详解Python logging调用Logger.info方法的处理过程

本次分析一下Logger.info的流程

1. Logger.info源码:

def info(self, msg, *args, **kwargs):
 """
 Log 'msg ％ args' with severity 'INFO'.

To pass exception information, use the keyword argument exc_info with
 a true value, e.g.

logger.info("Houston, we have a ％s", "interesting problem", exc_info=1)
 """
 if self.isEnabledFor(INFO):
  self._log(INFO, msg, args, **kwargs)

注释中反应了可以通过 msg和不定参数args来进行日志的格式化。
真实的调用为：_log方法：

2. Logger._log方法:

def _log(self, level, msg, args, exc_info=None, extra=None, stack_info=False):
 """
 Low-level logging routine which creates a LogRecord and then calls
 all the handlers of this logger to handle the record.
 """
 sinfo = None
 if _srcfile:
  #IronPython doesn't track Python frames, so findCaller raises an
  #exception on some versions of IronPython. We trap it here so that
  #IronPython can use logging.
  try:
   fn, lno, func, sinfo = self.findCaller(stack_info)
  except ValueError: # pragma: no cover
   fn, lno, func = "(unknown file)", 0, "(unknown function)"
 else: # pragma: no cover
  fn, lno, func = "(unknown file)", 0, "(unknown function)"
 if exc_info:
  if isinstance(exc_info, BaseException):
   exc_info = (type(exc_info), exc_info, exc_info.__traceback__)
  elif not isinstance(exc_info, tuple):
   exc_info = sys.exc_info()
 record = self.makeRecord(self.name, level, fn, lno, msg, args,
        exc_info, func, extra, sinfo)
 self.handle(record)

最后两行：

生成日志记录：

record = self.makeRecord(self.name, level, fn, lno, msg, args, exc_info, func, extra, sinfo)

处理日志记录

self.handle(record)

2 生成日志记录:

def makeRecord(self, name, level, fn, lno, msg, args, exc_info,
    func=None, extra=None, sinfo=None):
 """
 A factory method which can be overridden in subclasses to create
 specialized LogRecords.
 """
 rv = _logRecordFactory(name, level, fn, lno, msg, args, exc_info, func,
       sinfo)
 if extra is not None:
  for key in extra:
   if (key in ["message", "asctime"]) or (key in rv.__dict__):
    raise KeyError("Attempt to overwrite ％r in LogRecord" ％ key)
   rv.__dict__[key] = extra[key]
 return rv

调用_logRecordFactory初始化一个日志记录实例，_logRecordFactory 其实就是LogRecord类，初始化时，可能包含logger的name, level、调用的函数、行号、日志字符串、模板参数、堆栈信息等。

再看extra信息，extra到底有何用？现在从代码中可以看到，只是更新到生成的日志记录实例的__dict__中去.猜测：肯定会在生成最终的日志字符串的时候会用到。继续往下看。

3 处理日志记录self.handle(record):

Logger继承自Filterer,

def handle(self, record):
 """
 Call the handlers for the specified record.

This method is used for unpickled records received from a socket, as
 well as those created locally. Logger-level filtering is applied.
 """
 if (not self.disabled) and self.filter(record):
  self.callHandlers(record)

3.1 if语句中有一self.filter(record)的判断，看函数名，是来筛选是否要继续处理消息的，其核心源码如下：

def filter(self, record):
 """
 Determine if a record is loggable by consulting all the filters.

The default is to allow the record to be logged; any filter can veto
 this and the record is then dropped. Returns a zero value if a record
 is to be dropped, else non-zero.

.. versionchanged:: 3.2

Allow filters to be just callables.
 """
 rv = True
 for f in self.filters:
  if hasattr(f, 'filter'):
   result = f.filter(record)
  else:
   result = f(record) # assume callable - will raise if not
  if not result:
   rv = False
   break
 return rv

可以看到， 如果在handler中的filter中如果有返回为False或空，则会屏蔽对应的record,返回True或部位空的值，则会将record放行。那么我们就可以自定义自己的filter。

3.2 让Logger中所有的handles去处理record:

def callHandlers(self, record):
 """
 Pass a record to all relevant handlers.

Loop through all handlers for this logger and its parents in the
 logger hierarchy. If no handler was found, output a one-off error
 message to sys.stderr. Stop searching up the hierarchy whenever a
 logger with the "propagate" attribute set to zero is found - that
 will be the last logger whose handlers are called.
 """
 c = self
 found = 0
 while c:
  for hdlr in c.handlers:
   found = found + 1
   if record.levelno >= hdlr.level:
    hdlr.handle(record)
  if not c.propagate:
   c = None #break out
  else:
   c = c.parent
 if (found == 0):
  if lastResort:
   if record.levelno >= lastResort.level:
    lastResort.handle(record)
  elif raiseExceptions and not self.manager.emittedNoHandlerWarning:
   sys.stderr.write("No handlers could be found for logger"
        " \"％s\"\n" ％ self.name)
   self.manager.emittedNoHandlerWarning = True

代码中会去循环调用当前logger的所有handlers去处理record,for循环部分,之后，如果当前的logger的propagate的值为False或空，则不向logger的父logger传递，即向上传递。

4. Handler 中的 handler(record) 部分：

def handle(self, record):
 """
 Conditionally emit the specified logging record.

Emission depends on filters which may have been added to the handler.
 Wrap the actual emission of the record with acquisition/release of
 the I/O thread lock. Returns whether the filter passed the record for
 emission.
 """
 rv = self.filter(record)
 if rv:
  self.acquire()
  try:
   self.emit(record)
  finally:
   self.release()
 return rv

可以看到， Handler在处理record时， 会去加锁，然后调用self.emit(record)方法去处理。

4.1 emit(record)

def emit(self, record):
 """
 Do whatever it takes to actually log the specified logging record.

This version is intended to be implemented by subclasses and so
 raises a NotImplementedError.
 """
 raise NotImplementedError('emit must be implemented '
        'by Handler subclasses')

看到需要由子类去实现，以StreamHandler为例子：

def emit(self, record):
 """
 Emit a record.

If a formatter is specified, it is used to format the record.
 The record is then written to the stream with a trailing newline. If
 exception information is present, it is formatted using
 traceback.print_exception and appended to the stream. If the stream
 has an 'encoding' attribute, it is used to determine how to do the
 output to the stream.
 """
 try:
  msg = self.format(record)
  stream = self.stream
  stream.write(msg)
  stream.write(self.terminator)
  self.flush()
 except Exception:
  self.handleError(record)

4.2 Handler.format(record):

def format(self, record):
 """
 Format the specified record.

If a formatter is set, use it. Otherwise, use the default formatter
 for the module.
 """
 if self.formatter:
  fmt = self.formatter
 else:
  fmt = _defaultFormatter
 return fmt.format(record)

如果handler有自定义的formatter就用自定义的，如果没有则用默认的Formatter的实例， 初始化元源码为：

def __init__(self, fmt=None, datefmt=None, style='％'):
 """
 Initialize the formatter with specified format strings.

Initialize the formatter either with the specified format string, or a
 default as described above. Allow for specialized date formatting with
 the optional datefmt argument (if omitted, you get the ISO8601 format).

Use a style parameter of '％', '{' or '$' to specify that you want to
 use one of ％-formatting, :meth:`str.format` (``{}``) formatting or
 :class:`string.Template` formatting in your format string.

.. versionchanged:: 3.2
  Added the ``style`` parameter.
 """
 if style not in _STYLES:
  raise ValueError('Style must be one of: ％s' ％ ','.join(
       _STYLES.keys()))
 self._style = _STYLES[style][0](fmt)
 self._fmt = self._style._fmt
 self.datefmt = datefmt

有三个参数：

• fmt: 格式化模板

• datefmt: 时间格式化参数

• style: 日志格式化的样式。

style有三种：

_STYLES = {
'％': (PercentStyle, BASIC_FORMAT),
'{': (StrFormatStyle, '{levelname}:{name}:{message}'),
'$': (StringTemplateStyle, '${levelname}:${name}:${message}'),

可以看出对应到：％ 操作符的格式化， format方法的格式化以及Template的格式化。

Formatter的format方法源码为：

def format(self, record):
 """
 Format the specified record as text.
 The record's attribute dictionary is used as the operand to a
 string formatting operation which yields the returned string.
 Before formatting the dictionary, a couple of preparatory steps
 are carried out. The message attribute of the record is computed
 using LogRecord.getMessage(). If the formatting string uses the
 time (as determined by a call to usesTime(), formatTime() is
 called to format the event time. If there is exception information,
 it is formatted using formatException() and appended to the message.
 """
 record.message = record.getMessage()
 if self.usesTime():
  record.asctime = self.formatTime(record, self.datefmt)
 s = self.formatMessage(record)
 if record.exc_info:
  # Cache the traceback text to avoid converting it multiple times
  # (it's constant anyway)
  if not record.exc_text:
   record.exc_text = self.formatException(record.exc_info)
 if record.exc_text:
  if s[-1:] != "\n":
   s = s + "\n"
  s = s + record.exc_text
 if record.stack_info:
  if s[-1:] != "\n":
   s = s + "\n"
  s = s + self.formatStack(record.stack_info)

看到会调用record.getMessage(),这里仅仅是获取我们需要的日志信息。

之后会调用s = self.formatMessage(record):

def formatMessage(self, record):
 return self._style.format(record)

其实是调用了当前style的format方法，以％这一类型为例PercentStyle：

class PercentStyle(object):

default_format = '％(message)s'
asctime_format = '％(asctime)s'
asctime_search = '％(asctime)'

def __init__(self, fmt):
 self._fmt = fmt or self.default_format

def usesTime(self):
 return self._fmt.find(self.asctime_search) >= 0

def format(self, record):
 return self._fmt ％ record.__dict__

从其中的format方法可以看出，是针对record的__dict__属性中的所有参数进行格式化，这下，就清楚了之前的extra参数是干嘛用的了：可以在formatter中加入自己自定义的一些参数，如固定的用户信息等等。

之后，将最终的message flush到对应的Stream里面去就行了，就是整个流程:

来源：https://segmentfault.com/a/1190000018119406