logging_level(runlevel))
self.formatter = logging.Formatter(
fmt = '%(asctime)s [%(levelname)s] %(message)s',
datefmt = '%m/%d/%Y %H:%M:%S')
def emit(self, record):
print(self.formatter.format(record))
[docs]def log_to_stdout(runlevel):
"""
Logs further events to stdout.
:param stem.util.log.Runlevel runlevel: minimum runlevel a message needs to be to be logged
"""
get_logger().addHandler(_StdoutLogger(runlevel))