logger_tt 1.5.2

Make logging simple, log even exception that you forgot to catch. Make multiprocessing logging becomes a breeze

Logger_tt

Make configuring logging simpler and log even exceptions that you forgot to catch.
Even multiprocessing logging becomes a breeze.

Install

• From PYPI: pip install logger_tt
• From Github: clone or download this repo then python setup.py install

Overview:

In the most simple case, add the following code into your main python script of your project:

from logger_tt import setup_logging    

setup_logging(full_context=1)

Then from any of your modules, you just need to get a logger and start logging. There are two ways that you could obtain a logger.

• Conventional way: as you have always done it that way

  from logging import getLogger
  
  logger = getLogger(__name__)
  

• Convenient way: use a pre-made logger named logger_tt from this package.

  from logger_tt import logger
  

After that, you start logging as usual:

logger.debug('Module is initialized')
logger.info('Making connection ...')

# output
[2020-07-21 11:24:19] [__main__:5 DEBUG] Module is initialized
[2020-07-21 11:24:19] [__main__:6 INFO] Making connection ...

Both ways give you the same output except the line number, obviously.
The pre-made logger also has an advantage that it will automatically inject threadName and processName to the output in case of multithreading or multiprocessing.

Both ways will provide your project with the following default log behaviors:

• log file: Assume that your working directory is project_root, log.txt is stored at your project_root/logs/ folder.
  If the log path doesn't exist, it will be created.
  The log file is time rotated at midnight. A maximum of 15 dates of logs will be kept. This log file's level is DEBUG.
  The log format is [%(asctime)s] [%(name)s:%(lineno)d %(levelname)s] %(message)s,
  where the time format is %Y-%m-%d %H:%M:%S.
  Example: [2020-05-09 00:31:33] [myproject.mymodule:26 DEBUG] Module is initialized

• console: log records with level INFO and above will be printed to stdout of the console.
  The format for console log is simpler: [%(asctime)s] %(levelname)s: %(message)s.
  Example: [2020-05-09 00:31:34] INFO: Making connection ...

• urllib3 logger: this ready-made logger is to silent unwanted messages from requests library.

• suppressed logger: exchangelib. This sets logging level of exchangelib logger to WARNING.
  This is another way to silent unwanted messages from other module, read below for details.

• root logger: if there is no logger initialized in your module, this logger will be used with the above behaviors. This logger is also used to log uncaught exception in your project. Example:

  raise RecursionError
  

  # log.txt
  [2020-05-31 19:16:01] [root ERROR] Uncaught exception
  Traceback (most recent call last):
    File "D:/MyProject/Echelon/eyes.py", line 13, in <module>
      raise RecursionError
      => var_in = Customer(name='John', member_id=123456)
      => arg = (456, 789)
      => kwargs = {'my_kw': 'hello', 'another_kw': 'world'}
  RecursionError
  

• context logging: When an exception occur, variables used in the line of error are also logged.
  To log full local variables of current function scope, pass full_context=1 to setup_logging.
  If you need the outer scope too, set full_context to 2, 3 and so on...

Usage:

All configs are done through setup_logging function:

setup_logging(config_path="", log_path="", 
              capture_print=False, strict=False, guess_level=False,
              full_context=False,
              suppress_level_below=logging.WARNING,
              use_multiprocessing=False)

This function also return a LogConfig object. Except config_path, log_path and use_multiprocessing, other parameters are attributes of this object and can be changed on the fly.

Except config_path, log_path, all other parameters can be defined in logger_tt section in the config file (see Sample config chapter below). Parameter with the same name passed in setup_logging function will override the one in the config file.

1. You can overwrite the default log path with your own as follows:

   setup_logging(log_path='new/path/to/your_log.txt')
   

2. You can config your own logger and handler by providing either yaml or json config file as follows:

   setup_logging(config_path='path/to/.yaml_or_.json')
   

   Without providing a config file, the default config file with the above default log behavior is used. You could copy log_conf.yaml or log_conf.json shipped with this package to start making your version.

   Warning: To process .yaml config file, you need pyyaml package: pip install pyyaml

3. Capture stdout:

   If you have an old code base with a lot of print(msg) or sys.stdout.write(msg) and don't have access or time to refactor them into something like logger.info(msg), you can capture these msg and log them to file, too.

   To capture only msg that is printed out by print(msg), simply do as follows:

   setup_logging(capture_print=True)
   

   Example:

   print('To be or not to be')
   sys.stdout.write('That is the question')
   

   # log.txt
   [2020-05-09 11:42:08] [PrintCapture INFO] To be or not to be
   

   ---

   Yes, That is the question is not captured. Some libraries may directly use sys.stdout.write to draw on the screen (eg. progress bar) or do something quirk. This kind of information is usually not useful for users. But when you do need it, you can capture it as follows:

   setup_logging(capture_print=True, strict=True)
   

   Example:

   sys.stdout.write('The plane VJ-723 has been delayed')
   sys.stdout.write('New departure time has not been scheduled')
   

   # log.txt
   [2020-05-09 11:42:08] [PrintCapture INFO] The plane VJ-723 has been delayed
   [2020-05-09 11:42:08] [PrintCapture INFO] New departure time has not been scheduled
   

   ---

   As you have seen, the log level of the captured message is INFO . What if the code base prints something like An error has occurred. Abort operation. and you want to log it as Error? Just add guess_level=True to setup_logging().

   setup_logging(capture_print=True, guess_level=True)
   

   Example:

   print('An error has occurred. Abort operation.')
   print('A critical error has occurred during making request to database')
   

   # log.txt
   [2020-05-09 11:42:08] [PrintCapture ERROR] An error has occurred. Abort operation.
   [2020-05-09 11:42:08] [PrintCapture CRITICAL] A critical error has occurred during making request to database
   

   Note: Capturing stdout ignores messages of blank line. That means messages like \n\n or (spaces) will not appear in the log. But messages that contain blank line(s) and other characters will be fully logged. For example, \nTo day is a beautiful day\n will be logged as is.

4. Exception logging:

   Consider the following error code snippet:

   API_KEY = "asdjhfbhbsdf82340hsdf09u3ionf98230234ilsfd"
   TIMEOUT = 60
   
   class MyProfile:
       def __init__(self, name):
           self.my_boss = None
           self.name = name
   
   def my_faulty_func(my_var, *args, **kwargs):
       new_var = 'local scope variable'
       me = MyProfile('John Wick')
       boss = MyProfile('Winston')
       me.my_boss = boss
       print(f'Information: {var} and {me.my_boss.name}' 
              ' at {me.my_boss.location} with {API_KEY}')
   
   if __name__ == '__main__':
       cpu_no = 4
       max_concurrent_processes = 3
       my_faulty_func(max_concurrent_processes, 'ryzen 7', freq=3.4)
   

   In our hypothetical code above,print function will raise an exception. This exception, by default, will not only be logged but also analyzed with objects that appeared in the line:

   [2020-06-06 09:36:01] ERROR: Uncaught exception:
   Traceback (most recent call last):
     File "D:/MyProject/AutoBanking/main.py", line 31, in <module>
       my_faulty_func(max_concurrent_processes, 'ryzen 7', freq=3.4)
        |-> my_faulty_func = <function my_faulty_func at 0x0000023770C6A288>
        |-> max_concurrent_processes = 3
   
     File "D:/MyProject/AutoBanking/main.py", line 25, in my_faulty_func
       print(f'Information: {var} and {me.my_boss.name}'
              ' at {me.my_boss.location} with {API_KEY}')
        |-> me.my_boss.name = 'Winston'
        |-> me.my_boss.location = '!!! Not Exists'
        |-> (outer) API_KEY = 'asdjhfbhbsdf82340hsdf09u3ionf98230234ilsfd'
   NameError: name 'var' is not defined
   

   Note: look at the print(f'Information... line, logger-tt print this error line different from normal python traceback! With normal traceback, multi-line python statement has its only first line printed out. With logger-tt, full statement is grabbed for you.

   For each level in the stack, any object that appears in the error line is shown with its readable representation. This representation may not necessarily be __repr__. The choice between __str__ and __repr__ are as follows:

   • __str__ : __str__ is present and the object class's __repr__ is default with <class name at Address>.
   • __repr__: __str__ is present but the object class's __repr__ is anything else, such as ClassName(var=value).
     Also, when __str__ is missing, even if __repr__ is <class name at Address>, it is used.

   Currently, if an object doesn't exist and is directly accessed, as var in this case, it will not be shown up. But if it is attribute accessed with dot ., as location in me.my_boss.location, then its value is an explicit string '!!! Not Exists'.

   As you may have noticed, a variable API_KEY has its name prefixed with outer.
   This tells you that the variable is defined in the outer scope, not local.

   More often than not, only objects in the error line are not sufficient to diagnose what has happened. You want to know what the inputs of the function were. You want to know what the intermediate calculated results were. You want to know other objects that appeared during runtime, not only local but also outer scope. In other words, you want to know the full context of what has happened. logger-tt is here with you:

   setup_logging(full_context=2)
   

   With the above hypothetical code snippet, the error log becomes the following:

   [2020-06-06 10:35:21] ERROR: Uncaught exception:
   Traceback (most recent call last):
     File "D:/MyProject/AutoBanking/main.py", line 31, in <module>
       my_faulty_func(max_concurrent_processes, 'ryzen 7', freq=3.4)
        |-> my_faulty_func = <function my_faulty_func at 0x0000019E3599A288>
        |-> max_concurrent_processes = 3
        => __name__ = '__main__'
        => __doc__ = None
        => __package__ = None
        => __loader__ = <_frozen_importlib_external.SourceFileLoader object at 0x0000019E35840E48>
        => __spec__ = None
        => __annotations__ = {}
        => __builtins__ = <module 'builtins' (built-in)>
        => __file__ = 'D:/MyProject/AutoBanking/main.py'
        => __cached__ = None
        => setup_logging = <function setup_logging at 0x0000019E35D111F8>
        => getLogger = <function getLogger at 0x0000019E35BC7C18>
        => logger = <Logger __main__ (DEBUG)>
        => API_KEY = 'asdjhfbhbsdf82340hsdf09u3ionf98230234ilsfd'
        => TIMEOUT = 60
        => MyProfile = <class '__main__.MyProfile'>
        => cpu_no = 4
   
     File "D:/MyProject/AutoBanking/main.py", line 25, in my_faulty_func
       print(f'Information: {var} and {me.my_boss.name} at {me.my_boss.location} with {API_KEY}')
        |-> me.my_boss.name = 'Winston'
        |-> me.my_boss.location = '!!! Not Exists'
        |-> (outer) API_KEY = 'asdjhfbhbsdf82340hsdf09u3ionf98230234ilsfd'
        => my_var = 3
        => args = ('ryzen 7',)
        => kwargs = {'freq': 3.4}
        => new_var = 'local scope variable'
        => me = <__main__.MyProfile object at 0x0000019E35D3BA48>
        => boss = <__main__.MyProfile object at 0x0000019E35D3B9C8>
   NameError: name 'var' is not defined
   

   Additional objects that not appear in the error line are prefixed with =>.

5. try-except exception logging:

   exception context logging also applies for try-except block. This means that if you call logger.exception() inside except block, you would have all variables' value at the line of exception. For example,

   def my_faulty_func():
       a = 10
       b = 0
       c = a/b
       return c
   
   def my_main():
       try:
           my_faulty_func()
       except Exception as e:
           logger.exception('some error has occured')
           print('Clean up resource')
   
   my_main()
   

   Then the log will show up as follows:

   [2020-06-12 21:37:00] ERROR: some error has occured
   Traceback (most recent call last):
     File "D:/MyProject/exception_log.py", line 19, in my_main
       my_faulty_func()
        |-> my_faulty_func = <function my_faulty_func at 0x000001875DD4B168>
   
     File "D:/MyProject/exception_log.py", line 13, in my_faulty_func
       c = a / b
        |-> a = 10
        |-> b = 0
   ZeroDivisionError: division by zero
   Clean up resource
   

   Note: As in python's logging document, logger.exception() should only be called from an exception handler, eg. inside except block.

   You don't need to pass exception object to logger.exception(). It already knows how to get a traceback internally. This enable you to pass any string in as a hint or a short description of what may have happened.
   Otherwise, passing exception object, as logger.exception(e), will cause the first line of error report to be the message of exception. In the case of the above example, it would be come [2020-06-12 21:37:00] ERROR: division by zero.

6. Silent unwanted logger:

   Third party modules also have logger and their messages are usually not related to your code. A bunch of unwanted messages may hide the one that come from your own module. To prevent that and also reduce log file size, we need to silent unwanted loggers.

   By config file, there are two ways to silent a logger:

   • Create a new logger: in logger section of config file, add a new logger whose name is the same with the one you want to silent. Set it level to WARNING or above. If you add handlers, you should also set propagate to no or False. Otherwise, the same message may be logged multiple times. Ex:

     urllib3:
       level: WARNING
       handlers: [console, error_file_handler]
       propagate: no
     

     Above setting only allow messages with level WARNING and above to be processed. Usually that is enough to silent most of unwanted messages. If you need to silent more messages, try ERROR or CRITICAL.

   • Add logger's name to suppress list: Then a new logger with level default to WARNING will be automatically created for you. Ex:

     suppress: [exchangelib, urllib3]
     

     If you need to suppress at even higher level, use suppress_level_below in setup_logging. For example suppress any message below ERROR level that comes from loggers in suppress list:

     setup_logging(suppress_level_below=logging.ERROR)
     

   You could also suppress loggers directly by setup_logging:

   from logger_tt import setup_logging    
   
   setup_logging(suppress=['urllib3', 'exchangelib'])
   

7. Logging in multiprocessing:

   This is archived by using multiprocessing queues or a socket server.

   For linux, copy-on-write while forking carries over logger's information. So multiprocess.Queue is enough in this case.

   For Windows, it is important that setup_logging() must be call out side of if __name__ == '__main__': guard block. Because child processes run from scratch and re-import __main__, by re-executing setup_logging(), logger SocketHandler can be setup automatically.

   This also means that the same config can work with both multiprocessing.Process and multiprocessing.Pool magically without user doing anything special.

   Below is a minimal example:

   import time
   from random import randint
   from multiprocessing import Process
   
   from logger_tt import setup_logging, logger
   
   
   setup_logging(use_multiprocessing=True)        # for Windows, this line must be outside of guard block
   
   
   def worker(arg):
       logger.info(f'child process {arg}: started')
       time.sleep(randint(1,10))                  # imitate time consuming process
       logger.info(f'child process {arg}: stopped')
   
   
   if __name__ == '__main__':
       all_processes = []
       logger.info('Parent process is ready to spawn child')
       for i in range(3):
           p = Process(target=worker, args=(i,))
           all_processes.append(p)
           p.daemon = True
           p.start()
   
       for p in all_processes:
           p.join()
   

   The content of log.txt should be similar to below:

   [2020-10-28 20:39:14] [root:129 DEBUG] _________________New log started__________________
   [2020-10-28 20:39:17] [root:130 DEBUG] Log config file: D:\my_project\log_config.json
   [2020-10-28 20:39:17] [root:131 DEBUG] Logging server started!
   [2020-10-28 20:39:22] [__main__:28 INFO] Parent process is ready to spawn child
   [2020-10-28 20:39:22] [__mp_main__:16 INFO] Process-3 child process 2: started
   [2020-10-28 20:39:22] [__mp_main__:16 INFO] Process-2 child process 1: started
   [2020-10-28 20:39:22] [__mp_main__:16 INFO] Process-1 child process 0: started
   [2020-10-28 20:39:23] [__mp_main__:18 INFO] Process-2 child process 1: stopped
   [2020-10-28 20:39:23] [__mp_main__:18 INFO] Process-3 child process 2: stopped
   [2020-10-28 20:39:24] [__mp_main__:18 INFO] Process-1 child process 0: stopped
   

   Note: Under linux, to use queueHandler, you must pass use_multiprocessing="fork" to setup_logging.
   Other options True, spawn, forkserver will use socketHandler by default.
   This is to prevent you set_start_method as spawn under linux and thus queueHandler won't work.

8. Temporary disable logging:

   Some block of code contain critical information, such as password processing, that should not be logged. You can disable logging for that block with a logging_disabled context:

   from logger_tt import logging_disabled, getLogger
   
   logger = getLogger(__name__) 
   
   
   logger.debug('Begin a secret process')
   with logging_disabled():
       logger.info('This will not appear in any log')
   
   logger.debug('Finish')
   

Sample config:

1. Yaml format:

   log_config.yaml:

   # This is an example of config file
   # In case of no config provided, log_config.json file will be loaded
   # If you need a yaml file, install pyyaml package and copy this file
   version: 1
   disable_existing_loggers: False
   formatters:
     simple:
       format: "[%(asctime)s] [%(name)s:%(lineno)d %(levelname)s] %(message)s"
       datefmt: "%Y-%m-%d %H:%M:%S"
     brief:
       format: "[%(asctime)s] %(levelname)s: %(message)s"
       datefmt: "%Y-%m-%d %H:%M:%S"
   
   handlers:
     console:
       class: logging.StreamHandler
       level: INFO
       formatter: brief
       stream: ext://sys.stdout
   
     error_file_handler:
       class: logging.handlers.TimedRotatingFileHandler
       level: DEBUG
       formatter: simple
       filename: logs/log.txt
       backupCount: 15
       encoding: utf8
       when: midnight
   
   loggers:
     urllib3:
       level: WARNING
       handlers: [console, error_file_handler]
       propagate: no
   
   root:
     level: DEBUG
     handlers: [console, error_file_handler]
   
   logger_tt:
     suppress: ["exchangelib"]
     suppress_level_below: "WARNING"
     capture_print: False
     strict: False
     guess_level: False
     full_context: 0
     use_multiprocessing: False
     default_logger_formats:
       normal: ["%(name)s", "%(filename)s"]
       thread: ["%(message)s", "%(threadName)s %(message)s"]
       multiprocess: ["%(message)s", "%(processName)s %(message)s"]
       both: ["%(message)s", "%(processName)s %(threadName)s %(message)s"]
   

2. Json format:

   log_config.json:

   {
     "version": 1,
     "disable_existing_loggers": false,
     "formatters": {
       "simple": {
         "format": "[%(asctime)s] [%(name)s:%(lineno)d %(levelname)s] %(message)s",
         "datefmt": "%Y-%m-%d %H:%M:%S"
       },
       "brief": {
         "format": "[%(asctime)s] %(levelname)s: %(message)s",
         "datefmt": "%Y-%m-%d %H:%M:%S"
       }
     },
   
     "handlers": {
       "console": {
         "class": "logging.StreamHandler",
         "level": "INFO",
         "formatter": "brief",
         "stream": "ext://sys.stdout"
       },
   
       "error_file_handler": {
         "class": "logging.handlers.TimedRotatingFileHandler",
         "level": "DEBUG",
         "formatter": "simple",
         "filename": "logs/log.txt",
         "backupCount": 15,
         "encoding": "utf8",
         "when": "midnight"
       }
     },
   
     "loggers": {
       "urllib3": {
         "level": "ERROR",
         "handlers": ["console", "error_file_handler"],
         "propagate": false
       }
     },
   
     "root": {
       "level": "DEBUG",
       "handlers": ["console", "error_file_handler"]
     },
   
     "logger_tt": {
       "suppress": ["exchangelib"],
       "suppress_level_below": "WARNING",
       "capture_print": false,
       "strict": false,
       "guess_level": false,
       "full_context": 0,
       "use_multiprocessing": false,
       "default_logger_formats": {
          "normal": ["%(name)s", "%(filename)s"],
          "thread": ["%(message)s", "%(threadName)s %(message)s"],
          "multiprocess": ["%(message)s", "%(processName)s %(message)s"],
          "both": ["%(message)s", "%(processName)s %(threadName)s %(message)s"]
       }
     }
   }
   

changelog

1.5.2

Improved the pre-made logger named logger_tt

• logger_tt now can detect the qualified __name__ of the module that calls it. Instead of filename, output log line will have the __name__ as regular logger.

  For example:

  [2020-07-21 11:24:19] [my_module.py:5 DEBUG] Module is initialized
  [2020-07-21 11:24:19] [sub_module.py:15 DEBUG] Entering sub module
  

  Now becomes

  [2020-07-21 11:24:19] [__main__:5 DEBUG] Module is initialized
  [2020-07-21 11:24:19] [my_module.submodule:15 DEBUG] Entering sub module
  

• Suppressing loggers also works with log records output by logger_tt by using the qualified __name__ too. For example, suppressing my_module.submodule will tell logger_tt not to output the second line.

  This is much better than suppressing logger_tt if you use this same logger in other modules too.

• You now can define fields of log record for logger_tt in the log config file too. Just looks for default_logger_formats section. It works by replacing the field in the formatters that are used by any handler of the root logger.

Pre-existing loggers:
Before this version, if you import submodules before importing logger_tt and there are loggers in submodules, these loggers do not inspect exception when you call logger.exception(). That is because there class was different than loggers created after importing logger_tt. Now all loggers have the same new class regardless the point of importing logger_tt.

setup_logging(): This function should only be called once. Add a warning if it is called more than one time.

1.5.1

• Use socketHandler as default for multiprocessing. Under linux, to use queueHandler, user must pass use_multiprocessing="fork" to setup_logging
• Expose logging_disabled function to user: from logger_tt import logging_disabled. Then this function can be used as a context with with statement.
• For convenient, user can import a pre-made logger from logger_tt to use right away in sub modules. The built-in getLogger function can be imported from logger_tt, too.
• Added line number to a default simple log record formatter in the config file.
• Most parameters of setup_logging() function can be specified in the config file, too. If the same parameter is specified in both setup_logging() function and in the config file, the parameter passed in setup_logging() will be used.

1.5.0

• Logging is off-loaded to another thread and uses Queue to communicate. This allow critical thread can do there job why time-consuming logging can be done later or in parallel.
• Support for multiprocessing logging. For linux, a multiprocessing queue is used. For Windows and macOS, a socket server is used instead.
• setup_logging now return a LogConfig object. You can set/change parameters of this object instead of passing arguments directly to setup_logging.
  Only config_path, log_path and use_multiprocessing argument must be set with setup_logging.

Behaviors changed:

• full_context is now an int that indicate the depth level from the bottom, where surrounding variables should be parsed.
• Turned off parsing full context for raise exception since many exception names are enough to understand the problem.
• log_config file: move suppress section into logger_tt section. Future settings will also be put into this section for the sake of managing. If you need to hanging the logging framework, you just need to delete this section and move on.

1.4.2

To prevent exception during logging, the following actions have been applied:

• Catch exception while parsing for object's value (property of a class)
• Catch exception while evaluating __repr__ and __str__ of object
• Disable logging while inspecting objects' value and representation
• Disable logging after an uncaught exception is logged. Because the interpreter is shutting down, objects get deleted. Logging put inside magic function __del__ will raise error.

1.4.1

• Fix print_capture ignoring print() line in global scope due to lacking code_context frame
• If __str__ of an object has multiple lines, also indent the second line and so on accordingly.
• If there is an exception during getting object's representation, return !!! Attribute error instead of Error in sys.excepthook

1.4.0

• Add an extra field suppress in config file. Any logger's name appeared in this list will have its messages suppressed.

1.3.2

• change extended ascii dash ─ to normal dash - so that it is displayed consistently in different encoding

1.3.1

• change extended ascii vertical bar ├ to normal bar | so that it is displayed consistently in different encoding

1.3.0

• Exception analyzing now fetch full multi-line python statement. This means that variables lie at seconds and below of the same statement can also be seen without the need of full_context=True.

1.2.1

• Extend logging context to logger.exception() as well. Now you can do try-except a block of code and still have a full context at error line.

1.2.0

• Add logging context for uncaught exception. Now automatically log variables surrounding the error line, too.
• Add test cases for logging exception

1.1.1

• Fixed typos and grammar
• Add config file sample to README
• using full name log_config.json instead of log_conf.json, the same for yaml file
• add test cases for capture print

1.1.0

• Add capture print functionality with guess level for the message.