cobweb-launcher 0.1.8__py3-none-any.whl → 1.2.42__py3-none-any.whl

cobweb/launchers/launcher.py

@@ -0,0 +1,235 @@
+import time
+import inspect
+import threading
+import importlib
+from functools import wraps
+
+
+from cobweb import setting
+from cobweb.base import Seed, Queue, logger
+from cobweb.utils.tools import dynamic_load_class
+
+
+def check_pause(func):
+    @wraps(func)
+    def wrapper(self, *args, **kwargs):
+        while not self._pause.is_set():
+            try:
+                func(self, *args, **kwargs)
+            except Exception as e:
+                logger.info(f"{func.__name__}: " + str(e))
+            finally:
+                time.sleep(0.1)
+
+    return wrapper
+
+
+class Launcher(threading.Thread):
+
+    SEEDS = []
+
+    __DOING__ = {}
+
+    __CUSTOM_FUNC__ = {
+        # "download": None,
+        # "request": None,
+        # "parse": None,
+    }
+
+    __LAUNCHER_QUEUE__ = {
+        "new": Queue(),
+        "todo": Queue(),
+        "done": Queue(),
+        "upload": Queue()
+    }
+
+    __LAUNCHER_FUNC__ = [
+        "_reset",
+        "_scheduler",
+        "_insert",
+        "_refresh",
+        "_delete",
+    ]
+
+    def __init__(self, task, project, custom_setting=None, **kwargs):
+        super().__init__()
+        self.task = task
+        self.project = project
+
+        self._app_time = int(time.time())
+        self._stop = threading.Event()  # 结束事件
+        self._pause = threading.Event()  # 暂停事件
+
+        _setting = dict()
+
+        if custom_setting:
+            if isinstance(custom_setting, dict):
+                _setting = custom_setting
+            else:
+                if isinstance(custom_setting, str):
+                    custom_setting = importlib.import_module(custom_setting)
+                if not inspect.ismodule(custom_setting):
+                    raise Exception
+                for k, v in custom_setting.__dict__.items():
+                    if not k.startswith("__") and not inspect.ismodule(v):
+                        _setting[k] = v
+
+        _setting.update(**kwargs)
+
+        for k, v in _setting.items():
+            setattr(setting, k.upper(), v)
+
+        self._Crawler = dynamic_load_class(setting.CRAWLER)
+        self._Pipeline = dynamic_load_class(setting.PIPELINE)
+
+        self._before_scheduler_wait_seconds = setting.BEFORE_SCHEDULER_WAIT_SECONDS
+        self._scheduler_wait_seconds = setting.SCHEDULER_WAIT_SECONDS
+        self._todo_queue_full_wait_seconds = setting.TODO_QUEUE_FULL_WAIT_SECONDS
+        self._new_queue_wait_seconds = setting.NEW_QUEUE_WAIT_SECONDS
+        self._done_queue_wait_seconds = setting.DONE_QUEUE_WAIT_SECONDS
+        self._upload_queue_wait_seconds = setting.UPLOAD_QUEUE_WAIT_SECONDS
+        self._seed_reset_seconds = setting.SEED_RESET_SECONDS
+
+        self._todo_queue_size = setting.TODO_QUEUE_SIZE
+        self._new_queue_max_size = setting.NEW_QUEUE_MAX_SIZE
+        self._done_queue_max_size = setting.DONE_QUEUE_MAX_SIZE
+        self._upload_queue_max_size = setting.UPLOAD_QUEUE_MAX_SIZE
+
+        self._spider_max_retries = setting.SPIDER_MAX_RETRIES
+        self._spider_thread_num = setting.SPIDER_THREAD_NUM
+        self._spider_time_sleep = setting.SPIDER_TIME_SLEEP
+        self._spider_max_count = setting.SPIDER_MAX_COUNT
+        self._time_window = setting.TIME_WINDOW
+
+        self._done_model = setting.DONE_MODEL
+        self._task_model = setting.TASK_MODEL
+
+        self._filter_field = setting.FILTER_FIELD
+
+    @property
+    def request(self):
+        """
+        自定义request函数
+        use case:
+            from cobweb.base import Request, BaseItem
+            @launcher.request
+            def request(seed: Seed) -> Union[Request, BaseItem]:
+                ...
+                yield Request(seed.url, seed)
+        """
+        def decorator(func):
+            self.__CUSTOM_FUNC__["request"] = func
+        return decorator
+
+    @property
+    def download(self):
+        """
+        自定义download函数
+        use case:
+            from cobweb.base import Request, Response, Seed, BaseItem
+            @launcher.download
+            def download(item: Request) -> Union[Seed, BaseItem, Response, str]:
+                ...
+                yield Response(item.seed, response)
+        """
+        def decorator(func):
+            self.__CUSTOM_FUNC__["download"] = func
+        return decorator
+
+    @property
+    def parse(self):
+        """
+        自定义parse函数, xxxItem为自定义的存储数据类型
+        use case:
+            from cobweb.base import Request, Response
+            @launcher.parse
+            def parse(item: Response) -> BaseItem:
+               ...
+               yield xxxItem(seed, **kwargs)
+        """
+        def decorator(func):
+            self.__CUSTOM_FUNC__["parse"] = func
+        return decorator
+
+    def start_seeds(self):
+        seeds = [Seed(seed) for seed in self.SEEDS]
+        self.__LAUNCHER_QUEUE__['todo'].push(seeds)
+        return seeds
+
+    def _remove_doing_seeds(self, seeds):
+        for seed in seeds:
+            self.__DOING__.pop(seed, None)
+        # logger.info("remove %s seeds from __DOING__" % len(seeds))
+
+    def _get_seed(self) -> Seed:
+        return self.__LAUNCHER_QUEUE__["todo"].pop()
+
+    def _set_seed(self, seed, **kwargs):
+        self.__LAUNCHER_QUEUE__["todo"].push(seed, **kwargs)
+
+    def _upload_data(self, data, **kwargs):
+        self.__LAUNCHER_QUEUE__["upload"].push(data, **kwargs)
+
+    def _add_seed(self, seed, **kwargs):
+        self.__LAUNCHER_QUEUE__["new"].push(seed, **kwargs)
+
+    def _delete_seed(self, seed, **kwargs):
+        self.__LAUNCHER_QUEUE__["done"].push(seed, **kwargs)
+
+    def _execute(self):
+        for func_name in self.__LAUNCHER_FUNC__:
+            threading.Thread(name=func_name, target=getattr(self, func_name)).start()
+            time.sleep(1)
+
+    def run(self):
+        threading.Thread(target=self._execute_heartbeat).start()
+
+        self.start_seeds()
+
+        self._Crawler(
+            task=self.task, project=self.project,
+            stop=self._stop, pause=self._pause,
+            # launcher_queue=self.__LAUNCHER_QUEUE__,
+            get_seed=self._get_seed,
+            set_seed=self._set_seed,
+            add_seed=self._add_seed,
+            delete_seed=self._delete_seed,
+            upload_data=self._upload_data,
+            custom_func=self.__CUSTOM_FUNC__,
+            thread_num = self._spider_thread_num,
+            max_retries = self._spider_max_retries,
+            time_sleep=self._spider_time_sleep
+        ).start()
+
+        self._Pipeline(
+            stop=self._stop, pause=self._pause,
+            upload=self.__LAUNCHER_QUEUE__["upload"],
+            done=self.__LAUNCHER_QUEUE__["done"],
+            upload_size=self._upload_queue_max_size,
+            wait_seconds=self._upload_queue_wait_seconds
+        ).start()
+
+        self._execute()
+        self._polling()
+
+    def _execute_heartbeat(self):
+        pass
+
+    def _reset(self):
+        pass
+
+    def _scheduler(self):
+        pass
+
+    def _insert(self):
+        pass
+
+    def _refresh(self):
+        pass
+
+    def _delete(self):
+        pass
+
+    def _polling(self):
+        pass
+

cobweb/launchers/launcher_air.py

@@ -0,0 +1,88 @@
+import time
+
+from cobweb.base import logger
+from cobweb.constant import LogTemplate
+from .launcher import Launcher, check_pause
+
+
+class LauncherAir(Launcher):
+
+    # def _scheduler(self):
+    #     if self.start_seeds:
+    #         self.__LAUNCHER_QUEUE__['todo'].push(self.start_seeds)
+
+    @check_pause
+    def _insert(self):
+        seeds = {}
+        status = self.__LAUNCHER_QUEUE__['new'].length < self._new_queue_max_size
+        for _ in range(self._new_queue_max_size):
+            seed = self.__LAUNCHER_QUEUE__['new'].pop()
+            if not seed:
+                break
+            seeds[seed.to_string] = seed.params.priority
+        if seeds:
+            self.__LAUNCHER_QUEUE__['todo'].push(seeds)
+        if status:
+            time.sleep(self._new_queue_wait_seconds)
+
+    @check_pause
+    def _delete(self):
+        seeds = []
+        status = self.__LAUNCHER_QUEUE__['done'].length < self._done_queue_max_size
+
+        for _ in range(self._done_queue_max_size):
+            seed = self.__LAUNCHER_QUEUE__['done'].pop()
+            if not seed:
+                break
+            seeds.append(seed.to_string)
+
+        if seeds:
+            self._remove_doing_seeds(seeds)
+
+        if status:
+            time.sleep(self._done_queue_wait_seconds)
+
+    def _polling(self):
+
+        check_emtpy_times = 0
+
+        while not self._stop.is_set():
+
+            queue_not_empty_count = 0
+            pooling_wait_seconds = 30
+
+            for q in self.__LAUNCHER_QUEUE__.values():
+                if q.length != 0:
+                    queue_not_empty_count += 1
+
+            if queue_not_empty_count == 0:
+                pooling_wait_seconds = 3
+                if self._pause.is_set():
+                    check_emtpy_times = 0
+                    if not self._task_model:
+                        logger.info("Done! Ready to close thread...")
+                        self._stop.set()
+                elif check_emtpy_times > 2:
+                    self.__DOING__ = {}
+                    self._pause.set()
+                else:
+                    logger.info(
+                        "check whether the task is complete, "
+                        f"reset times {3 - check_emtpy_times}"
+                    )
+                    check_emtpy_times += 1
+            elif self._pause.is_set():
+                self._pause.clear()
+                self._execute()
+            else:
+                logger.info(LogTemplate.launcher_air_polling.format(
+                    task=self.task,
+                    doing_len=len(self.__DOING__.keys()),
+                    todo_len=self.__LAUNCHER_QUEUE__['todo'].length,
+                    done_len=self.__LAUNCHER_QUEUE__['done'].length,
+                    upload_len=self.__LAUNCHER_QUEUE__['upload'].length,
+                ))
+
+            time.sleep(pooling_wait_seconds)
+
+

cobweb/launchers/launcher_api.py

@@ -0,0 +1,209 @@
+import time
+import threading
+
+from cobweb.db import ApiDB
+from cobweb.base import Seed, logger
+from cobweb.constant import DealModel, LogTemplate
+from .launcher import Launcher, check_pause
+
+
+class LauncherApi(Launcher):
+
+    def __init__(self, task, project, custom_setting=None, **kwargs):
+        super().__init__(task, project, custom_setting, **kwargs)
+        self._db = ApiDB()
+
+        self._todo_key = "{%s:%s}:todo" % (project, task)
+        self._done_key = "{%s:%s}:done" % (project, task)
+        self._fail_key = "{%s:%s}:fail" % (project, task)
+        self._heartbeat_key = "heartbeat:%s_%s" % (project, task)
+
+        self._statistics_done_key = "statistics:%s:%s:done" % (project, task)
+        self._statistics_fail_key = "statistics:%s:%s:fail" % (project, task)
+        self._speed_control_key = "speed_control:%s_%s" % (project, task)
+
+        self._reset_lock_key = "lock:reset:%s_%s" % (project, task)
+
+        # self._bf_key = "bloom_%s_%s" % (project, task)
+        # self._bf = BloomFilter(self._bf_key)
+
+        self._heartbeat_start_event = threading.Event()
+        self._redis_queue_empty_event = threading.Event()
+
+    @property
+    def heartbeat(self):
+        return self._db.exists(self._heartbeat_key)
+
+    def statistics(self, key, count):
+        if not self._task_model and not self._db.exists(key):
+            self._db.setex(key, 86400 * 30, int(count))
+        else:
+            self._db.incrby(key, count)
+
+    def _get_seed(self) -> Seed:
+        """
+        从队列中获取种子（频控）
+        设置时间窗口为self._time_window（秒），判断在该窗口内的采集量是否满足阈值（self._spider_max_speed）
+        :return: True -> 种子, False -> None
+        """
+        if (self.__LAUNCHER_QUEUE__["todo"].length and
+                not self._db.auto_incr(self._speed_control_key, t=self._time_window, limit=self._spider_max_count)):
+            expire_time = self._db.ttl(self._speed_control_key)
+            logger.info(f"Too fast! Please wait {expire_time} seconds...")
+            time.sleep(expire_time / 2)
+            return None
+        seed = self.__LAUNCHER_QUEUE__["todo"].pop()
+        return seed
+
+    @check_pause
+    def _execute_heartbeat(self):
+        if self._heartbeat_start_event.is_set():
+            self._db.setex(self._heartbeat_key, 5)
+        time.sleep(3)
+
+    @check_pause
+    def _reset(self):
+        """
+        检查过期种子，重新添加到redis缓存中
+        """
+        reset_wait_seconds = 30
+        if self._db.lock(self._reset_lock_key, t=120):
+
+            _min = -int(time.time()) + self._seed_reset_seconds \
+                if self.heartbeat else "-inf"
+
+            self._db.members(self._todo_key, 0, _min=_min, _max="(0")
+
+            if not self.heartbeat:
+                self._heartbeat_start_event.set()
+
+            self._db.delete(self._reset_lock_key)
+
+        time.sleep(reset_wait_seconds)
+
+    @check_pause
+    def _scheduler(self):
+        """
+        调度任务，获取redis队列种子，同时添加到doing字典中
+        """
+        if not self._db.zcount(self._todo_key, 0, "(1000"):
+            time.sleep(self._scheduler_wait_seconds)
+        elif self.__LAUNCHER_QUEUE__['todo'].length >= self._todo_queue_size:
+            time.sleep(self._todo_queue_full_wait_seconds)
+        else:
+            members = self._db.members(
+                self._todo_key, int(time.time()),
+                count=self._todo_queue_size,
+                _min=0, _max="(1000"
+            )
+            for member, priority in members:
+                seed = Seed(member, priority=priority)
+                self.__LAUNCHER_QUEUE__['todo'].push(seed)
+                self.__DOING__[seed.to_string] = seed.params.priority
+
+    @check_pause
+    def _insert(self):
+        """
+        添加新种子到redis队列中
+        """
+        seeds = {}
+        status = self.__LAUNCHER_QUEUE__['new'].length < self._new_queue_max_size
+        for _ in range(self._new_queue_max_size):
+            seed = self.__LAUNCHER_QUEUE__['new'].pop()
+            if seed:
+                seeds[seed.to_string] = seed.params.priority
+        if seeds:
+            self._db.zadd(self._todo_key, seeds, nx=True)
+        if status:
+            time.sleep(self._new_queue_wait_seconds)
+
+    @check_pause
+    def _refresh(self):
+        """
+        刷新doing种子过期时间，防止reset重新消费
+        """
+        if self.__DOING__:
+            refresh_time = int(time.time())
+            seeds = {k:-refresh_time - v / 1000 for k, v in self.__DOING__.items()}
+            self._db.zadd(self._todo_key, item=seeds, xx=True)
+        time.sleep(15)
+
+    @check_pause
+    def _delete(self):
+        """
+        删除队列种子，根据状态添加至成功或失败队列，移除doing字典种子索引
+        """
+        # seed_info = {"count": 0, "failed": [], "succeed": [], "common": []}
+
+        seed_list = []
+        status = self.__LAUNCHER_QUEUE__['done'].length < self._done_queue_max_size
+
+        for _ in range(self._done_queue_max_size):
+            seed = self.__LAUNCHER_QUEUE__['done'].pop()
+            if not seed:
+                break
+            seed_list.append(seed.to_string)
+
+        if seed_list:
+
+            self._db.zrem(self._todo_key, *seed_list)
+            self._remove_doing_seeds(seed_list)
+
+        if status:
+            time.sleep(self._done_queue_wait_seconds)
+
+    def _polling(self):
+        wait_scheduler_execute = True
+        check_emtpy_times = 0
+        while not self._stop.is_set():
+            queue_not_empty_count = 0
+            pooling_wait_seconds = 30
+
+            for q in self.__LAUNCHER_QUEUE__.values():
+                if q.length != 0:
+                    queue_not_empty_count += 1
+                    wait_scheduler_execute = False
+
+            if queue_not_empty_count == 0:
+                pooling_wait_seconds = 3
+                if self._pause.is_set():
+                    check_emtpy_times = 0
+                    if not self._task_model and (
+                            not wait_scheduler_execute or
+                            int(time.time()) - self._app_time > self._before_scheduler_wait_seconds
+                    ):
+                        logger.info("Done! ready to close thread...")
+                        self._stop.set()
+
+                    elif self._db.zcount(self._todo_key, _min=0, _max="(1000"):
+                        logger.info(f"Recovery {self.task} task run！")
+                        self._pause.clear()
+                        self._execute()
+                    else:
+                        logger.info("pause! waiting for resume...")
+                elif check_emtpy_times > 2:
+                    self.__DOING__ = {}
+                    if not self._db.zcount(self._todo_key, _min="-inf", _max="(1000"):
+                        self._pause.set()
+                else:
+                    logger.info(
+                        "check whether the task is complete, "
+                        f"reset times {3 - check_emtpy_times}"
+                    )
+                    check_emtpy_times += 1
+            else:
+                logger.info(LogTemplate.launcher_pro_polling.format(
+                    task=self.task,
+                    doing_len=len(self.__DOING__.keys()),
+                    todo_len=self.__LAUNCHER_QUEUE__['todo'].length,
+                    done_len=self.__LAUNCHER_QUEUE__['done'].length,
+                    redis_seed_count=self._db.zcount(self._todo_key, "-inf", "+inf"),
+                    redis_todo_len=self._db.zcount(self._todo_key, 0, "(1000"),
+                    redis_doing_len=self._db.zcount(self._todo_key, "-inf", "(0"),
+                    upload_len=self.__LAUNCHER_QUEUE__['upload'].length,
+                ))
+
+            time.sleep(pooling_wait_seconds)
+
+        logger.info("Done! Ready to close thread...")
+