Format log messages as a tree

Question

I would like to have python logging output to be in form of tree corresponding to logger tree. Just look at example.

Lets say we have a code:

import logging

logger_a = logging.getLogger("a")
logger_a_b = logging.getLogger("a.b")
logger_a_b_c = logging.getLogger("a.b.c")
# ...


logger_a.debug("One")

logger_a_b.warning("two")
logger_a_b.warning("three")

logger_a_b_c.critical("Four")

logger_a_b.warning("Five")

The output should be something like:

<--"a"
   |   DEBUG: One
   |
   o<--"a.b"
   |   | WARNING: Two
   |   | WARNING: Three
   |   |
   |   o<--"a.b.c"
   |       | CRITICAL: Four   
   |       |
   |   | WARNING: Five

I could write formatters for each of the log by hand, but it doesn't solve the problem of inserting something like o<--"a.b" right and I would prefer to calculate offset automatically by logging structure.

There is a module called logging tree. It prints the logging layout. What I would like, is to print log messages approximately the same way.

Do you know any libraries, ways of doing it easy way?

Answer 1

Based on your example, I created a custom Formatter which will handle the tree.

import logging

# custom tree formatter
class TreeFormatter(logging.Formatter):
    formatPrefix = {} # map loggername, formatPrefix

    def format(self, record):
        s = ""
        # first time this name is encountered: create the prefix and print the name
        if not record.name in self.formatPrefix: 
            f = self.getFormatPrefix(record)
            s += "%s \"%s\"\n" % (f, record.name)

        # print the actual message
        s += "%s %s: %s" % (self.formatPrefix[record.name], record.levelname, record.msg) 
        return s


    # create the format prefix for the given package name 
    # (stored in self.formatPrefix[record.name])
    # and return the first line to print
    def getFormatPrefix(self, record):
        depth = record.name.count(".")
        self.formatPrefix[record.name] = "   |" * (depth+1)

        if depth == 0:
            return "<--"

        return "%so<--" % ( ("   |" * depth)[:-1])

You can then use it to create the first-level logger (here a). The rest of the code is unchanged.

Here is an example:

# use this to create the first level logger
def createTreeLogger(name, level=logging.DEBUG):
    logger = logging.getLogger(name)
    logger.setLevel(level)
    ch = logging.StreamHandler()
    ch.setLevel(level)
    ch.setFormatter(TreeFormatter())
    logger.addHandler(ch)
    return logger


if __name__ == '__main__':

    logger_a = createTreeLogger("a") # first level: use createLogger
    # then create your loggers as always
    logger_a_b = logging.getLogger("a.b") 
    logger_a_b_c = logging.getLogger("a.b.c")


    logger_a.debug("One")

    logger_a_b.warning("two")
    logger_a_b.warning("three")

    logger_a_b_c.critical("Four")

    logger_a_b.warning("Five")
    logger_a.warning("Six")

What is nice is that the logging package internals will automatically use the same handler for the subpackages (a.b, a.b.c). So, by running this code, you get:

<-- "a"
   | DEBUG: One
   o<-- "a.b"
   |   | WARNING: two
   |   | WARNING: three
   |   o<-- "a.b.c"
   |   |   | CRITICAL: Four
   |   | WARNING: Five
   | WARNING: Six

One drawback is that the logs become confusing if you have more than one package hierarchy. But the TreeFormatter class is easy to tweak to your needs.