Since this thing is on the innerwebs I suppose it should have a
[python_utils.git] / logging_utils.py
index a15ccd64b24d9d3393fea65ab62ee2fcf27c348a..6ceba65f7b8ee2413ae8e1595bf653d169856e95 100644 (file)
@@ -1,4 +1,7 @@
 #!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+
+# © Copyright 2021-2022, Scott Gasch
 
 """Utilities related to logging."""
 
@@ -8,14 +11,15 @@ import datetime
 import enum
 import io
 import logging
-from logging.handlers import RotatingFileHandler, SysLogHandler
 import os
 import random
 import sys
-from typing import Callable, Iterable, Mapping, Optional
+from logging.config import fileConfig
+from logging.handlers import RotatingFileHandler, SysLogHandler
+from typing import Any, Callable, Dict, Iterable, List, Optional
 
-from overrides import overrides
 import pytz
+from overrides import overrides
 
 # This module is commonly used by others in here and should avoid
 # taking any unnecessary dependencies back on them.
@@ -173,12 +177,12 @@ cfg.add_argument(
     ),
 )
 
-built_in_print = print
-logging_initialized = False
+BUILT_IN_PRINT = print
+LOGGING_INITIALIZED = False
 
 
 # A map from logging_callsite_id -> count of logged messages.
-squelched_logging_counts: Mapping[str, int] = {}
+squelched_logging_counts: Dict[str, int] = {}
 
 
 def squelch_repeated_log_messages(squelch_after_n_repeats: int) -> Callable:
@@ -222,8 +226,8 @@ class SquelchRepeatedMessagesFilter(logging.Filter):
     """
 
     def __init__(self) -> None:
-        self.counters = collections.Counter()
         super().__init__()
+        self.counters: collections.Counter = collections.Counter()
 
     @overrides
     def filter(self, record: logging.LogRecord) -> bool:
@@ -256,9 +260,7 @@ class DynamicPerScopeLoggingLevelFilter(logging.Filter):
         per_scope_logging_levels: str,
     ) -> None:
         super().__init__()
-        self.valid_levels = set(
-            ['NOTSET', 'DEBUG', 'INFO', 'WARNING', 'ERROR', 'CRITICAL']
-        )
+        self.valid_levels = set(['NOTSET', 'DEBUG', 'INFO', 'WARNING', 'ERROR', 'CRITICAL'])
         self.default_logging_level = default_logging_level
         self.level_by_scope = {}
         if per_scope_logging_levels is not None:
@@ -285,9 +287,9 @@ class DynamicPerScopeLoggingLevelFilter(logging.Filter):
                         file=sys.stderr,
                     )
                     continue
-                self.level_by_scope[
-                    scope
-                ] = DynamicPerScopeLoggingLevelFilter.level_name_to_level(level)
+                self.level_by_scope[scope] = DynamicPerScopeLoggingLevelFilter.level_name_to_level(
+                    level
+                )
 
     @overrides
     def filter(self, record: logging.LogRecord) -> bool:
@@ -313,7 +315,7 @@ class DynamicPerScopeLoggingLevelFilter(logging.Filter):
 
 
 # A map from function_identifier -> probability of logging (0.0%..100.0%)
-probabilistic_logging_levels: Mapping[str, float] = {}
+probabilistic_logging_levels: Dict[str, float] = {}
 
 
 def logging_is_probabilistic(probability_of_logging: float) -> Callable:
@@ -381,30 +383,101 @@ class MillisecondAwareFormatter(logging.Formatter):
 
     """
 
-    converter = datetime.datetime.fromtimestamp
+    converter = datetime.datetime.fromtimestamp  # type: ignore
 
     @overrides
     def formatTime(self, record, datefmt=None):
-        ct = MillisecondAwareFormatter.converter(
-            record.created, pytz.timezone("US/Pacific")
-        )
+        ct = MillisecondAwareFormatter.converter(record.created, pytz.timezone("US/Pacific"))
         if datefmt:
             s = ct.strftime(datefmt)
         else:
             t = ct.strftime("%Y-%m-%d %H:%M:%S")
-            s = "%s,%03d" % (t, record.msecs)
+            s = f"{t},{record.msecs:%03d}"
         return s
 
 
+def log_about_logging(
+    logger,
+    default_logging_level,
+    preexisting_handlers_count,
+    fmt,
+    facility_name,
+):
+    level_name = logging._levelToName.get(default_logging_level, str(default_logging_level))
+    logger.debug('Initialized global logging; default logging level is %s.', level_name)
+    if config.config['logging_clear_preexisting_handlers'] and preexisting_handlers_count > 0:
+        logger.warning(
+            'Logging cleared %d global handlers (--logging_clear_preexisting_handlers)',
+            preexisting_handlers_count,
+        )
+    logger.debug('Logging format specification is "%s"', fmt)
+    if config.config['logging_debug_threads']:
+        logger.debug('...Logging format spec captures tid/pid. (--logging_debug_threads)')
+    if config.config['logging_debug_modules']:
+        logger.debug(
+            '...Logging format spec captures files/functions/lineno. (--logging_debug_modules)'
+        )
+    if config.config['logging_syslog']:
+        logger.debug(
+            'Logging to syslog as %s with priority mapping based on level. (--logging_syslog)',
+            facility_name,
+        )
+    if config.config['logging_filename']:
+        logger.debug(
+            'Logging to file "%s". (--logging_filename)', config.config["logging_filename"]
+        )
+        logger.debug(
+            '...with %d bytes max file size. (--logging_filename_maxsize)',
+            config.config["logging_filename_maxsize"],
+        )
+        logger.debug(
+            '...and %d rotating backup file count. (--logging_filename_count)',
+            config.config["logging_filename_count"],
+        )
+    if config.config['logging_console']:
+        logger.debug('Logging to the console (stderr). (--logging_console)')
+    if config.config['logging_info_is_print']:
+        logger.debug(
+            'Logging logger.info messages will be repeated on stdout. (--logging_info_is_print)'
+        )
+    if config.config['logging_squelch_repeats']:
+        logger.debug(
+            'Logging code allowed to request repeated messages be squelched. (--logging_squelch_repeats)'
+        )
+    else:
+        logger.debug(
+            'Logging code forbidden to request messages be squelched; all messages logged. (--no_logging_squelch_repeats)'
+        )
+    if config.config['logging_probabilistically']:
+        logger.debug(
+            'Logging code is allowed to request probabilistic logging. (--logging_probabilistically)'
+        )
+    else:
+        logger.debug(
+            'Logging code is forbidden to request probabilistic logging; messages always logged. (--no_logging_probabilistically)'
+        )
+    if config.config['lmodule']:
+        logger.debug(
+            f'Logging dynamic per-module logging enabled. (--lmodule={config.config["lmodule"]})'
+        )
+    if config.config['logging_captures_prints']:
+        logger.debug(
+            'Logging will capture printed data as logger.info messages. (--logging_captures_prints)'
+        )
+
+
 def initialize_logging(logger=None) -> logging.Logger:
-    global logging_initialized
-    if logging_initialized:
-        return
-    logging_initialized = True
+    global LOGGING_INITIALIZED
+    if LOGGING_INITIALIZED:
+        return logging.getLogger()
+    LOGGING_INITIALIZED = True
 
     if logger is None:
         logger = logging.getLogger()
 
+    # --logging_clear_preexisting_handlers removes logging handlers
+    # that were registered by global statements during imported module
+    # setup.
     preexisting_handlers_count = 0
     assert config.has_been_parsed()
     if config.config['logging_clear_preexisting_handlers']:
@@ -412,19 +485,22 @@ def initialize_logging(logger=None) -> logging.Logger:
             logger.removeHandler(logger.handlers[0])
             preexisting_handlers_count += 1
 
+    # --logging_config_file pulls logging settings from a config file
+    # skipping the rest of this setup.
     if config.config['logging_config_file'] is not None:
-        logging.config.fileConfig('logging.conf')
+        fileConfig(config.config['logging_config_file'])
         return logger
 
-    handlers = []
+    handlers: List[logging.Handler] = []
+    handler: Optional[logging.Handler] = None
 
-    # Global default logging level (--logging_level)
-    default_logging_level = getattr(
-        logging, config.config['logging_level'].upper(), None
-    )
+    # Global default logging level (--logging_level); messages below
+    # this level will be silenced.
+    default_logging_level = getattr(logging, config.config['logging_level'].upper(), None)
     if not isinstance(default_logging_level, int):
-        raise ValueError('Invalid level: %s' % config.config['logging_level'])
+        raise ValueError(f'Invalid level: {config.config["logging_level"]}')
 
+    # Custom or default --logging_format?
     if config.config['logging_format']:
         fmt = config.config['logging_format']
     else:
@@ -432,16 +508,25 @@ def initialize_logging(logger=None) -> logging.Logger:
             fmt = '%(levelname).1s:%(filename)s[%(process)d]: %(message)s'
         else:
             fmt = '%(levelname).1s:%(asctime)s: %(message)s'
+
+    # --logging_debug_threads and --logging_debug_modules both affect
+    # the format by prepending information about the pid/tid or
+    # file/function.
     if config.config['logging_debug_threads']:
         fmt = f'%(process)d.%(thread)d|{fmt}'
     if config.config['logging_debug_modules']:
         fmt = f'%(filename)s:%(funcName)s:%(lineno)s|{fmt}'
 
+    # --logging_syslog (optionally with --logging_syslog_facility)
+    # sets up for logging to use the standard system syslogd as a
+    # sink.
+    facility_name = None
     if config.config['logging_syslog']:
         if sys.platform not in ('win32', 'cygwin'):
             if config.config['logging_syslog_facility']:
                 facility_name = 'LOG_' + config.config['logging_syslog_facility']
-            facility = SysLogHandler.__dict__.get(facility_name, SysLogHandler.LOG_USER)
+            facility = SysLogHandler.__dict__.get(facility_name, SysLogHandler.LOG_USER)  # type: ignore
+            assert facility is not None
             handler = SysLogHandler(facility=facility, address='/dev/log')
             handler.setFormatter(
                 MillisecondAwareFormatter(
@@ -451,6 +536,9 @@ def initialize_logging(logger=None) -> logging.Logger:
             )
             handlers.append(handler)
 
+    # --logging_filename (with friends --logging_filename_count and
+    # --logging_filename_maxsize) set up logging to a file on the
+    # filesystem with automatic rotation when it gets too big.
     if config.config['logging_filename']:
         handler = RotatingFileHandler(
             config.config['logging_filename'],
@@ -465,6 +553,7 @@ def initialize_logging(logger=None) -> logging.Logger:
         )
         handlers.append(handler)
 
+    # --logging_console is, ahem, logging to the console.
     if config.config['logging_console']:
         handler = logging.StreamHandler(sys.stderr)
         handler.setFormatter(
@@ -477,23 +566,37 @@ def initialize_logging(logger=None) -> logging.Logger:
 
     if len(handlers) == 0:
         handlers.append(logging.NullHandler())
-
     for handler in handlers:
         logger.addHandler(handler)
 
+    # --logging_info_is_print echoes any message to logger.info(x) as
+    # a print statement on stdout.
     if config.config['logging_info_is_print']:
         handler = logging.StreamHandler(sys.stdout)
         handler.addFilter(OnlyInfoFilter())
         logger.addHandler(handler)
 
+    # --logging_squelch_repeats allows code to request repeat logging
+    # messages (identical log site and message contents) to be
+    # silenced.  Logging code must request this explicitly, it isn't
+    # automatic.  This option just allows the silencing to happen.
     if config.config['logging_squelch_repeats']:
         for handler in handlers:
             handler.addFilter(SquelchRepeatedMessagesFilter())
 
+    # --logging_probabilistically allows code to request
+    # non-deterministic logging where messages have some probability
+    # of being produced.  Logging code must request this explicitly.
+    # This option just allows the non-deterministic behavior to
+    # happen.  Disabling it will cause every log message to be
+    # produced.
     if config.config['logging_probabilistically']:
         for handler in handlers:
             handler.addFilter(ProbabilisticFilter())
 
+    # --lmodule is a way to have a special logging level for just on
+    # module or one set of modules that is different than the one set
+    # globally via --logging_level.
     for handler in handlers:
         handler.addFilter(
             DynamicPerScopeLoggingLevelFilter(
@@ -504,85 +607,30 @@ def initialize_logging(logger=None) -> logging.Logger:
     logger.setLevel(0)
     logger.propagate = False
 
+    # --logging_captures_prints, if set, will capture and log.info
+    # anything printed on stdout.
     if config.config['logging_captures_prints']:
         import builtins
 
-        global built_in_print
-
         def print_and_also_log(*arg, **kwarg):
             f = kwarg.get('file', None)
             if f == sys.stderr:
                 logger.warning(*arg)
             else:
                 logger.info(*arg)
-            built_in_print(*arg, **kwarg)
+            BUILT_IN_PRINT(*arg, **kwarg)
 
         builtins.print = print_and_also_log
 
     # At this point the logger is ready, handlers are set up,
     # etc... so log about the logging configuration.
-
-    level_name = logging._levelToName.get(
-        default_logging_level, str(default_logging_level)
+    log_about_logging(
+        logger,
+        default_logging_level,
+        preexisting_handlers_count,
+        fmt,
+        facility_name,
     )
-    logger.debug(f'Initialized global logging; default logging level is {level_name}.')
-    if (
-        config.config['logging_clear_preexisting_handlers']
-        and preexisting_handlers_count > 0
-    ):
-        msg = f'Logging cleared {preexisting_handlers_count} global handlers (--logging_clear_preexisting_handlers)'
-        logger.warning(msg)
-    logger.debug(f'Logging format specification is "{fmt}"')
-    if config.config['logging_debug_threads']:
-        logger.debug(
-            '...Logging format spec captures tid/pid (--logging_debug_threads)'
-        )
-    if config.config['logging_debug_modules']:
-        logger.debug(
-            '...Logging format spec captures files/functions/lineno (--logging_debug_modules)'
-        )
-    if config.config['logging_syslog']:
-        logger.debug(
-            f'Logging to syslog as {facility_name} with priority mapping based on level'
-        )
-    if config.config['logging_filename']:
-        logger.debug(f'Logging to filename {config.config["logging_filename"]}')
-        logger.debug(
-            f'...with {config.config["logging_filename_maxsize"]} bytes max file size.'
-        )
-        logger.debug(
-            f'...and {config.config["logging_filename_count"]} rotating backup file count.'
-        )
-    if config.config['logging_console']:
-        logger.debug('Logging to the console (stderr).')
-    if config.config['logging_info_is_print']:
-        logger.debug(
-            'Logging logger.info messages will be repeated on stdout (--logging_info_is_print)'
-        )
-    if config.config['logging_squelch_repeats']:
-        logger.debug(
-            'Logging code allowed to request repeated messages be squelched (--logging_squelch_repeats)'
-        )
-    else:
-        logger.debug(
-            'Logging code forbidden to request messages be squelched; all messages logged (--no_logging_squelch_repeats)'
-        )
-    if config.config['logging_probabilistically']:
-        logger.debug(
-            'Logging code is allowed to request probabilistic logging (--logging_probabilistically)'
-        )
-    else:
-        logger.debug(
-            'Logging code is forbidden to request probabilistic logging; messages always logged (--no_logging_probabilistically)'
-        )
-    if config.config['lmodule']:
-        logger.debug(
-            f'Logging dynamic per-module logging enabled (--lmodule={config.config["lmodule"]})'
-        )
-    if config.config['logging_captures_prints']:
-        logger.debug(
-            'Logging will capture printed data as logger.info messages (--logging_captures_prints)'
-        )
     return logger
 
 
@@ -628,18 +676,20 @@ class OutputMultiplexer(object):
         """Bits in the destination_bitv bitvector.  Used to indicate the
         output destination."""
 
-        LOG_DEBUG = 0x01  #  ⎫
-        LOG_INFO = 0x02  #  ⎪
-        LOG_WARNING = 0x04  #  ⎬ Must provide logger to the c'tor.
-        LOG_ERROR = 0x08  #  ⎪
+        # fmt: off
+        LOG_DEBUG = 0x01     #  ⎫
+        LOG_INFO = 0x02      #  ⎪
+        LOG_WARNING = 0x04   #  ⎬ Must provide logger to the c'tor.
+        LOG_ERROR = 0x08     #  ⎪
         LOG_CRITICAL = 0x10  #  ⎭
-        FILENAMES = 0x20  # Must provide a filename to the c'tor.
-        FILEHANDLES = 0x40  # Must provide a handle to the c'tor.
+        FILENAMES = 0x20     # Must provide a filename to the c'tor.
+        FILEHANDLES = 0x40   # Must provide a handle to the c'tor.
         HLOG = 0x80
         ALL_LOG_DESTINATIONS = (
             LOG_DEBUG | LOG_INFO | LOG_WARNING | LOG_ERROR | LOG_CRITICAL
         )
         ALL_OUTPUT_DESTINATIONS = 0x8F
+        # fmt: on
 
     def __init__(
         self,
@@ -653,15 +703,17 @@ class OutputMultiplexer(object):
             logger = logging.getLogger(None)
         self.logger = logger
 
+        self.f: Optional[List[Any]] = None
         if filenames is not None:
             self.f = [open(filename, 'wb', buffering=0) for filename in filenames]
         else:
-            if destination_bitv & OutputMultiplexer.FILENAMES:
+            if destination_bitv & OutputMultiplexer.Destination.FILENAMES:
                 raise ValueError("Filenames argument is required if bitv & FILENAMES")
             self.f = None
 
+        self.h: Optional[List[Any]] = None
         if handles is not None:
-            self.h = [handle for handle in handles]
+            self.h = list(handles)
         else:
             if destination_bitv & OutputMultiplexer.Destination.FILEHANDLES:
                 raise ValueError("Handle argument is required if bitv & FILEHANDLES")
@@ -755,7 +807,10 @@ class OutputMultiplexerContext(OutputMultiplexer, contextlib.ContextDecorator):
         handles=None,
     ):
         super().__init__(
-            destination_bitv, logger=logger, filenames=filenames, handles=handles
+            destination_bitv,
+            logger=logger,
+            filenames=filenames,
+            handles=handles,
         )
 
     def __enter__(self):