cobweb-launcher 1.2.49__py3-none-any.whl → 1.3.2__py3-none-any.whl

cobweb/launchers/launcher_api.py

@@ -2,9 +2,9 @@ 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
+from cobweb.base import Seed, TaskQueue,logger, stop, pause
+from cobweb.constant import DealModel
+from .launcher import Launcher
 
 
 class LauncherApi(Launcher):
@@ -24,18 +24,14 @@ class LauncherApi(Launcher):
 
         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):
+        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)
@@ -46,30 +42,25 @@ class LauncherApi(Launcher):
         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)):
+        if TaskQueue.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)
+        return TaskQueue.TODO.pop()
 
-    @check_pause
+    @stop
     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 \
+            _min = -int(time.time()) + self.seed_reset_seconds \
                 if self.heartbeat else "-inf"
 
             self._db.members(self._todo_key, 0, _min=_min, _max="(0")
@@ -79,131 +70,92 @@ class LauncherApi(Launcher):
 
             self._db.delete(self._reset_lock_key)
 
-        time.sleep(reset_wait_seconds)
+        time.sleep(30)
+
+    @stop
+    def _refresh(self):
+        """
+        刷新doing种子过期时间，防止reset重新消费
+        """
+        if self.doing_seeds:
+            refresh_time = int(time.time())
+            seeds = {k: -refresh_time - v / 1e3 for k, v in self.doing_seeds.items()}
+            self._db.zadd(self._todo_key, item=seeds, xx=True)
+        time.sleep(3)
 
-    @check_pause
+    @stop
     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)
+            time.sleep(self.scheduler_wait_seconds)
+        elif TaskQueue.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,
+                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
+                TaskQueue.TODO.push(seed)
+                self.doing_seeds[seed.to_string] = seed.params.priority
+
+    @pause
+    def _heartbeat(self):
+        if self._heartbeat_start_event.is_set():
+            self._db.setex(self._heartbeat_key, t=5)
+        time.sleep(3)
 
-    @check_pause
+    @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:
+        for _ in range(self.new_queue_max_size):
+            if seed := TaskQueue.SEED.pop():
                 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)
+        if TaskQueue.SEED.length < self.new_queue_max_size:
+            time.sleep(self.new_queue_wait_seconds)
 
-    @check_pause
+    @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_info = {"count": 0, "failed": [], "succeed": [], "common": []}
+        status = TaskQueue.DONE.length < self.done_queue_max_size
 
-        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()
+        for _ in range(self.done_queue_max_size):
+            seed = TaskQueue.DONE.pop()
             if not seed:
                 break
-            seed_list.append(seed.to_string)
+            if seed.params.seed_status == DealModel.fail:
+                seed_info["failed"].append(seed.to_string)
+            elif self.done_model == 1:
+                seed_info["succeed"].append(seed.to_string)
+            else:
+                seed_info["common"].append(seed.to_string)
+            seed_info['count'] += 1
 
-        if seed_list:
+        if seed_info["count"]:
 
-            self._db.zrem(self._todo_key, *seed_list)
-            self._remove_doing_seeds(seed_list)
+            succeed_count = int(self._db.zrem(self._todo_key, *seed_info["common"]) or 0)
+            succeed_count += int(self._db.done([self._todo_key, self._done_key], *seed_info["succeed"]) or 0)
+            failed_count = int(self._db.done([self._todo_key, self._fail_key], *seed_info["failed"]) or 0)
+
+            if failed_count:
+                self.statistics(self._statistics_fail_key, failed_count)
+            if succeed_count:
+                self.statistics(self._statistics_done_key, succeed_count)
+
+            self._remove_doing_seeds(seed_info["common"] + seed_info["succeed"] + seed_info["failed"])
 
         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...")
+            time.sleep(self.done_queue_wait_seconds)
 

cobweb/launchers/launcher_pro.py

@@ -1,208 +1,90 @@
 import time
-import threading
 
-from cobweb.db import RedisDB
-from cobweb.base import Seed, logger
-from cobweb.utils import BloomFilter
-from cobweb.constant import DealModel, LogTemplate
-from .launcher import Launcher, check_pause
+from cobweb.base import TaskQueue, Decorators
+from cobweb.schedulers import RedisScheduler
+from .launcher import Launcher
 
 
 class LauncherPro(Launcher):
 
     def __init__(self, task, project, custom_setting=None, **kwargs):
         super().__init__(task, project, custom_setting, **kwargs)
-        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._db = RedisDB()
-        #
-        # 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._client.incrby(key, count)
-
-    def _get_seed(self) -> Seed:
-        spider_speed = self._db._client.get(self._speed_control_key)
-        if int(spider_speed or 0) > self._spider_max_count:
-            expire_time = self._db.ttl(self._speed_control_key)
-            if expire_time == -1:
-                self._db.delete(self._speed_control_key)
-            else:
-                logger.info(f"Too fast! Please wait {expire_time} seconds...")
-                time.sleep(expire_time / 2)
-            return None
-        seed = self.__LAUNCHER_QUEUE__["todo"].pop()
-        if seed and not self._db.lock(self._speed_control_key, t=self._time_window):
-            self._db._client.incrby(self._speed_control_key, 1)
-        return seed
-
-    @check_pause
-    def _execute_heartbeat(self):
-        if self._heartbeat_start_event.is_set():
-            self._db.setex(self._heartbeat_key, 5)
+        self._redis_download = "{%s:%s}:download" % (project, task)
+        self._redis_todo = "{%s:%s}:todo" % (project, task)
+        self._scheduler = RedisScheduler(task, project)
+
+    @Decorators.stop
+    def _schedule(self):
+        thread_sleep = self.scheduling_wait_time
+        for q, key, size, item_info in [
+            (TaskQueue.TODO, self._redis_todo, self.todo_queue_size, self._task_info["todo"]),
+            (TaskQueue.DOWNLOAD, self._redis_download, self.download_queue_size, self._task_info["download"]),
+        ]:
+            if q.length < size:
+                for member, priority in self._scheduler.schedule(
+                    key, self.scheduling_size
+                ):
+                    q.push((member, priority), direct_insertion=True)
+                    self.add_working_item(key.split(":")[-1], member, priority)
+                thread_sleep = 0.1
+        time.sleep(thread_sleep)
+
+    @Decorators.pause
+    def _heartbeat(self):
+        if self._scheduler.working.is_set():
+            self._scheduler.set_heartbeat()
         time.sleep(3)
 
-    @check_pause
+    @Decorators.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")
-            self._db.delete(self._reset_lock_key)
-
-            if not self.heartbeat:
-                self._heartbeat_start_event.set()
-
-        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
+        self._scheduler.reset(
+            keys=[self._redis_todo, self._redis_download],
+            reset_time=self.seed_reset_seconds
+        )
+        time.sleep(30)
 
-    @check_pause
+    @Decorators.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
+        thread_sleep = 0.1
+        for q, key, size in [
+            (TaskQueue.SEED, self._redis_todo, self.seed_queue_size),
+            (TaskQueue.REQUEST, self._redis_download, self.request_queue_size),
+        ]:
+            item_info = {}
+            while item := q.pop() and len(item_info.keys()) < self.inserting_size:
+                item_info[item.seed] = item.params.priority
+            if q.length >= size:
+                thread_sleep = self.inserting_wait_time
+            self._scheduler.insert(key, item_info)
+        time.sleep(thread_sleep)
+
+    @Decorators.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_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...")
-
+        self._scheduler.refresh(self._redis_todo, self._task_info["todo"])
+        self._scheduler.refresh(self._redis_download, self._task_info["download"])
+        time.sleep(10)
+
+    @Decorators.pause
+    def _remove(self):
+        thread_sleep = self.removing_wait_time
+        for q, key, size in [
+            (TaskQueue.DELETE, self._redis_todo, self.delete_queue_size),
+            (TaskQueue.DONE, self._redis_download, self.done_queue_size),
+        ]:
+            items = []
+            while item := q.pop() and len(items) < self.removing_size:
+                items.append(item)
+            self._scheduler.delete(key, *items)
+            self.remove_working_items(key.split(":")[-1], items)
+            if q.length >= size:
+                thread_sleep = 0.1
+        time.sleep(thread_sleep)
+
+    def _init_schedule_thread(self):
+        self._add_thread(func=self._heartbeat)
+        self._add_thread(func=self._reset)
+        self._add_thread(func=self._refresh)
+        self._add_thread(func=self._schedule)
+        self._add_thread(func=self._insert)
+        self._add_thread(func=self._remove)
+        self._add_thread(func=self._polling)

cobweb/pipelines/base_pipeline.py

@@ -0,0 +1,54 @@
+import time
+import threading
+
+from abc import ABC, abstractmethod
+from cobweb.base import BaseItem, Queue, logger
+
+
+class Pipeline(threading.Thread, ABC):
+
+    def __init__(
+            self,
+            done_queue: Queue,
+            upload_queue: Queue,
+            upload_queue_size: int,
+            upload_wait_seconds: int
+    ):
+        super().__init__()
+        self.done_queue = done_queue
+        self.upload_queue = upload_queue
+        self.upload_queue_size = upload_queue_size
+        self.upload_wait_seconds = upload_wait_seconds
+
+    @abstractmethod
+    def build(self, item: BaseItem) -> dict:
+        pass
+
+    @abstractmethod
+    def upload(self, table: str, data: list) -> bool:
+        pass
+
+    def run(self):
+        while True:
+            status = self.upload_queue.length < self.upload_queue_size
+            if status:
+                time.sleep(self.upload_wait_seconds)
+            data_info, seeds = {}, []
+            for _ in range(self.upload_queue_size):
+                item = self.upload_queue.pop()
+                if not item:
+                    break
+                data = self.build(item)
+                seeds.append(item.seed)
+                data_info.setdefault(item.table, []).append(data)
+            for table, datas in data_info.items():
+                try:
+                    self.upload(table, datas)
+                    status = True
+                except Exception as e:
+                    logger.info(e)
+                    status = False
+                if status:
+                    self.done_queue.push(seeds)
+
+

cobweb/pipelines/loghub_pipeline.py

@@ -0,0 +1,34 @@
+import json
+
+from cobweb import setting
+from cobweb.base import BaseItem
+from cobweb.pipelines import Pipeline
+from aliyun.log import LogClient, LogItem, PutLogsRequest
+
+
+class LoghubPipeline(Pipeline):
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.client = LogClient(**setting.LOGHUB_CONFIG)
+
+    def build(self, item: BaseItem):
+        log_item = LogItem()
+        temp = item.to_dict
+        for key, value in temp.items():
+            if not isinstance(value, str):
+                temp[key] = json.dumps(value, ensure_ascii=False)
+        contents = sorted(temp.items())
+        log_item.set_contents(contents)
+        return log_item
+
+    def upload(self, table, datas):
+        request = PutLogsRequest(
+            project=setting.LOGHUB_PROJECT,
+            logstore=table,
+            topic=setting.LOGHUB_TOPIC,
+            source=setting.LOGHUB_SOURCE,
+            logitems=datas,
+            compress=True
+        )
+        self.client.put_logs(request=request)