Test your AI Agents with the all-new Agent to Agent Testing Platform.Learn More

How to use ld15iqr method in pytest-benchmark

Best Python code snippet using pytest-benchmark

benchmark.py

Source:benchmark.py

1# Copyright 2021 Xilinx Inc.2#3# Licensed under the Apache License, Version 2.0 (the "License");4# you may not use this file except in compliance with the License.5# You may obtain a copy of the License at6#7#      http://www.apache.org/licenses/LICENSE-2.08#9# Unless required by applicable law or agreed to in writing, software10# distributed under the License is distributed on an "AS IS" BASIS,11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.12# See the License for the specific language governing permissions and13# limitations under the License.14import argparse15import copy16from enum import Enum, auto17import functools18import ipaddress19import json20import itertools21import math22import os23import pathlib24import pprint25import socket26import statistics27import subprocess28import sys29from typing import Optional30import proteus31from proteus.server import Server32from proteus.rest import Client33import pytest_benchmark.utils34from _pytest.mark import KeywordMatcher35from _pytest.mark.expression import Expression36from rich.console import Console37from rich.table import Table38from rich.progress import Progress39import yaml40class Highlight(Enum):41    smallest = auto()42    largest = auto()43    none = auto()44class Options:45    def __init__(self, options) -> None:46        self.options = options47    def __getattr__(self, name):48        if name in self.options:49            return self.options[name]50        raise AttributeError(f"'Options' object has no attribute '{name}'")51class Config:52    def __init__(self, path) -> None:53        with open(path, "r") as f:54            self.config = yaml.safe_load(f)55        self.pytest = Options(self.config["pytest"])56        self.wrk = Options(self.config["wrk"])57        self.cpp = Options(self.config["cpp"])58    @property59    def benchmarks(self):60        return self.config["benchmarks"]61    @benchmarks.setter62    def benchmarks(self, value):63        self.config["benchmarks"] = value64    @property65    def repeat(self):66        return self.config["repeat_count"]67    @property68    def verbosity(self):69        return self.config["verbosity"]70    @property71    def http_address(self):72        return self.config["http_address"]73    @property74    def start_server(self):75        return self.config["start_local_server"]76    def __str__(self):77        return pprint.pformat(self.config, 2)78class Benchmark:79    """80    This class holds a single group's benchmark results. Each statistic and81    metadata associated with a benchmark is stored in lists internally so it82    can be plotted in order when printing the table to the console.83    """84    def __init__(self, benchmark: dict) -> None:85        """86        Initialize the internal lists using the stats from the initial benchmark87        used to create this object88        Args:89            benchmark (dict): The dict defining the benchmark. This is in the90            same format as output from pytest_benchmark.91        """92        try:93            load = benchmark["load"]94            benchmark_type = benchmark["type"]95            benchmark_config = benchmark["config"]96            self.stats = {97                "load": [load],98                "type": [benchmark_type],99                "config": [benchmark_config],100                "name": [benchmark["name"]],101                "min": [benchmark["stats"]["min"]],102                "max": [benchmark["stats"]["max"]],103                "mean": [benchmark["stats"]["mean"]],104                "stddev": [benchmark["stats"]["stddev"]],105                "rounds": [benchmark["stats"]["rounds"]],106                "median": [benchmark["stats"]["median"]],107                "iqr": [benchmark["stats"]["iqr"]],108                "q1": [benchmark["stats"]["q1"]],109                "q3": [benchmark["stats"]["q3"]],110                "iqr_outliers": [benchmark["stats"]["iqr_outliers"]],111                "stddev_outliers": [benchmark["stats"]["stddev_outliers"]],112                "outliers": [benchmark["stats"]["outliers"]],113                "ld15iqr": [benchmark["stats"]["ld15iqr"]],114                "hd15iqr": [benchmark["stats"]["hd15iqr"]],115                "ops": [benchmark["stats"]["ops"]],116                "ops_uncertainty": [benchmark["stats"]["ops_uncertainty"]],117                "total": [benchmark["stats"]["total"]],118                "iterations": [benchmark["stats"]["iterations"]],119            }120        except KeyError:121            print(122                "This benchmark cannot be opened as it has not been normalized due to an error"123            )124            sys.exit(1)125        self._final_stats = {}126    def add(self, benchmark: dict):127        """128        Add a benchmark to this group129        Args:130            benchmark (dict): The dict defining the benchmark. This is in the131            same format as output from pytest_benchmark.132        """133        load = benchmark["load"]134        benchmark_type = benchmark["type"]135        benchmark_config = benchmark["config"]136        self.stats["name"].append(benchmark["name"])137        self.stats["load"].append(load)138        self.stats["type"].append(benchmark_type)139        self.stats["config"].append(benchmark_config)140        for key in self.stats.keys():141            try:142                self.stats[key].append(benchmark["stats"][key])143            except KeyError:144                continue145    @staticmethod146    def _data_format(arg):147        """148        Define the highlighting and formatting rules for the different metadata149        in the benchmarks150        Args:151            arg (str): The metadata type to format (e.g. load, min, max etc.)152        Returns:153            partial([data]): A function that accepts [data] of the type of arg and154            returns it as a list of formatted strings for rich to print out155        """156        data_format = {157            "load": {"highlight": Highlight.none, "format": "d"},158            "name": {"highlight": Highlight.none, "format": "s"},159            "type": {"highlight": Highlight.none, "format": "s"},160            "config": {"highlight": Highlight.none, "format": "s"},161            "min": {"highlight": Highlight.smallest, "format": "1.2e"},162            "max": {"highlight": Highlight.smallest, "format": "1.2e"},163            "mean": {"highlight": Highlight.smallest, "format": "1.2e"},164            "stddev": {"highlight": Highlight.smallest, "format": "1.2e"},165            "rounds": {"highlight": Highlight.none, "format": "1.2e"},166            "median": {"highlight": Highlight.smallest, "format": "1.2e"},167            "iqr": {"highlight": Highlight.smallest, "format": "1.2e"},168            "q1": {"highlight": Highlight.smallest, "format": "1.2e"},169            "q3": {"highlight": Highlight.smallest, "format": "1.2e"},170            "iqr_outliers": {"highlight": Highlight.none, "format": "1.2e"},171            "stddev_outliers": {"highlight": Highlight.none, "format": "1.2e"},172            "outliers": {"highlight": Highlight.none, "format": "1.2e"},173            "ld15iqr": {"highlight": Highlight.none, "format": "1.2e"},174            "hd15iqr": {"highlight": Highlight.none, "format": "1.2e"},175            "ops": {"highlight": Highlight.largest, "format": ",.2f"},176            "ops_uncertainty": {"highlight": Highlight.smallest, "format": ",.2f"},177            "total": {"highlight": Highlight.none, "format": "1.2e"},178            "iterations": {"highlight": Highlight.none, "format": "1.2e"},179        }180        def format_func(colors, data):181            """182            Formats the list of data using the given colors to highlight the min183            or max if formatting is enabled for the data type184            Args:185                colors (iterable): Iterable of colors (first is used for min, second for max) or None186                data (iterable): List of data to format187            Returns:188                list: Formatted data in rich format189            """190            numeric_values = [191                value for value in data if isinstance(value, (int, float))192            ]193            if numeric_values:194                min_value = min(numeric_values)195                max_value = max(numeric_values)196                enable_colors = len(numeric_values) != 1197            formatted_data = []198            for datum in data:199                if isinstance(datum, (int, float)):200                    if colors is not None and enable_colors:201                        if datum == min_value:202                            formatted_data.append(203                                f"[{colors[0]}]{datum:{data_format[arg]['format']}}[/{colors[0]}]"204                            )205                        elif datum == max_value:206                            formatted_data.append(207                                f"[{colors[1]}]{datum:{data_format[arg]['format']}}[/{colors[1]}]"208                            )209                        else:210                            formatted_data.append(211                                f"{datum:{data_format[arg]['format']}}"212                            )213                    else:214                        formatted_data.append(f"{datum:{data_format[arg]['format']}}")215                else:216                    formatted_data.append(str(datum))217            return formatted_data218        if data_format[arg]["highlight"] != Highlight.none:219            if data_format[arg]["highlight"] == Highlight.smallest:220                return functools.partial(format_func, ("green", "red"))221            return functools.partial(format_func, ("red", "green"))222        return functools.partial(format_func, None)223    def finalize(self, *args):224        """225        This should be called prior to printing the benchmark. The arguments to this226        function are the stats that are going to be printed. For these stats,227        it computes the formatted strings and saves them for rich to print later.228        """229        for arg in args:230            format_func = self._data_format(arg)231            self._final_stats[arg] = format_func(self.stats[arg])232    def has_type(self, search_str: str):233        for benchmark_type in self.stats["type"]:234            if search_str.upper() in str(benchmark_type).upper():235                return True236        return False237    def get_row(self):238        """239        Get rows of data for this benchmark based on what was done by finalize()240        Yields:241            list[str]: one row242        """243        for index, _ in enumerate(self.stats["min"]):244            yield [245                str(self._final_stats[arg][index]) for arg in self._final_stats.keys()246            ]247class Benchmarks:248    """249    This class holds all the benchmarks and can be used to print them to console.250    """251    def __init__(self, benchmark, path=None, normalize=False) -> None:252        if normalize:253            normalized_benchmark = self._normalize(benchmark)254            self._benchmark = normalized_benchmark255        else:256            self._benchmark = benchmark257        self.benchmarks = {}258        self.path = path259        for test in benchmark["benchmarks"]:260            self._analyze(test)261    @property262    def machine_info(self):263        return self._benchmark["machine_info"]264    @property265    def commit_info(self):266        return self._benchmark["commit_info"]267    def _analyze(self, test: dict):268        """269        For each benchmark, add a new Benchmark object for it if it's new or add270        it to an existing object271        Args:272            test (dict): The dict defining the benchmark. This is in the273            same format as output from pytest_benchmark.274        """275        group = test["group"]276        if group not in self.benchmarks:277            self.benchmarks[group] = Benchmark(test)278        else:279            self.benchmarks[group].add(test)280    def _normalize(self, benchmarks: dict):281        normalized_benchmarks = []282        for benchmark in benchmarks["benchmarks"]:283            if "load" not in benchmark:284                benchmark["load"] = 1285            if "type" not in benchmark:286                if "type" in benchmark["extra_info"]:287                    benchmark["type"] = benchmark["extra_info"]["type"]288                else:289                    benchmark["type"] = "Unknown"290            if "config" not in benchmark:291                if "config" in benchmark["extra_info"]:292                    benchmark["config"] = benchmark["extra_info"]["config"]293                else:294                    benchmark["config"] = "Unknown"295            if "ops_uncertainty" not in benchmark["stats"]:296                benchmark["stats"]["ops_uncertainty"] = benchmark["stats"]["ops"] - (297                    1 / (benchmark["stats"]["mean"] + benchmark["stats"]["stddev"])298                )299            normalized_benchmarks.append(benchmark)300        benchmarks["benchmarks"] = normalized_benchmarks301        return benchmarks302    def add(self, benchmark):303        self._benchmark["benchmarks"].append(benchmark)304        self._analyze(benchmark)305    def get(self) -> list:306        return self._benchmark["benchmarks"]307    def has_type(self, search_str):308        for _, benchmark in self.benchmarks.items():309            if benchmark.has_type(search_str):310                return True311        return False312    def write(self, path=None):313        if path is None:314            path = self.path315        if path is None:316            print(f"No path specified to write to")317            return318        with open(path, "w") as f:319            json.dump(self._benchmark, f, indent=4)320    def clear(self):321        self.benchmarks = {}322        self._benchmark["benchmarks"] = []323    def print(self):324        console = Console()325        table = Table(326            show_header=True,327            header_style="bold magenta",328            title="Legend",329            title_style="bold yellow",330        )331        table.add_column("Term")332        table.add_column("Meaning", max_width=80)333        show_legend = False334        if self.has_type("wrk"):335            table.add_row(336                "Tx:Cy:tz",337                "Indicates wrk's configuration i.e. that the wrk application was run with x threads, y total TCP connections and for z time",338            )339            show_legend = True340        if self.has_type("cpp"):341            table.add_row(342                "Ix:Ty",343                "Indicates the cpp benchmark's configuration i.e. it was run with x images and y threads",344            )345            show_legend = True346        if show_legend:347            console.print(table)348        for key, value in self.benchmarks.items():349            table = Table(350                show_header=True,351                header_style="bold magenta",352                title=key,353                title_style="bold yellow",354            )355            table.add_column("Name")356            table.add_column("Type")357            table.add_column("Config")358            table.add_column("Workers")359            table.add_column("Min (s)")360            table.add_column("Max (s)")361            table.add_column("Mean (s)")362            table.add_column("StdDev (s)")363            # table.add_column("Median (s)")364            table.add_column("Op/s")365            table.add_column("Op/s +/-")366            value.finalize(367                "name",368                "type",369                "config",370                "load",371                "min",372                "max",373                "mean",374                "stddev",375                "ops",376                "ops_uncertainty",377            )378            for row in value.get_row():379                table.add_row(*row)380            console.print(table)381    def __str__(self):382        return json.dumps(self._benchmark)383def get_last_benchmark_path(index) -> Optional[str]:384    """385    Get the path to the last generated benchmark in the benchmark directory386    Args:387        index (int): return the nth benchmark (0 means newest)388    Returns:389        Optional[str]: Path to the benchmark if found390    """391    dir = os.getenv("PROTEUS_ROOT") + "/tests/.benchmarks"392    machine_id = pytest_benchmark.utils.get_machine_id()393    list_of_files = pathlib.Path(f"{dir}/{machine_id}").rglob("*.json")394    if list_of_files:395        return sorted(list_of_files, key=os.path.getctime, reverse=True)[index]396    return None397def get_benchmark(398    path: Optional[str] = None, index=0, normalize=False399) -> Optional[Benchmarks]:400    """401    Get the last nth generated benchmark from the directory402    Args:403        path (Optional[str], optional): Get a specific benchmark. Defaults to None.404        index (int, optional): Get the nth newest benchmark. Defaults to 0.405        normalize (bool, optional): Normalize the benchmark. Defaults to False.406    Returns:407        Optional[Benchmarks]: The chosen benchmark or None if an empty directory is used408    """409    if path is not None:410        last_benchmark = path411    else:412        last_benchmark = get_last_benchmark_path(index)413    if last_benchmark is not None:414        with open(last_benchmark, "r") as f:415            return Benchmarks(json.load(f), last_benchmark, normalize)416    return None417def parse_wrk_output(output: str) -> dict:418    """419    Parse the terminal output of wrk into a dictionary. The output format is420    defined by write_lua() in the tests.421    Args:422        output (str): Raw terminal output from wrk423    Returns:424        dict: Organized results from wrk425    """426    lines = output.split("\n")427    raw_stats = lines[-4:-1]428    latencies = raw_stats[0].split(",")429    requests = raw_stats[1].split(",")430    summaries = raw_stats[2].split(",")431    stats = {432        "latencies": {},433        "requests": {},434        "summaries": {},435    }436    metrics = ["min", "max", "mean", "stdev"]437    metrics_summary = ["duration", "requests", "bytes"]438    for latency, label in zip(latencies[:2], metrics[:2]):439        stats["latencies"][label] = int(latency)440    for latency, label in zip(latencies[2:], metrics[2:]):441        stats["latencies"][label] = float(latency)442    for request, label in zip(requests[:2], metrics[:2]):443        stats["requests"][label] = int(request)444    for request, label in zip(requests[2:], metrics[2:]):445        stats["requests"][label] = float(request)446    for summary, label in zip(summaries, metrics_summary):447        stats["summaries"][label] = int(summary)448    return stats449BASE_BENCHMARK = {450    "group": "facedetect_dpucadf8h",451    "name": "test_benchmark_facedetect_dpucadf8h_1",452    "fullname": "test_facedetect.py::TestInferImageFacedetectDPUCADF8H::test_benchmark_facedetect_dpucadf8h_1",453    "params": None,454    "param": None,455    "stats": {456        "min": 0.0855791429639794,457        "max": 0.10293014405760914,458        "mean": 0.093361033460083,459        "stddev": 0.005477697831711018,460        "rounds": 11,461        "median": 0.09155373001703992,462        "iqr": 0.005927730540861376,463        "q1": 0.09056435847014654,464        "q3": 0.09649208901100792,465        "iqr_outliers": 0,466        "stddev_outliers": 4,467        "outliers": "4;0",468        "ld15iqr": 0.0855791429639794,469        "hd15iqr": 0.10293014405760914,470        "ops": 10.71110679625837,471        "ops_uncertainty": 10.71110679625837,472        "total": 1.026971368060913,473        "iterations": 1,474    },475}476def make_wrk_benchmarks(raw_stats, benchmark, wrk_config, load):477    wrk_setting = f"T{wrk_config[0]}:C{wrk_config[1]}:t{wrk_config[2]}"478    benchmark_wrk = copy.deepcopy(benchmark)479    # clear other stats from pytest-benchmark480    del benchmark_wrk["extra_info"]481    del benchmark_wrk["options"]482    benchmark_wrk["load"] = load483    benchmark_wrk["type"] = "rest (wrk)"484    benchmark_wrk["config"] = wrk_setting485    benchmark_wrk["name"] = benchmark["name"]486    benchmark_wrk["stats"]["rounds"] = 1487    benchmark_wrk["stats"]["median"] = None488    benchmark_wrk["stats"]["iqr"] = None489    benchmark_wrk["stats"]["q1"] = None490    benchmark_wrk["stats"]["q3"] = None491    benchmark_wrk["stats"]["iqr_outliers"] = None492    benchmark_wrk["stats"]["stddev_outliers"] = None493    benchmark_wrk["stats"]["outliers"] = None494    benchmark_wrk["stats"]["ld15iqr"] = None495    benchmark_wrk["stats"]["hd15iqr"] = None496    benchmark_wrk["stats"]["total"] = None497    benchmark_wrk["stats"]["iterations"] = 1498    # these tables are not currently being used499    # benchmark_latency = copy.deepcopy(benchmark_wrk)500    # benchmark_latency["group"] = benchmark_latency["group"] + " (wrk latencies per thread)"501    # # convert latencies from us to s502    # benchmark_latency["stats"]["min"] = raw_stats["latencies"]["min"] / 1E6503    # benchmark_latency["stats"]["max"] = raw_stats["latencies"]["max"] / 1E6504    # benchmark_latency["stats"]["mean"] = raw_stats["latencies"]["mean"] / 1E6505    # benchmark_latency["stats"]["stddev"] = raw_stats["latencies"]["stdev"] / 1E6506    # benchmark_latency["stats"]["ops"] = 1 / (raw_stats["latencies"]["mean"] / 1E6)507    # benchmark_request = copy.deepcopy(benchmark_wrk)508    # benchmark_request["group"] = benchmark_request["group"] + " (wrk requests/time per thread)"509    # benchmark_request["stats"]["min"] = raw_stats["requests"]["min"]510    # benchmark_request["stats"]["max"] = raw_stats["requests"]["max"]511    # benchmark_request["stats"]["mean"] = raw_stats["requests"]["mean"]512    # benchmark_request["stats"]["stddev"] = raw_stats["requests"]["stdev"]513    # benchmark_request["stats"]["ops"] = raw_stats["requests"]["mean"]514    benchmark_summary = copy.deepcopy(benchmark_wrk)515    benchmark_summary["stats"]["min"] = raw_stats["latencies"]["min"] / 1e6516    benchmark_summary["stats"]["max"] = raw_stats["latencies"]["max"] / 1e6517    benchmark_summary["stats"]["mean"] = raw_stats["latencies"]["mean"] / 1e6518    benchmark_summary["stats"]["stddev"] = raw_stats["latencies"]["stdev"] / 1e6519    benchmark_summary["stats"]["ops"] = raw_stats["summaries"]["requests"] / (520        raw_stats["summaries"]["duration"] / 1e6521    )522    # wrk's stddev and means are per thread which doesn't correlate directly to the mean523    # so calculating uncertainty with them is not accurate524    benchmark_summary["stats"]["ops_uncertainty"] = "N/A"525    return (526        # (benchmark_latency, "wrk latencies per thread"),527        # (benchmark_request, "wrk requests/time per thread"),528        benchmark_summary,529    )530def combine_wrk_stats(samples):531    stats = {532        "latencies": {},533        "requests": {},534        "summaries": {},535    }536    stats["latencies"]["min"] = min([x["latencies"]["min"] for x in samples])537    stats["latencies"]["max"] = max([x["latencies"]["max"] for x in samples])538    stats["latencies"]["mean"] = statistics.mean(539        [x["latencies"]["mean"] for x in samples]540    )541    # pooled std dev542    stats["latencies"]["stdev"] = math.sqrt(543        sum([math.pow(x["latencies"]["stdev"], 2) for x in samples])544    )545    stats["summaries"]["requests"] = sum([x["summaries"]["requests"] for x in samples])546    stats["summaries"]["duration"] = sum([x["summaries"]["duration"] for x in samples])547    return stats548def wrk_benchmarks(config: Config, benchmarks: Benchmarks):549    client = Client(config.http_address)550    addr = socket.gethostbyname(config.http_address.split(":")[0])551    if not ipaddress.ip_address(addr).is_loopback:552        assert client.server_live()553        server = None554    else:555        server = Server()556    wrk_options = config.wrk557    with Progress() as progress:558        task0 = progress.add_task(559            "Running wrk benchmarks...", total=len(benchmarks.get())560        )561        for benchmark in benchmarks.get():562            try:563                extra_info = benchmark["extra_info"]564            except KeyError:565                progress.update(task0, advance=1)566                continue567            if "lua" in extra_info:568                lua_file = (569                    os.getenv("PROTEUS_ROOT")570                    + f"/tests/workers/{extra_info['lua']}.lua"571                )572                if not os.path.exists(lua_file):573                    print(f"Lua file not found, skipping: {lua_file}")574                    continue575                if wrk_options.workers is not None:576                    loads = wrk_options.workers577                else:578                    loads = [1]579                if server is not None and config.start_server:580                    server.start(True)581                    client.wait_until_live()582                else:583                    assert client.server_live()584                repeat_wrk_count = config.repeat585                task1 = progress.add_task(586                    f"Running {benchmark['name']}", total=len(loads)587                )588                for load in loads:589                    model = extra_info["model"]590                    parameters = extra_info["parameters"]591                    if parameters is None:592                        parameters = {"share": False}593                    else:594                        parameters["share"] = False595                    for _ in range(load):596                        client.load(model, parameters)597                    while not client.model_ready(model):598                        pass599                    infer_endpoint = client.get_address("infer", model)600                    # print(f"Loading {load} copies of the {model} model")601                    total = (602                        len(wrk_options.threads)603                        * len(wrk_options.connections)604                        * len(wrk_options.time)605                    )606                    task2 = progress.add_task(607                        f"Running with {load} worker(s)...", total=total608                    )609                    for wrk_config in itertools.product(610                        wrk_options.threads, wrk_options.connections, wrk_options.time611                    ):612                        # TODO(varunsh): we need to check how many requests actually succeeded613                        wrk_command = [614                            "wrk",615                            f"-t{wrk_config[0]}",616                            f"-c{wrk_config[1]}",617                            f"-d{wrk_config[2]}",618                            "-s",619                            lua_file,620                            infer_endpoint,621                        ]622                        if config.verbosity > 0:623                            print(f"wrk command: \n  {' '.join(wrk_command)}")624                        wrk_stats = []625                        task3 = progress.add_task(626                            f"Running wrk: threads: {wrk_config[0]}, TCP connections: {wrk_config[1]}, time: {wrk_config[2]}",627                            total=repeat_wrk_count,628                        )629                        for _ in range(repeat_wrk_count):630                            try:631                                ret = subprocess.run(632                                    wrk_command,633                                    check=True,634                                    stdout=subprocess.PIPE,635                                    stderr=subprocess.PIPE,636                                )637                            except subprocess.CalledProcessError as ex:638                                print(ex.stdout)639                                print(ex.stderr)640                                sys.exit(ex.returncode)641                            wrk_output = ret.stdout.decode("utf-8")642                            wrk_stats.append(parse_wrk_output(wrk_output))643                            progress.update(task3, advance=1)644                        raw_stats = combine_wrk_stats(wrk_stats)645                        wrk_benchmarks = make_wrk_benchmarks(646                            raw_stats, benchmark, wrk_config, load647                        )648                        for wrk_benchmark in wrk_benchmarks:649                            benchmarks.add(wrk_benchmark)650                        progress.update(task2, advance=1)651                    for _ in range(load):652                        client.unload(model)653                    progress.update(task1, advance=1)654                if server is not None and config.start_server:655                    server.stop()656                    client.wait_until_stop()657            progress.update(task0, advance=1)658    return benchmarks659def parse_cpp_output(output: str) -> dict:660    """661    Parse the terminal output of cpp tests into a dictionary662    Args:663        output (str): Raw terminal output from the executable664    Returns:665        dict: Organized results from the executable666    """667    lines = output.split("\n")668    raw_stats = lines[-3:-1]669    queries = raw_stats[0].split(" ")[4]670    time = raw_stats[0].split(" ")[6]671    queries_per_sec = raw_stats[1].split(" ")[4]672    stats = {673        "queries": int(queries),674        "time": float(time) / 1000,  # convert to seconds675        "qps": float(queries_per_sec),676    }677    return stats678def combine_cpp_stats(samples):679    stats = {}680    stats["min"] = min([x["time"] / x["queries"] for x in samples])681    stats["max"] = max([x["time"] / x["queries"] for x in samples])682    stats["mean"] = statistics.mean([x["time"] / x["queries"] for x in samples])683    if len(samples) > 1:684        stats["stdev"] = statistics.stdev(685            [x["time"] / x["queries"] for x in samples], stats["mean"]686        )687    else:688        stats["stdev"] = 0689    stats["ops"] = statistics.mean([x["qps"] for x in samples])690    return stats691def make_cpp_benchmarks(raw_stats, path: pathlib.Path, cpp_config, repeat):692    cpp_setting = f"I{cpp_config[0]}:T{cpp_config[1]}"693    name = path.stem694    benchmark = copy.deepcopy(BASE_BENCHMARK)695    if name.startswith("test_"):696        benchmark["group"] = name[len("test_") :]697    else:698        benchmark["group"] = name699    benchmark["type"] = "native (cpp)"700    benchmark["config"] = cpp_setting701    benchmark["fullname"] = str(path)702    benchmark["load"] = cpp_config[2]703    benchmark["name"] = name704    if cpp_config[3]:705        benchmark["name"] += f" (reference)"706    benchmark["stats"]["rounds"] = 1707    benchmark["stats"]["median"] = None708    benchmark["stats"]["iqr"] = None709    benchmark["stats"]["q1"] = None710    benchmark["stats"]["q3"] = None711    benchmark["stats"]["iqr_outliers"] = None712    benchmark["stats"]["stddev_outliers"] = None713    benchmark["stats"]["outliers"] = None714    benchmark["stats"]["ld15iqr"] = None715    benchmark["stats"]["hd15iqr"] = None716    benchmark["stats"]["total"] = None717    benchmark["stats"]["iterations"] = repeat718    benchmark["stats"]["min"] = raw_stats["min"]719    benchmark["stats"]["max"] = raw_stats["max"]720    benchmark["stats"]["mean"] = raw_stats["mean"]721    benchmark["stats"]["stddev"] = raw_stats["stdev"]722    benchmark["stats"]["ops"] = raw_stats["ops"]723    mean = benchmark["stats"]["mean"]724    stddev = benchmark["stats"]["stddev"]725    ops = benchmark["stats"]["ops"]726    benchmark["stats"]["ops_uncertainty"] = ops - (1 / (mean + stddev))727    return benchmark728def get_benchmark_exe(path: pathlib.Path):729    relative_path_to_exe = (730        str(path.parent)[len(os.getenv("PROTEUS_ROOT")) :] + f"/{path.stem}"731    )732    benchmark_path = os.getenv("PROTEUS_ROOT") + f"/build/Release{relative_path_to_exe}"733    if not os.path.exists(benchmark_path):734        return None735    with open(path, "r") as f:736        for line in f:737            if "@brief Benchmark" in line:738                return benchmark_path739            if line.strip().startswith("#include"):740                return None741    return None742def cpp_benchmarks(config: Config, benchmarks: Benchmarks):743    benchmarks_to_run = set()744    benchmark_dir = os.getenv("PROTEUS_ROOT") + "/tests"745    accept_all_benchmarks = True if config.benchmarks is None else False746    for path in pathlib.Path(benchmark_dir).rglob("*.cpp"):747        if not accept_all_benchmarks:748            expression = Expression.compile(config.benchmarks)749            if not expression.evaluate(KeywordMatcher([path.stem])):750                continue751        benchmark = get_benchmark_exe(path)752        if benchmark is not None:753            benchmarks_to_run.add(benchmark)754    cpp_options = config.cpp755    repeat_count = config.repeat756    with Progress() as progress:757        task0 = progress.add_task(758            f"Running cpp benchmarks...", total=len(benchmarks_to_run)759        )760        for benchmark in benchmarks_to_run:761            valid_flags = [""]762            for flag in cpp_options.flags:763                if not flag:764                    continue765                cpp_command = [benchmark, "--help"]766                ret = subprocess.run(cpp_command, check=True, stdout=subprocess.PIPE)767                help_output = ret.stdout.decode("utf-8")768                if help_output.find(flag.split(" ")[0]) != -1:769                    valid_flags.append(flag)770            total = (771                len(cpp_options.images)772                * len(cpp_options.threads)773                * len(cpp_options.workers * len(valid_flags))774            )775            task1 = progress.add_task(f"Running {benchmark}", total=total)776            for cpp_config in itertools.product(777                cpp_options.images,778                cpp_options.threads,779                cpp_options.workers,780                valid_flags,781            ):782                cpp_command = [783                    benchmark,784                    "-i",785                    str(cpp_config[0]),786                    "-t",787                    str(cpp_config[1]),788                    "-r",789                    str(cpp_config[2]),790                    cpp_config[3],791                ]792                cpp_stats = []793                task2 = progress.add_task(794                    f"Configuration: Images: {cpp_config[0]}, Threads: {cpp_config[1]}, Workers: {cpp_config[2]}, Flags: {cpp_config[3]}",795                    total=repeat_count,796                )797                if config.verbosity > 0:798                    print(f"cpp command: \n  {' '.join(cpp_command)}")799                for _ in range(repeat_count):800                    ret = subprocess.run(801                        cpp_command, check=True, stdout=subprocess.PIPE802                    )803                    cpp_output = ret.stdout.decode("utf-8")804                    # print(cpp_output)805                    cpp_stats.append(parse_cpp_output(cpp_output))806                    progress.update(task2, advance=1)807                raw_stats = combine_cpp_stats(cpp_stats)808                cpp_benchmark = make_cpp_benchmarks(809                    raw_stats, pathlib.Path(benchmark), cpp_config, repeat_count810                )811                benchmarks.add(cpp_benchmark)812                progress.update(task1, advance=1)813            progress.update(task0, advance=1)814    return benchmarks815def pytest_benchmarks(config: Config, quiet=False):816    hostname, port = config.http_address.split(":")817    client = Client(config.http_address)818    addr = socket.gethostbyname(hostname)819    if not ipaddress.ip_address(addr).is_loopback or not config.start_server:820        try:821            assert client.server_live()822        except proteus.exceptions.ConnectionError:823            print(824                f"Cannot connect to HTTP server at {config.http_address}. Check the address or set it to start automatically"825            )826            sys.exit(1)827    cmd = (828        os.getenv("PROTEUS_ROOT")829        + f"/tests/test.sh --benchmark only --hostname {hostname} --http_port {port}"830    )831    if config.benchmarks:832        cmd += f' -k "{config.benchmarks}"'833    try:834        if not quiet:835            subprocess.run(836                ["/bin/bash", "-c", cmd],837                stdout=sys.stdout,838                stderr=subprocess.PIPE,839                check=True,840            )841        else:842            subprocess.run(843                ["/bin/bash", "-c", cmd],844                stdout=subprocess.DEVNULL,845                stderr=subprocess.DEVNULL,846                check=True,847            )848    except subprocess.CalledProcessError as ex:849        print("Running pytest tests failed. Exiting benchmarking")850        sys.exit(ex.returncode)851    benchmarks = get_benchmark(normalize=True)852    benchmarks.write()853if __name__ == "__main__":854    parser = argparse.ArgumentParser(description="Run Proteus benchmarking")855    parser.add_argument(856        "-k",857        action="store",858        default="",859        help="choose which benchmarks to run by substring",860    )861    parser.add_argument(862        "-print",863        action="store",864        default=None,865        type=int,866        help="print the last nth benchmark and exit",867    )868    parser.add_argument(869        "--force",870        action="store_true",871        default=False,872        help="don't prompt to verify configuration",873    )874    args = parser.parse_args()875    if args.print is not None:876        benchmarks = get_benchmark(index=int(args.print))877        benchmarks.print()878        sys.exit(0)879    config = Config(os.getenv("PROTEUS_ROOT") + "/tools/benchmark.yml")880    if args.k:881        config.benchmarks = args.k882    if not args.force:883        print(config)884        retval = input("Run benchmarking with this configuration? [Y/N] ")885        if retval.upper() != "Y":886            print("Exiting...")887            sys.exit(0)888    # if wrk tests are run, we need to also run pytest tests889    if config.pytest.enabled or config.wrk.enabled:890        pytest_benchmarks(config)891    else:892        # if pytest is disabled, run anyway with a small test suite to create a893        # new benchmark file that can be populated by later benchmarks894        tmp = config.benchmarks895        config.benchmarks = "echo"896        pytest_benchmarks(config, True)897        config.benchmarks = tmp898        benchmarks = get_benchmark(normalize=True)899        benchmarks.clear()900        benchmarks.write()901    # normalize the pytest generated benchmarks to be in line with the others902    benchmarks = get_benchmark(normalize=True)903    if config.wrk.enabled:904        benchmarks = wrk_benchmarks(config, benchmarks)905    if config.cpp.enabled:906        benchmarks = cpp_benchmarks(config, benchmarks)907    benchmarks.write()...

stats.py

Source:stats.py

...64    @cached_property65    def median(self):66        return statistics.median(self.data)67    @cached_property68    def ld15iqr(self):69        """70        Tukey-style Lowest Datum within 1.5 IQR under Q1.71        """72        if len(self.data) == 1:73            return self.data[0]74        else:75            return self.sorted_data[bisect_left(self.sorted_data, self.q1 - 1.5 * self.iqr)]76    @cached_property77    def hd15iqr(self):78        """79        Tukey-style Highest Datum within 1.5 IQR over Q3.80        """81        if len(self.data) == 1:82            return self.data[0]...

Automation Testing Tutorials

Learn to execute automation testing from scratch with LambdaTest Learning Hub. Right from setting up the prerequisites to run your first automation test, to following best practices and diving deeper into advanced test scenarios. LambdaTest Learning Hubs compile a list of step-by-step guides to help you be proficient with different test automation frameworks i.e. Selenium, Cypress, TestNG etc.