"""Script for testing the performance of pickling/unpickling. This will pickle/unpickle several real world-representative objects a few thousand times. The methodology below was chosen for was chosen to be similar to real-world scenarios which operate on single objects at a time. Note that if we did something like pickle.dumps([dict(some_dict) for _ in range(10000)]) this isn't equivalent to dumping the dict 10000 times: pickle uses a highly-efficient encoding for the n-1 following copies. """ import datetime import random import sys import pyperf from memray_helper import get_tracker IS_PYPY = pyperf.python_implementation() == "pypy" __author__ = "collinwinter@google.com (Collin Winter)" DICT = { "ads_flags": 0, "age": 18, "birthday": datetime.date(1980, 5, 7), "bulletin_count": 0, "comment_count": 0, "country": "BR", "encrypted_id": "G9urXXAJwjE", "favorite_count": 9, "first_name": "", "flags": 412317970704, "friend_count": 0, "gender": "m", "gender_for_display": "Male", "id": 302935349, "is_custom_profile_icon": 0, "last_name": "", "locale_preference": "pt_BR", "member": 0, "tags": ["a", "b", "c", "d", "e", "f", "g"], "profile_foo_id": 827119638, "secure_encrypted_id": "Z_xxx2dYx3t4YAdnmfgyKw", "session_number": 2, "signup_id": "201-19225-223", "status": "A", "theme": 1, "time_created": 1225237014, "time_updated": 1233134493, "unread_message_count": 0, "user_group": "0", "username": "collinwinter", "play_count": 9, "view_count": 7, "zip": "", } TUPLE = ( [ 265867233, 265868503, 265252341, 265243910, 265879514, 266219766, 266021701, 265843726, 265592821, 265246784, 265853180, 45526486, 265463699, 265848143, 265863062, 265392591, 265877490, 265823665, 265828884, 265753032, ], 60, ) def mutate_dict(orig_dict, random_source): new_dict = dict(orig_dict) for key, value in new_dict.items(): rand_val = random_source.random() * sys.maxsize if isinstance(key, (int, bytes, str)): new_dict[key] = type(key)(rand_val) return new_dict random_source = random.Random(5) # Fixed seed. DICT_GROUP = [mutate_dict(DICT, random_source) for _ in range(3)] def bench_pickle(loops, pickle, options): range_it = range(loops) # micro-optimization: use fast local variables dumps = pickle.dumps objs = (DICT, TUPLE, DICT_GROUP) protocol = options.protocol with get_tracker(): t0 = pyperf.perf_counter() for _ in range_it: for obj in objs: # 20 dumps dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) return pyperf.perf_counter() - t0 def bench_unpickle(loops, pickle, options): pickled_dict = pickle.dumps(DICT, options.protocol) pickled_tuple = pickle.dumps(TUPLE, options.protocol) pickled_dict_group = pickle.dumps(DICT_GROUP, options.protocol) range_it = range(loops) # micro-optimization: use fast local variables loads = pickle.loads objs = (pickled_dict, pickled_tuple, pickled_dict_group) with get_tracker(): t0 = pyperf.perf_counter() for _ in range_it: for obj in objs: # 20 loads dict loads(obj) loads(obj) loads(obj) loads(obj) loads(obj) loads(obj) loads(obj) loads(obj) loads(obj) loads(obj) loads(obj) loads(obj) loads(obj) loads(obj) loads(obj) loads(obj) loads(obj) loads(obj) loads(obj) loads(obj) return pyperf.perf_counter() - t0 LIST = [[list(range(10)), list(range(10))] for _ in range(10)] def bench_pickle_list(loops, pickle, options): range_it = range(loops) # micro-optimization: use fast local variables dumps = pickle.dumps obj = LIST protocol = options.protocol with get_tracker(): t0 = pyperf.perf_counter() for _ in range_it: # 10 dumps list dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) dumps(obj, protocol) return pyperf.perf_counter() - t0 def bench_unpickle_list(loops, pickle, options): pickled_list = pickle.dumps(LIST, options.protocol) range_it = range(loops) # micro-optimization: use fast local variables loads = pickle.loads with get_tracker(): t0 = pyperf.perf_counter() for _ in range_it: # 10 loads list loads(pickled_list) loads(pickled_list) loads(pickled_list) loads(pickled_list) loads(pickled_list) loads(pickled_list) loads(pickled_list) loads(pickled_list) loads(pickled_list) loads(pickled_list) return pyperf.perf_counter() - t0 MICRO_DICT = dict((key, dict.fromkeys(range(10))) for key in range(100)) def bench_pickle_dict(loops, pickle, options): range_it = range(loops) # micro-optimization: use fast local variables protocol = options.protocol obj = MICRO_DICT with get_tracker(): t0 = pyperf.perf_counter() for _ in range_it: # 5 dumps dict pickle.dumps(obj, protocol) pickle.dumps(obj, protocol) pickle.dumps(obj, protocol) pickle.dumps(obj, protocol) pickle.dumps(obj, protocol) return pyperf.perf_counter() - t0 BENCHMARKS = { # 20 inner-loops: don't count the 3 pickled objects "pickle": (bench_pickle, 20), # 20 inner-loops: don't count the 3 unpickled objects "unpickle": (bench_unpickle, 20), "pickle_list": (bench_pickle_list, 10), "unpickle_list": (bench_unpickle_list, 10), "pickle_dict": (bench_pickle_dict, 5), } def is_accelerated_module(module): return getattr(module.Pickler, "__module__", "") != "pickle" def add_cmdline_args(cmd, args): if args.pure_python: cmd.append("--pure-python") cmd.extend(("--protocol", str(args.protocol))) cmd.append(args.benchmark) if __name__ == "__main__": runner = pyperf.Runner(add_cmdline_args=add_cmdline_args) runner.metadata["description"] = "Test the performance of pickling." parser = runner.argparser parser.add_argument( "--pure-python", action="store_true", help="Use the C version of pickle." ) parser.add_argument( "--protocol", action="store", default=None, type=int, help="Which protocol to use (default: highest protocol).", ) benchmarks = sorted(BENCHMARKS) parser.add_argument("benchmark", choices=benchmarks) options = runner.parse_args() benchmark, inner_loops = BENCHMARKS[options.benchmark] name = options.benchmark if options.pure_python: name += "_pure_python" if not (options.pure_python or IS_PYPY): # C accelerators are enabled by default on 3.x import pickle if not is_accelerated_module(pickle): raise RuntimeError("Missing C accelerators for pickle") else: sys.modules["_pickle"] = None import pickle if is_accelerated_module(pickle): raise RuntimeError("Unexpected C accelerators for pickle") if options.protocol is None: options.protocol = pickle.HIGHEST_PROTOCOL runner.metadata["pickle_protocol"] = str(options.protocol) runner.metadata["pickle_module"] = pickle.__name__ runner.bench_time_func(name, benchmark, pickle, options, inner_loops=inner_loops)