X-Git-Url: https://wannabe.guru.org/gitweb/?a=blobdiff_plain;f=bootstrap.py;h=117468623b94e578894561d0ff6fede8a9a33135;hb=532df2c5b57c7517dfb3dddd8c1358fbadf8baf3;hp=98da78cf6c1755c5fc3f5e42379e8d900abb10b7;hpb=a4bf4d05230474ad14243d67ac7f8c938f670e58;p=python_utils.git diff --git a/bootstrap.py b/bootstrap.py index 98da78c..1174686 100644 --- a/bootstrap.py +++ b/bootstrap.py @@ -1,39 +1,49 @@ #!/usr/bin/env python3 +# © Copyright 2021-2022, Scott Gasch + +"""This is a module for wrapping around python programs and doing some +minor setup and tear down work for them. With it, you can break into +pdb on unhandled top level exceptions, profile your code by passing a +commandline argument in, audit module import events, examine where +memory is being used in your program, and so on. + +""" + import functools +import importlib import logging import os -import importlib -from inspect import stack -from typing import List import sys +from inspect import stack + +import config +import logging_utils +from argparse_utils import ActionNoYes # This module is commonly used by others in here and should avoid # taking any unnecessary dependencies back on them. -from argparse_utils import ActionNoYes -import config -import logging_utils logger = logging.getLogger(__name__) -args = config.add_commandline_args( +cfg = config.add_commandline_args( f'Bootstrap ({__file__})', 'Args related to python program bootstrapper and Swiss army knife', ) -args.add_argument( +cfg.add_argument( '--debug_unhandled_exceptions', action=ActionNoYes, default=False, help='Break into pdb on top level unhandled exceptions.', ) -args.add_argument( +cfg.add_argument( '--show_random_seed', action=ActionNoYes, default=False, help='Should we display (and log.debug) the global random seed?', ) -args.add_argument( +cfg.add_argument( '--set_random_seed', type=int, nargs=1, @@ -41,32 +51,32 @@ args.add_argument( metavar='SEED_INT', help='Override the global random seed with a particular number.', ) -args.add_argument( +cfg.add_argument( '--dump_all_objects', action=ActionNoYes, default=False, help='Should we dump the Python import tree before main?', ) -args.add_argument( +cfg.add_argument( '--audit_import_events', action=ActionNoYes, default=False, help='Should we audit all import events?', ) -args.add_argument( +cfg.add_argument( '--run_profiler', action=ActionNoYes, default=False, help='Should we run cProfile on this code?', ) -args.add_argument( +cfg.add_argument( '--trace_memory', action=ActionNoYes, default=False, help='Should we record/report on memory utilization?', ) -original_hook = sys.excepthook +ORIGINAL_EXCEPTION_HOOK = sys.excepthook def handle_uncaught_exception(exc_type, exc_value, exc_tb): @@ -76,7 +86,6 @@ def handle_uncaught_exception(exc_type, exc_value, exc_tb): maybe attaches a debugger. """ - global original_hook msg = f'Unhandled top level exception {exc_type}' logger.exception(msg) print(msg, file=sys.stderr) @@ -86,7 +95,7 @@ def handle_uncaught_exception(exc_type, exc_value, exc_tb): else: if not sys.stderr.isatty() or not sys.stdin.isatty(): # stdin or stderr is redirected, just do the normal thing - original_hook(exc_type, exc_value, exc_tb) + ORIGINAL_EXCEPTION_HOOK(exc_type, exc_value, exc_tb) else: # a terminal is attached and stderr is not redirected, maybe debug. import traceback @@ -98,10 +107,19 @@ def handle_uncaught_exception(exc_type, exc_value, exc_tb): logger.info("Invoking the debugger...") pdb.pm() else: - original_hook(exc_type, exc_value, exc_tb) + ORIGINAL_EXCEPTION_HOOK(exc_type, exc_value, exc_tb) class ImportInterceptor(importlib.abc.MetaPathFinder): + """An interceptor that always allows module load events but dumps a + record into the log and onto stdout when modules are loaded and + produces an audit of who imported what at the end of the run. It + can't see any load events that happen before it, though, so move + bootstrap up in your __main__'s import list just temporarily to + get a good view. + + """ + def __init__(self): import collect.trie @@ -112,26 +130,25 @@ class ImportInterceptor(importlib.abc.MetaPathFinder): def repopulate_modules_by_filename(self): self.module_by_filename_cache.clear() - for mod in sys.modules: - if hasattr(sys.modules[mod], '__file__'): - fname = getattr(sys.modules[mod], '__file__') + for _, mod in sys.modules.copy().items(): # copy here because modules is volatile + if hasattr(mod, '__file__'): + fname = getattr(mod, '__file__') else: fname = 'unknown' self.module_by_filename_cache[fname] = mod - def should_ignore_filename(self, filename: str) -> bool: + @staticmethod + def should_ignore_filename(filename: str) -> bool: return 'importlib' in filename or 'six.py' in filename def find_module(self, fullname, path): - raise Exception( - "This method has been deprecated since Python 3.4, please upgrade." - ) + raise Exception("This method has been deprecated since Python 3.4, please upgrade.") - def find_spec(self, loaded_module, path=None, target=None): + def find_spec(self, loaded_module, path=None, _=None): s = stack() for x in range(3, len(s)): filename = s[x].filename - if self.should_ignore_filename(filename): + if ImportInterceptor.should_ignore_filename(filename): continue loading_function = s[x].function @@ -172,15 +189,14 @@ class ImportInterceptor(importlib.abc.MetaPathFinder): # # Also note: move bootstrap up in the global import list to catch # more import events and have a more complete record. -import_interceptor = None +IMPORT_INTERCEPTOR = None for arg in sys.argv: if arg == '--audit_import_events': - import_interceptor = ImportInterceptor() - sys.meta_path.insert(0, import_interceptor) + IMPORT_INTERCEPTOR = ImportInterceptor() + sys.meta_path.insert(0, IMPORT_INTERCEPTOR) def dump_all_objects() -> None: - global import_interceptor messages = {} all_modules = sys.modules for obj in object.__subclasses__(): @@ -200,8 +216,8 @@ def dump_all_objects() -> None: mod_file = mod.__file__ else: mod_file = 'unknown' - if import_interceptor is not None: - import_path = import_interceptor.find_importer(mod_name) + if IMPORT_INTERCEPTOR is not None: + import_path = IMPORT_INTERCEPTOR.find_importer(mod_name) else: import_path = 'unknown' msg = f'{class_mod_name}::{klass} ({mod_file})' @@ -229,13 +245,16 @@ def initialize(entry_point): # Try to figure out the name of the program entry point. Then # parse configuration (based on cmdline flags, environment vars # etc...) - if ( - '__globals__' in entry_point.__dict__ - and '__file__' in entry_point.__globals__ - ): - config.parse(entry_point.__globals__['__file__']) - else: - config.parse(None) + entry_filename = None + entry_descr = None + try: + entry_filename = entry_point.__code__.co_filename + entry_descr = entry_point.__code__.__repr__() + except Exception: + if '__globals__' in entry_point.__dict__ and '__file__' in entry_point.__globals__: + entry_filename = entry_point.__globals__['__file__'] + entry_descr = entry_filename + config.parse(entry_filename) if config.config['trace_memory']: import tracemalloc @@ -249,12 +268,21 @@ def initialize(entry_point): # Maybe log some info about the python interpreter itself. logger.debug( - f'Platform: {sys.platform}, maxint=0x{sys.maxsize:x}, byteorder={sys.byteorder}' + 'Platform: %s, maxint=0x%x, byteorder=%s', sys.platform, sys.maxsize, sys.byteorder ) - logger.debug(f'Python interpreter version: {sys.version}') - logger.debug(f'Python implementation: {sys.implementation}') - logger.debug(f'Python C API version: {sys.api_version}') - logger.debug(f'Python path: {sys.path}') + logger.debug('Python interpreter version: %s', sys.version) + logger.debug('Python implementation: %s', sys.implementation) + logger.debug('Python C API version: %s', sys.api_version) + if __debug__: + logger.debug('Python interpreter running in __debug__ mode.') + else: + logger.debug('Python interpreter running in optimized mode.') + logger.debug('Python path: %s', sys.path) + + # Log something about the site_config, many things use it. + import site_config + + logger.debug('Global site_config: %s', site_config.get_config()) # Allow programs that don't bother to override the random seed # to be replayed via the commandline. @@ -273,7 +301,7 @@ def initialize(entry_point): random.seed(random_seed) # Do it, invoke the user's code. Pay attention to how long it takes. - logger.debug(f'Starting {entry_point.__name__} (program entry point)') + logger.debug('Starting %s (program entry point)', entry_descr) ret = None import stopwatch @@ -293,7 +321,7 @@ def initialize(entry_point): with stopwatch.Timer() as t: ret = entry_point(*args, **kwargs) - logger.debug(f'{entry_point.__name__} (program entry point) returned {ret}.') + logger.debug('%s (program entry point) returned %s.', entry_descr, ret) if config.config['trace_memory']: snapshot = tracemalloc.take_snapshot() @@ -307,27 +335,32 @@ def initialize(entry_point): dump_all_objects() if config.config['audit_import_events']: - global import_interceptor - if import_interceptor is not None: - print(import_interceptor.tree) + if IMPORT_INTERCEPTOR is not None: + print(IMPORT_INTERCEPTOR.tree) walltime = t() (utime, stime, cutime, cstime, elapsed_time) = os.times() logger.debug( '\n' - f'user: {utime}s\n' - f'system: {stime}s\n' - f'child user: {cutime}s\n' - f'child system: {cstime}s\n' - f'machine uptime: {elapsed_time}s\n' - f'walltime: {walltime}s' + 'user: %.4fs\n' + 'system: %.4fs\n' + 'child user: %.4fs\n' + 'child system: %.4fs\n' + 'machine uptime: %.4fs\n' + 'walltime: %.4fs', + utime, + stime, + cutime, + cstime, + elapsed_time, + walltime, ) # If it doesn't return cleanly, call attention to the return value. if ret is not None and ret != 0: - logger.error(f'Exit {ret}') + logger.error('Exit %s', ret) else: - logger.debug(f'Exit {ret}') + logger.debug('Exit %s', ret) sys.exit(ret) return initialize_wrapper