cobweb-launcher 1.3.3__tar.gz → 1.3.4__tar.gz

{cobweb-launcher-1.3.3 → cobweb-launcher-1.3.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: cobweb-launcher
-Version: 1.3.3
+Version: 1.3.4
 Summary: spider_hole
 Home-page: https://github.com/Juannie-PP/cobweb
 Author: Juannie-PP

cobweb-launcher-1.3.4/cobweb/__init__.py

@@ -0,0 +1,2 @@
+from .launchers import LauncherPro, LauncherApi
+from .constant import CrawlerModel

cobweb-launcher-1.3.4/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-launcher-1.3.4/cobweb/launchers/launcher_api.py

@@ -0,0 +1,88 @@
+import time
+
+from cobweb.base import TaskQueue, Decorators
+from cobweb.schedulers import ApiScheduler
+from .launcher import Launcher
+
+
+class LauncherPro(Launcher):
+
+    def __init__(self, task, project, custom_setting=None, **kwargs):
+        super().__init__(task, project, custom_setting, **kwargs)
+        self._redis_download = "{%s:%s}:download" % (project, task)
+        self._redis_todo = "{%s:%s}:todo" % (project, task)
+        self._scheduler = ApiScheduler(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.stop
+    def _heartbeat(self):
+        if self._scheduler.working.is_set():
+            self._scheduler.set_heartbeat()
+        time.sleep(3)
+
+    @Decorators.stop
+    def _reset(self):
+        self._scheduler.reset(
+            keys=[self._redis_todo, self._redis_download],
+            reset_time=self.seed_reset_seconds
+        )
+        time.sleep(30)
+
+    @Decorators.pause
+    def _insert(self):
+        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):
+        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-launcher-1.3.3 → cobweb-launcher-1.3.4}/cobweb/schedulers/__init__.py

@@ -1 +1,3 @@
 from .scheduler_redis import RedisScheduler
+from .scheduler_api import ApiScheduler
+

cobweb-launcher-1.3.4/cobweb/schedulers/scheduler_api.py

@@ -0,0 +1,69 @@
+import threading
+import time
+
+# from cobweb.base import Seed
+from cobweb.db import ApiDB
+
+
+class ApiScheduler:
+
+    def __init__(self, task, project, scheduler_wait_seconds=30):
+        self._todo_key = "{%s:%s}:todo" % (project, task)
+        self._download_key = "{%s:%s}:download" % (project, task)
+        self._heartbeat_key = "heartbeat:%s_%s" % (project, task)
+        self._speed_control_key = "speed_control:%s_%s" % (project, task)
+        self._reset_lock_key = "lock:reset:%s_%s" % (project, task)
+        self._db = ApiDB()
+
+        self.scheduler_wait_seconds = scheduler_wait_seconds
+        self.working = threading.Event()
+
+    @property
+    def heartbeat(self):
+        return self._db.exists(self._heartbeat_key)
+
+    def set_heartbeat(self):
+        return self._db.setex(self._heartbeat_key, 5)
+
+    def schedule(self, key, count):
+        if not self._db.zcount(key, 0, "(1000"):
+            time.sleep(self.scheduler_wait_seconds)
+        else:
+            source = int(time.time())
+            members = self._db.members(key, source, count=count, _min=0, _max="(1000")
+            for member, priority in members:
+                # seed = Seed(member, priority=priority)
+                yield member.decode(), priority
+
+    def insert(self, key, items):
+        if items:
+            self._db.zadd(key, items, nx=True)
+
+    def reset(self, keys, reset_time=30):
+        if self._db.lock(self._reset_lock_key, t=120):
+
+            if isinstance(keys, str):
+                keys = [keys]
+
+            _min = reset_time - int(time.time()) if self.heartbeat else "-inf"
+
+            for key in keys:
+                self._db.members(key, 0, _min=_min, _max="(0")
+
+            if not self.heartbeat:
+                self.working.set()
+                time.sleep(10)
+
+            self._db.delete(self._reset_lock_key)
+
+    def refresh(self, key, items: dict[str, int]):
+        refresh_time = int(time.time())
+        its = {k: -refresh_time - v / 1000 for k, v in items}
+        self._db.zadd(key, item=its, xx=True)
+
+    def delete(self, key, values):
+        self._db.zrem(key, *values)
+
+
+
+

{cobweb-launcher-1.3.3 → cobweb-launcher-1.3.4}/cobweb_launcher.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: cobweb-launcher
-Version: 1.3.3
+Version: 1.3.4
 Summary: spider_hole
 Home-page: https://github.com/Juannie-PP/cobweb
 Author: Juannie-PP

{cobweb-launcher-1.3.3 → cobweb-launcher-1.3.4}/cobweb_launcher.egg-info/SOURCES.txt

@@ -30,6 +30,7 @@ cobweb/pipelines/pipeline.py
 cobweb/pipelines/pipeline_console.py
 cobweb/pipelines/pipeline_loghub.py
 cobweb/schedulers/__init__.py
+cobweb/schedulers/scheduler_api.py
 cobweb/schedulers/scheduler_redis.py
 cobweb/utils/__init__.py
 cobweb/utils/bloom.py

{cobweb-launcher-1.3.3 → cobweb-launcher-1.3.4}/setup.py

@@ -5,7 +5,7 @@ with open("README.md", "r", encoding="utf-8") as fh:
 
 setup(
     name="cobweb-launcher",
-    version="1.3.3",
+    version="1.3.4",
     packages=find_packages(),
     url="https://github.com/Juannie-PP/cobweb",
     license="MIT",

cobweb-launcher-1.3.3/cobweb/launchers/launcher_api.py

@@ -1,161 +0,0 @@
-import time
-import threading
-
-from cobweb.db import ApiDB
-from cobweb.base import Seed, TaskQueue,logger, stop, pause
-from cobweb.constant import DealModel
-from .launcher import Launcher
-
-
-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._heartbeat_start_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 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
-        return TaskQueue.TODO.pop()
-
-    @stop
-    def _reset(self):
-        """
-        检查过期种子，重新添加到redis缓存中
-        """
-        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(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)
-
-    @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 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,
-                _min=0, _max="(1000"
-            )
-            for member, priority in members:
-                seed = Seed(member, priority=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)
-
-    @pause
-    def _insert(self):
-        """
-        添加新种子到redis队列中
-        """
-        seeds = {}
-        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 TaskQueue.SEED.length < self.new_queue_max_size:
-            time.sleep(self.new_queue_wait_seconds)
-
-    @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": []}
-        status = TaskQueue.DONE.length < self.done_queue_max_size
-
-        for _ in range(self.done_queue_max_size):
-            seed = TaskQueue.DONE.pop()
-            if not seed:
-                break
-            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_info["count"]:
-
-            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)
-

cobweb-launcher-1.3.3/cobweb_/__init__.py

@@ -1,2 +0,0 @@
-from .launchers import LauncherAir, LauncherPro, LauncherApi
-from .constant import CrawlerModel

cobweb-launcher-1.3.3/cobweb_/launchers/launcher_air.py

@@ -1,88 +0,0 @@
-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)
-
-