Optionally surface exceptions that happen under executors by reading
[python_utils.git] / bootstrap.py
index 738fceaf39003141f08f9e39a7843c09f1239b43..e50cb386543aaef66fdd36cb9c149fbf009ef5ae 100644 (file)
@@ -17,18 +17,19 @@ logger = logging.getLogger(__name__)
 
 args = config.add_commandline_args(
     f'Bootstrap ({__file__})',
-    'Args related to python program bootstrapper and Swiss army knife')
+    'Args related to python program bootstrapper and Swiss army knife',
+)
 args.add_argument(
     '--debug_unhandled_exceptions',
     action=ActionNoYes,
     default=False,
-    help='Break into pdb on top level unhandled exceptions.'
+    help='Break into pdb on top level unhandled exceptions.',
 )
 args.add_argument(
     '--show_random_seed',
     action=ActionNoYes,
     default=False,
-    help='Should we display (and log.debug) the global random seed?'
+    help='Should we display (and log.debug) the global random seed?',
 )
 args.add_argument(
     '--set_random_seed',
@@ -36,13 +37,13 @@ args.add_argument(
     nargs=1,
     default=None,
     metavar='SEED_INT',
-    help='Override the global random seed with a particular number.'
+    help='Override the global random seed with a particular number.',
 )
 args.add_argument(
     '--dump_all_objects',
     action=ActionNoYes,
     default=False,
-    help='Should we dump the Python import tree before main?'
+    help='Should we dump the Python import tree before main?',
 )
 args.add_argument(
     '--audit_import_events',
@@ -50,7 +51,18 @@ args.add_argument(
     default=False,
     help='Should we audit all import events?',
 )
-
+args.add_argument(
+    '--run_profiler',
+    action=ActionNoYes,
+    default=False,
+    help='Should we run cProfile on this code?',
+)
+args.add_argument(
+    '--trace_memory',
+    action=ActionNoYes,
+    default=False,
+    help='Should we record/report on memory utilization?',
+)
 
 original_hook = sys.excepthook
 
@@ -70,18 +82,17 @@ def handle_uncaught_exception(exc_type, exc_value, exc_tb):
         sys.__excepthook__(exc_type, exc_value, exc_tb)
         return
     else:
-        if (
-                not sys.stderr.isatty() or
-                not sys.stdin.isatty()
-        ):
+        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)
         else:
             # a terminal is attached and stderr is not redirected, maybe debug.
             import traceback
+
             traceback.print_exception(exc_type, exc_value, exc_tb)
             if config.config['debug_unhandled_exceptions']:
                 import pdb
+
                 logger.info("Invoking the debugger...")
                 pdb.pm()
             else:
@@ -91,6 +102,7 @@ def handle_uncaught_exception(exc_type, exc_value, exc_tb):
 class ImportInterceptor(object):
     def __init__(self):
         import collect.trie
+
         self.module_by_filename_cache = {}
         self.repopulate_modules_by_filename()
         self.tree = collect.trie.Trie()
@@ -142,18 +154,6 @@ class ImportInterceptor(object):
         return []
 
 
-# # TODO: test this with python 3.8+
-# def audit_import_events(event, args):
-#     if event == 'import':
-#         module = args[0]
-#         filename = args[1]
-#         sys_path = args[2]
-#         sys_meta_path = args[3]
-#         sys_path_hooks = args[4]
-#         logger.debug(msg)
-#         print(msg)
-
-
 # Audit import events?  Note: this runs early in the lifetime of the
 # process (assuming that import bootstrap happens early); config has
 # (probably) not yet been loaded or parsed the commandline.  Also,
@@ -167,12 +167,6 @@ for arg in sys.argv:
     if arg == '--audit_import_events':
         import_interceptor = ImportInterceptor()
         sys.meta_path = [import_interceptor] + sys.meta_path
-        # if not hasattr(sys, 'frozen'):
-        #     if (
-        #             sys.version_info[0] == 3
-        #             and sys.version_info[1] >= 8
-        #     ):
-        #         sys.addaudithook(audit_import_events)
 
 
 def dump_all_objects() -> None:
@@ -215,6 +209,7 @@ def initialize(entry_point):
     seed, etc... before running main.
 
     """
+
     @functools.wraps(entry_point)
     def initialize_wrapper(*args, **kwargs):
         # Hook top level unhandled exceptions, maybe invoke debugger.
@@ -225,21 +220,36 @@ def initialize(entry_point):
         # parse configuration (based on cmdline flags, environment vars
         # etc...)
         if (
-                '__globals__' in entry_point.__dict__ and
-                '__file__' in entry_point.__globals__
+            '__globals__' in entry_point.__dict__
+            and '__file__' in entry_point.__globals__
         ):
             config.parse(entry_point.__globals__['__file__'])
         else:
             config.parse(None)
 
+        if config.config['trace_memory']:
+            import tracemalloc
+
+            tracemalloc.start()
+
         # Initialize logging... and log some remembered messages from
         # config module.
         logging_utils.initialize_logging(logging.getLogger())
         config.late_logging()
 
+        # Maybe log some info about the python interpreter itself.
+        logger.debug(
+            f'Platform: {sys.platform}, maxint=0x{sys.maxsize:x}, byteorder={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}')
+
         # Allow programs that don't bother to override the random seed
         # to be replayed via the commandline.
         import random
+
         random_seed = config.config['set_random_seed']
         if random_seed is not None:
             random_seed = random_seed[0]
@@ -256,11 +266,32 @@ def initialize(entry_point):
         logger.debug(f'Starting {entry_point.__name__} (program entry point)')
         ret = None
         import stopwatch
-        with stopwatch.Timer() as t:
-            ret = entry_point(*args, **kwargs)
-        logger.debug(
-            f'{entry_point.__name__} (program entry point) returned {ret}.'
-        )
+
+        if config.config['run_profiler']:
+            import cProfile
+            from pstats import SortKey
+
+            with stopwatch.Timer() as t:
+                cProfile.runctx(
+                    "ret = entry_point(*args, **kwargs)",
+                    globals(),
+                    locals(),
+                    None,
+                    SortKey.CUMULATIVE,
+                )
+        else:
+            with stopwatch.Timer() as t:
+                ret = entry_point(*args, **kwargs)
+
+        logger.debug(f'{entry_point.__name__} (program entry point) returned {ret}.')
+
+        if config.config['trace_memory']:
+            snapshot = tracemalloc.take_snapshot()
+            top_stats = snapshot.statistics('lineno')
+            print()
+            print("--trace_memory's top 10 memory using files:")
+            for stat in top_stats[:10]:
+                print(stat)
 
         if config.config['dump_all_objects']:
             dump_all_objects()
@@ -272,13 +303,15 @@ def initialize(entry_point):
 
         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')
+        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'
+        )
 
         # If it doesn't return cleanly, call attention to the return value.
         if ret is not None and ret != 0:
@@ -286,4 +319,5 @@ def initialize(entry_point):
         else:
             logger.debug(f'Exit {ret}')
         sys.exit(ret)
+
     return initialize_wrapper