crawlo 1.4.5__py3-none-any.whl → 1.4.7__py3-none-any.whl

crawlo/core/engine.py

@@ -1,439 +1,452 @@
-#!/usr/bin/python
-# -*- coding:UTF-8 -*-
-import asyncio
-import time
-from inspect import iscoroutine
-from typing import Optional, Generator, Callable
-
-from crawlo import Request, Item
-from crawlo.core.processor import Processor
-from crawlo.core.scheduler import Scheduler
-from crawlo.downloader import DownloaderBase
-from crawlo.event import spider_opened, spider_error, request_scheduled
-from crawlo.exceptions import OutputError
-from crawlo.utils.misc import load_object
-from crawlo.spider import Spider
-from crawlo.task_manager import TaskManager
-from crawlo.utils.func_tools import transform
-from crawlo.utils.log import get_logger
-
-
-class Engine(object):
-
-    def __init__(self, crawler):
-        self.running = False
-        self.normal = True
-        self.crawler = crawler
-        self.settings = crawler.settings
-        self.spider: Optional[Spider] = None
-        self.downloader: Optional[DownloaderBase] = None
-        self.scheduler: Optional[Scheduler] = None
-        self.processor: Optional[Processor] = None
-        self.start_requests: Optional[Generator] = None
-        self.task_manager: Optional[TaskManager] = TaskManager(self.settings.get_int('CONCURRENCY'))
-
-        # 增强控制参数
-        self.max_queue_size = self.settings.get_int('SCHEDULER_MAX_QUEUE_SIZE', 200)
-        self.generation_batch_size = self.settings.get_int('REQUEST_GENERATION_BATCH_SIZE', 10)
-        self.generation_interval = self.settings.get_float('REQUEST_GENERATION_INTERVAL', 0.01)  # 优化默认值
-        self.backpressure_ratio = self.settings.get_float('BACKPRESSURE_RATIO', 0.9)  # 优化默认值
-        
-        # 状态跟踪
-        self._generation_paused = False
-        self._last_generation_time = 0
-        self._generation_stats = {
-            'total_generated': 0,
-            'backpressure_events': 0
-        }
-
-        self.logger = get_logger(name=self.__class__.__name__)
-
-    def _get_downloader_cls(self):
-        """获取下载器类，支持多种配置方法"""
-        # 方式1: 使用 DOWNLOADER_TYPE 简化名称（推荐）
-        downloader_type = self.settings.get('DOWNLOADER_TYPE')
-        if downloader_type:
-            try:
-                from crawlo.downloader import get_downloader_class
-                downloader_cls = get_downloader_class(downloader_type)
-                self.logger.debug(f"使用下载器类型: {downloader_type} -> {downloader_cls.__name__}")
-                return downloader_cls
-            except (ImportError, ValueError) as e:
-                self.logger.warning(f"无法使用下载器类型 '{downloader_type}': {e}，回退到默认配置")
-        
-        # 方式2: 使用 DOWNLOADER 完整类路径（兼容旧版本）
-        downloader_cls = load_object(self.settings.get('DOWNLOADER'))
-        if not issubclass(downloader_cls, DownloaderBase):
-            raise TypeError(f'下载器 {downloader_cls.__name__} 不是 DownloaderBase 的子类。')
-        return downloader_cls
-
-    def engine_start(self):
-        self.running = True
-        # 获取版本号，如果获取失败则使用默认值
-        version = self.settings.get('VERSION', '1.0.0')
-        if not version or version == 'None':
-            version = '1.0.0'
-        # 将INFO级别日志改为DEBUG级别，避免与CrawlerProcess启动日志重复
-        self.logger.debug(f"Crawlo框架已启动 {version}")
-
-    async def start_spider(self, spider):
-        self.spider = spider
-
-        self.scheduler = Scheduler.create_instance(self.crawler)
-        if hasattr(self.scheduler, 'open'):
-            if asyncio.iscoroutinefunction(self.scheduler.open):
-                await self.scheduler.open()
-            else:
-                self.scheduler.open()
-
-        downloader_cls = self._get_downloader_cls()
-        self.downloader = downloader_cls(self.crawler)
-        if hasattr(self.downloader, 'open'):
-            if asyncio.iscoroutinefunction(self.downloader.open):
-                self.downloader.open()
-            else:
-                # DownloaderBase.open() 是同步方法，直接调用而不是await
-                self.downloader.open()
-
-        self.processor = Processor(self.crawler)
-        if hasattr(self.processor, 'open'):
-            if asyncio.iscoroutinefunction(self.processor.open):
-                await self.processor.open()
-            else:
-                # Processor.open() 是同步方法
-                self.processor.open()
-
-        # 在处理器初始化之后初始化扩展管理器，确保日志输出顺序正确
-        # 中间件 -> 管道 -> 扩展
-        if not hasattr(self.crawler, 'extension') or not self.crawler.extension:
-            self.crawler.extension = self.crawler._create_extension()
-
-        # 启动引擎
-        self.engine_start()
-        
-        self.logger.debug("开始创建start_requests迭代器")
-        try:
-            # 先收集所有请求到列表中，避免在检查时消耗迭代器
-            requests_list = list(spider.start_requests())
-            self.logger.debug(f"收集到 {len(requests_list)} 个请求")
-            self.start_requests = iter(requests_list)
-            self.logger.debug("start_requests迭代器创建成功")
-        except Exception as e:
-            self.logger.error(f"创建start_requests迭代器失败: {e}")
-            import traceback
-            self.logger.error(traceback.format_exc())
-        await self._open_spider()
-
-    async def crawl(self):
-        """
-        支持智能请求生成和背压控制
-        """
-        generation_task = None
-        
-        try:
-            # 启动请求生成任务（如果启用了受控生成）
-            if (self.start_requests and 
-                self.settings.get_bool('ENABLE_CONTROLLED_REQUEST_GENERATION', False)):
-                self.logger.debug("创建受控请求生成任务")
-                generation_task = asyncio.create_task(
-                    self._controlled_request_generation()
-                )
-            else:
-                # 传统方式处理启动请求
-                self.logger.debug("创建传统请求生成任务")
-                generation_task = asyncio.create_task(
-                    self._traditional_request_generation()
-                )
-            
-            self.logger.debug("请求生成任务创建完成")
-            
-            # 主爬取循环
-            loop_count = 0
-            last_exit_check = 0  # 记录上次检查退出条件的时间
-            exit_check_interval = 1  # 每1次循环检查一次退出条件，进一步提高检查频率
-            
-            while self.running:
-                loop_count += 1
-                # 获取并处理请求
-                if request := await self._get_next_request():
-                    await self._crawl(request)
-                
-                # 优化退出条件检查频率
-                if loop_count - last_exit_check >= exit_check_interval:
-                    should_exit = await self._should_exit()
-                    if should_exit:
-                        self.logger.debug("满足退出条件，准备退出循环")
-                        break
-                    last_exit_check = loop_count
-                
-                # 短暂休息避免忙等，但减少休息时间以提高效率
-                await asyncio.sleep(0.000001)  # 从0.00001减少到0.000001
-            
-            self.logger.debug(f"主爬取循环结束，总共执行了 {loop_count} 次")
-        
-        finally:
-            # 确保请求生成任务完成
-            if generation_task and not generation_task.done():
-                try:
-                    await generation_task
-                except asyncio.CancelledError:
-                    pass
-            
-            await self.close_spider()
-
-    async def _traditional_request_generation(self):
-        """传统请求生成方法（兼容旧版本）"""
-        self.logger.debug("开始处理传统请求生成")
-        processed_count = 0
-        while self.running:
-            try:
-                start_request = next(self.start_requests)
-                self.logger.debug(f"获取到请求: {getattr(start_request, 'url', 'Unknown URL')}")
-                # 请求入队
-                await self.enqueue_request(start_request)
-                processed_count += 1
-                self.logger.debug(f"已处理请求数量: {processed_count}")
-            except StopIteration:
-                self.logger.debug("所有起始请求处理完成")
-                self.start_requests = None
-                break
-            except Exception as exp:
-                self.logger.error(f"处理请求时发生异常: {exp}")
-                import traceback
-                self.logger.error(traceback.format_exc())
-                # 1. All requests have been processed
-                # 2. Is scheduler idle
-                # 3. Is downloader idle
-                if not await self._exit():
-                    continue
-                self.running = False
-                if self.start_requests is not None:
-                    self.logger.error(f"Error occurred while starting request: {str(exp)}")
-            # 减少等待时间以提高效率
-            await asyncio.sleep(0.00001)  # 从0.0001减少到0.00001
-        self.logger.debug(f"传统请求生成完成，总共处理了 {processed_count} 个请求")
-
-    async def _controlled_request_generation(self):
-        """Controlled request generation (enhanced features)"""
-        self.logger.debug("Starting controlled request generation")
-        
-        batch = []
-        total_generated = 0
-        
-        try:
-            for request in self.start_requests:
-                batch.append(request)
-                
-                # 批量处理
-                if len(batch) >= self.generation_batch_size:
-                    generated = await self._process_generation_batch(batch)
-                    total_generated += generated
-                    batch = []
-                
-                # 背压检查
-                if await self._should_pause_generation():
-                    await self._wait_for_capacity()
-            
-            # 处理剩余请求
-            if batch:
-                generated = await self._process_generation_batch(batch)
-                total_generated += generated
-        
-        except Exception as e:
-            self.logger.error(f"Request generation failed: {e}")
-        
-        finally:
-            self.start_requests = None
-            self.logger.debug(f"Request generation completed, total: {total_generated}")
-
-    async def _process_generation_batch(self, batch) -> int:
-        """Process a batch of requests"""
-        generated = 0
-        
-        for request in batch:
-            if not self.running:
-                break
-            
-            # 等待队列有空间
-            while await self._is_queue_full() and self.running:
-                await asyncio.sleep(0.01)  # 减少等待时间
-            
-            if self.running:
-                await self.enqueue_request(request)
-                generated += 1
-                self._generation_stats['total_generated'] += 1
-            
-            # 控制生成速度，但使用更小的间隔
-            if self.generation_interval > 0:
-                await asyncio.sleep(self.generation_interval)
-        
-        return generated
-
-    async def _should_pause_generation(self) -> bool:
-        """Determine whether generation should be paused"""
-        # 检查队列大小
-        if await self._is_queue_full():
-            return True
-        
-        # 检查任务管理器负载
-        if self.task_manager:
-            current_tasks = len(self.task_manager.current_task)
-            if hasattr(self.task_manager, 'semaphore'):
-                max_concurrency = getattr(self.task_manager.semaphore, '_initial_value', 8)
-                if current_tasks >= max_concurrency * self.backpressure_ratio:
-                    return True
-        
-        return False
-
-    async def _is_queue_full(self) -> bool:
-        """Check if queue is full"""
-        if not self.scheduler:
-            return False
-        
-        queue_size = len(self.scheduler)
-        return queue_size >= self.max_queue_size * self.backpressure_ratio
-
-    async def _wait_for_capacity(self):
-        """Wait for system to have sufficient capacity"""
-        self._generation_stats['backpressure_events'] += 1
-        self.logger.debug("Backpressure triggered, pausing request generation")
-        
-        wait_time = 0.01  # 减少初始等待时间
-        max_wait = 1.0  # 减少最大等待时间
-        
-        while await self._should_pause_generation() and self.running:
-            await asyncio.sleep(wait_time)
-            wait_time = min(wait_time * 1.1, max_wait)
-
-    async def _open_spider(self):
-        asyncio.create_task(self.crawler.subscriber.notify(spider_opened))
-        # 直接调用crawl方法而不是创建任务，确保等待完成
-        await self.crawl()
-
-    async def _crawl(self, request):
-        async def crawl_task():
-            start_time = time.time()
-            try:
-                outputs = await self._fetch(request)
-                # 记录响应时间
-                response_time = time.time() - start_time
-                if self.task_manager:
-                    self.task_manager.record_response_time(response_time)
-                
-                # TODO 处理output
-                if outputs:
-                    await self._handle_spider_output(outputs)
-            except Exception as e:
-                # 记录详细的异常信息
-                self.logger.error(
-                    f"处理请求失败: {getattr(request, 'url', 'Unknown URL')} - {type(e).__name__}: {e}"
-                )
-                self.logger.debug(f"详细异常信息", exc_info=True)
-                
-                # 发送统计事件
-                if hasattr(self.crawler, 'stats'):
-                    self.crawler.stats.inc_value('downloader/exception_count')
-                    self.crawler.stats.inc_value(f'downloader/exception_type_count/{type(e).__name__}')
-                    if hasattr(request, 'url'):
-                        self.crawler.stats.inc_value(f'downloader/failed_urls_count')
-                
-                # 不再重新抛出异常，避免未处理的Task异常
-                return None
-
-        # 使用异步任务创建，遵守并发限制
-        await self.task_manager.create_task(crawl_task())
-
-    async def _fetch(self, request):
-        async def _successful(_response):
-            callback: Callable = request.callback or self.spider.parse
-            if _outputs := callback(_response):
-                if iscoroutine(_outputs):
-                    await _outputs
-                else:
-                    return transform(_outputs, _response)
-
-        _response = await self.downloader.fetch(request)
-        if _response is None:
-            return None
-        output = await _successful(_response)
-        return output
-
-    async def enqueue_request(self, start_request):
-        await self._schedule_request(start_request)
-
-    async def _schedule_request(self, request):
-        # TODO 去重
-        if await self.scheduler.enqueue_request(request):
-            asyncio.create_task(self.crawler.subscriber.notify(request_scheduled, request, self.crawler.spider))
-
-    async def _get_next_request(self):
-        return await self.scheduler.next_request()
-
-    async def _handle_spider_output(self, outputs):
-        async for spider_output in outputs:
-            if isinstance(spider_output, (Request, Item)):
-                await self.processor.enqueue(spider_output)
-            elif isinstance(spider_output, Exception):
-                asyncio.create_task(
-                    self.crawler.subscriber.notify(spider_error, spider_output, self.spider)
-                )
-                raise spider_output
-            else:
-                raise OutputError(f'{type(self.spider)} must return `Request` or `Item`.')
-
-    async def _exit(self):
-        if self.scheduler.idle() and self.downloader.idle() and self.task_manager.all_done() and self.processor.idle():
-            return True
-        return False
-
-    async def _should_exit(self) -> bool:
-        """检查是否应该退出"""
-        self.logger.debug(f"检查退出条件: start_requests={self.start_requests is not None}")
-        # 没有启动请求，且所有队列都空闲
-        if self.start_requests is None:
-            self.logger.debug("start_requests 为 None，检查其他组件状态")
-            # 使用异步的idle检查方法以获得更精确的结果
-            scheduler_idle = await self.scheduler.async_idle() if hasattr(self.scheduler, 'async_idle') else self.scheduler.idle()
-            downloader_idle = self.downloader.idle()
-            task_manager_done = self.task_manager.all_done()
-            processor_idle = self.processor.idle()
-            
-            self.logger.debug(f"组件状态 - Scheduler: {scheduler_idle}, Downloader: {downloader_idle}, TaskManager: {task_manager_done}, Processor: {processor_idle}")
-            
-            if (scheduler_idle and 
-                downloader_idle and 
-                task_manager_done and 
-                processor_idle):
-                # 立即进行二次检查，不等待
-                scheduler_idle = await self.scheduler.async_idle() if hasattr(self.scheduler, 'async_idle') else self.scheduler.idle()
-                downloader_idle = self.downloader.idle()
-                task_manager_done = self.task_manager.all_done()
-                processor_idle = self.processor.idle()
-                
-                self.logger.debug(f"二次检查组件状态 - Scheduler: {scheduler_idle}, Downloader: {downloader_idle}, TaskManager: {task_manager_done}, Processor: {processor_idle}")
-                
-                if (scheduler_idle and 
-                    downloader_idle and 
-                    task_manager_done and 
-                    processor_idle):
-                    self.logger.info("所有组件都空闲，准备退出")
-                    return True
-        else:
-            self.logger.debug("start_requests 不为 None，不退出")
-        
-        return False
-
-    async def close_spider(self):
-        # 不再调用crawler.close()，避免重复清理
-        # 清理工作应该由crawler的_lifecycle_manager上下文管理器来处理
-        await asyncio.gather(*self.task_manager.current_task)
-        await self.scheduler.close()
-        await self.downloader.close()
-    
-    def get_generation_stats(self) -> dict:
-        """获取生成统计"""
-        return {
-            **self._generation_stats,
-            'queue_size': len(self.scheduler) if self.scheduler else 0,
-            'active_tasks': len(self.task_manager.current_task) if self.task_manager else 0
+#!/usr/bin/python
+# -*- coding:UTF-8 -*-
+import asyncio
+import time
+from inspect import iscoroutine
+from typing import Optional, Generator, Callable
+
+from crawlo import Request, Item
+from crawlo.core.processor import Processor
+from crawlo.core.scheduler import Scheduler
+from crawlo.downloader import DownloaderBase
+from crawlo.event import CrawlerEvent
+from crawlo.exceptions import OutputError
+from crawlo.utils.misc import load_object
+from crawlo.spider import Spider
+from crawlo.task_manager import TaskManager
+from crawlo.utils.func_tools import transform
+from crawlo.logging import get_logger
+
+
+class Engine(object):
+
+    def __init__(self, crawler):
+        self.running = False
+        self.normal = True
+        self.crawler = crawler
+        self.settings = crawler.settings
+        self.spider: Optional[Spider] = None
+        self.downloader: Optional[DownloaderBase] = None
+        self.scheduler: Optional[Scheduler] = None
+        self.processor: Optional[Processor] = None
+        self.start_requests: Optional[Generator] = None
+        self.task_manager: Optional[TaskManager] = TaskManager(self.settings.get_int('CONCURRENCY'))
+
+        # 增强控制参数
+        self.max_queue_size = self.settings.get_int('SCHEDULER_MAX_QUEUE_SIZE', 200)
+        self.generation_batch_size = self.settings.get_int('REQUEST_GENERATION_BATCH_SIZE', 10)
+        self.generation_interval = self.settings.get_float('REQUEST_GENERATION_INTERVAL', 0.01)  # 优化默认值
+        self.backpressure_ratio = self.settings.get_float('BACKPRESSURE_RATIO', 0.9)  # 优化默认值
+        
+        # 状态跟踪
+        self._generation_paused = False
+        self._last_generation_time = 0
+        self._generation_stats = {
+            'total_generated': 0,
+            'backpressure_events': 0
+        }
+
+        self.logger = get_logger(name=self.__class__.__name__)
+
+    def _get_downloader_cls(self):
+        """获取下载器类，支持多种配置方法"""
+        # 方式1: 使用 DOWNLOADER_TYPE 简化名称（推荐）
+        downloader_type = self.settings.get('DOWNLOADER_TYPE')
+        if downloader_type:
+            try:
+                from crawlo.downloader import get_downloader_class
+                downloader_cls = get_downloader_class(downloader_type)
+                self.logger.debug(f"使用下载器类型: {downloader_type} -> {downloader_cls.__name__}")
+                return downloader_cls
+            except (ImportError, ValueError) as e:
+                self.logger.warning(f"无法使用下载器类型 '{downloader_type}': {e}，回退到默认配置")
+        
+        # 方式2: 使用 DOWNLOADER 完整类路径（兼容旧版本）
+        downloader_cls = load_object(self.settings.get('DOWNLOADER'))
+        if not issubclass(downloader_cls, DownloaderBase):
+            raise TypeError(f'下载器 {downloader_cls.__name__} 不是 DownloaderBase 的子类。')
+        return downloader_cls
+
+    def engine_start(self):
+        self.running = True
+        # 获取版本号，如果获取失败则使用默认值
+        version = self.settings.get('VERSION', '1.0.0')
+        if not version or version == 'None':
+            version = '1.0.0'
+        # 将INFO级别日志改为DEBUG级别，避免与CrawlerProcess启动日志重复
+        self.logger.debug(f"Crawlo框架已启动 {version}")
+
+    async def start_spider(self, spider):
+        self.spider = spider
+
+        self.scheduler = Scheduler.create_instance(self.crawler)
+        if hasattr(self.scheduler, 'open'):
+            if asyncio.iscoroutinefunction(self.scheduler.open):
+                await self.scheduler.open()
+            else:
+                self.scheduler.open()
+
+        downloader_cls = self._get_downloader_cls()
+        self.downloader = downloader_cls(self.crawler)
+        if hasattr(self.downloader, 'open'):
+            if asyncio.iscoroutinefunction(self.downloader.open):
+                self.downloader.open()
+            else:
+                # DownloaderBase.open() 是同步方法，直接调用而不是await
+                self.downloader.open()
+        
+        # 注册下载器到资源管理器
+        if hasattr(self.crawler, '_resource_manager') and self.downloader:
+            from crawlo.utils.resource_manager import ResourceType
+            self.crawler._resource_manager.register(
+                self.downloader,
+                lambda d: d.close() if hasattr(d, 'close') else None,
+                ResourceType.DOWNLOADER,
+                name=f"downloader.{downloader_cls.__name__}"
+            )
+            self.logger.debug(f"Downloader registered to resource manager: {downloader_cls.__name__}")
+
+        self.processor = Processor(self.crawler)
+        if hasattr(self.processor, 'open'):
+            if asyncio.iscoroutinefunction(self.processor.open):
+                await self.processor.open()
+            else:
+                # Processor.open() 是同步方法
+                self.processor.open()
+
+        # 在处理器初始化之后初始化扩展管理器，确保日志输出顺序正确
+        # 中间件 -> 管道 -> 扩展
+        if not hasattr(self.crawler, 'extension') or not self.crawler.extension:
+            self.crawler.extension = self.crawler._create_extension()
+
+        # 启动引擎
+        self.engine_start()
+        
+        self.logger.debug("开始创建start_requests迭代器")
+        try:
+            # 先收集所有请求到列表中，避免在检查时消耗迭代器
+            requests_list = list(spider.start_requests())
+            self.logger.debug(f"收集到 {len(requests_list)} 个请求")
+            self.start_requests = iter(requests_list)
+            self.logger.debug("start_requests迭代器创建成功")
+        except Exception as e:
+            self.logger.error(f"创建start_requests迭代器失败: {e}")
+            import traceback
+            self.logger.error(traceback.format_exc())
+        await self._open_spider()
+
+    async def crawl(self):
+        """
+        支持智能请求生成和背压控制
+        """
+        generation_task = None
+        
+        try:
+            # 启动请求生成任务（如果启用了受控生成）
+            if (self.start_requests and 
+                self.settings.get_bool('ENABLE_CONTROLLED_REQUEST_GENERATION', False)):
+                self.logger.debug("创建受控请求生成任务")
+                generation_task = asyncio.create_task(
+                    self._controlled_request_generation()
+                )
+            else:
+                # 传统方式处理启动请求
+                self.logger.debug("创建传统请求生成任务")
+                generation_task = asyncio.create_task(
+                    self._traditional_request_generation()
+                )
+            
+            self.logger.debug("请求生成任务创建完成")
+            
+            # 主爬取循环
+            loop_count = 0
+            last_exit_check = 0  # 记录上次检查退出条件的时间
+            exit_check_interval = 1  # 每1次循环检查一次退出条件，进一步提高检查频率
+            
+            while self.running:
+                loop_count += 1
+                # 获取并处理请求
+                if request := await self._get_next_request():
+                    await self._crawl(request)
+                
+                # 优化退出条件检查频率
+                if loop_count - last_exit_check >= exit_check_interval:
+                    should_exit = await self._should_exit()
+                    if should_exit:
+                        self.logger.debug("满足退出条件，准备退出循环")
+                        break
+                    last_exit_check = loop_count
+                
+                # 短暂休息避免忙等，但减少休息时间以提高效率
+                await asyncio.sleep(0.000001)  # 从0.00001减少到0.000001
+            
+            self.logger.debug(f"主爬取循环结束，总共执行了 {loop_count} 次")
+        
+        finally:
+            # 确保请求生成任务完成
+            if generation_task and not generation_task.done():
+                try:
+                    await generation_task
+                except asyncio.CancelledError:
+                    pass
+            
+            await self.close_spider()
+
+    async def _traditional_request_generation(self):
+        """传统请求生成方法（兼容旧版本）"""
+        self.logger.debug("开始处理传统请求生成")
+        processed_count = 0
+        while self.running:
+            try:
+                start_request = next(self.start_requests)
+                # 将过于频繁的debug日志合并，减少输出
+                # self.logger.debug(f"获取到请求: {getattr(start_request, 'url', 'Unknown URL')}")
+                # 请求入队
+                await self.enqueue_request(start_request)
+                processed_count += 1
+                # 减少过于频繁的日志输出
+                # self.logger.debug(f"已处理请求数量: {processed_count}")
+            except StopIteration:
+                self.logger.debug("所有起始请求处理完成")
+                self.start_requests = None
+                break
+            except Exception as exp:
+                self.logger.error(f"处理请求时发生异常: {exp}")
+                import traceback
+                self.logger.error(traceback.format_exc())
+                # 1. All requests have been processed
+                # 2. Is scheduler idle
+                # 3. Is downloader idle
+                if not await self._exit():
+                    continue
+                self.running = False
+                if self.start_requests is not None:
+                    self.logger.error(f"Error occurred while starting request: {str(exp)}")
+            # 减少等待时间以提高效率
+            await asyncio.sleep(0.00001)  # 从0.0001减少到0.00001
+        self.logger.debug(f"传统请求生成完成，总共处理了 {processed_count} 个请求")
+
+    async def _controlled_request_generation(self):
+        """Controlled request generation (enhanced features)"""
+        self.logger.debug("Starting controlled request generation")
+        
+        batch = []
+        total_generated = 0
+        
+        try:
+            for request in self.start_requests:
+                batch.append(request)
+                
+                # 批量处理
+                if len(batch) >= self.generation_batch_size:
+                    generated = await self._process_generation_batch(batch)
+                    total_generated += generated
+                    batch = []
+                
+                # 背压检查
+                if await self._should_pause_generation():
+                    await self._wait_for_capacity()
+            
+            # 处理剩余请求
+            if batch:
+                generated = await self._process_generation_batch(batch)
+                total_generated += generated
+        
+        except Exception as e:
+            self.logger.error(f"Request generation failed: {e}")
+        
+        finally:
+            self.start_requests = None
+            self.logger.debug(f"Request generation completed, total: {total_generated}")
+
+    async def _process_generation_batch(self, batch) -> int:
+        """Process a batch of requests"""
+        generated = 0
+        
+        for request in batch:
+            if not self.running:
+                break
+            
+            # 等待队列有空间
+            while await self._is_queue_full() and self.running:
+                await asyncio.sleep(0.01)  # 减少等待时间
+            
+            if self.running:
+                await self.enqueue_request(request)
+                generated += 1
+                self._generation_stats['total_generated'] += 1
+            
+            # 控制生成速度，但使用更小的间隔
+            if self.generation_interval > 0:
+                await asyncio.sleep(self.generation_interval)
+        
+        return generated
+
+    async def _should_pause_generation(self) -> bool:
+        """Determine whether generation should be paused"""
+        # 检查队列大小
+        if await self._is_queue_full():
+            return True
+        
+        # 检查任务管理器负载
+        if self.task_manager:
+            current_tasks = len(self.task_manager.current_task)
+            if hasattr(self.task_manager, 'semaphore'):
+                max_concurrency = getattr(self.task_manager.semaphore, '_initial_value', 8)
+                if current_tasks >= max_concurrency * self.backpressure_ratio:
+                    return True
+        
+        return False
+
+    async def _is_queue_full(self) -> bool:
+        """Check if queue is full"""
+        if not self.scheduler:
+            return False
+        
+        queue_size = len(self.scheduler)
+        return queue_size >= self.max_queue_size * self.backpressure_ratio
+
+    async def _wait_for_capacity(self):
+        """Wait for system to have sufficient capacity"""
+        self._generation_stats['backpressure_events'] += 1
+        self.logger.debug("Backpressure triggered, pausing request generation")
+        
+        wait_time = 0.01  # 减少初始等待时间
+        max_wait = 1.0  # 减少最大等待时间
+        
+        while await self._should_pause_generation() and self.running:
+            await asyncio.sleep(wait_time)
+            wait_time = min(wait_time * 1.1, max_wait)
+
+    async def _open_spider(self):
+        asyncio.create_task(self.crawler.subscriber.notify(CrawlerEvent.SPIDER_OPENED))
+        # 直接调用crawl方法而不是创建任务，确保等待完成
+        await self.crawl()
+
+    async def _crawl(self, request):
+        async def crawl_task():
+            start_time = time.time()
+            try:
+                outputs = await self._fetch(request)
+                # 记录响应时间
+                response_time = time.time() - start_time
+                if self.task_manager:
+                    self.task_manager.record_response_time(response_time)
+                
+                # TODO 处理output
+                if outputs:
+                    await self._handle_spider_output(outputs)
+            except Exception as e:
+                # 记录详细的异常信息
+                self.logger.error(
+                    f"处理请求失败: {getattr(request, 'url', 'Unknown URL')} - {type(e).__name__}: {e}"
+                )
+                self.logger.debug(f"详细异常信息", exc_info=True)
+                
+                # 发送统计事件
+                if hasattr(self.crawler, 'stats'):
+                    self.crawler.stats.inc_value('downloader/exception_count')
+                    self.crawler.stats.inc_value(f'downloader/exception_type_count/{type(e).__name__}')
+                    if hasattr(request, 'url'):
+                        self.crawler.stats.inc_value(f'downloader/failed_urls_count')
+                
+                # 不再重新抛出异常，避免未处理的Task异常
+                return None
+
+        # 使用异步任务创建，遵守并发限制
+        await self.task_manager.create_task(crawl_task())
+
+    async def _fetch(self, request):
+        async def _successful(_response):
+            callback: Callable = request.callback or self.spider.parse
+            if _outputs := callback(_response):
+                if iscoroutine(_outputs):
+                    await _outputs
+                else:
+                    return transform(_outputs, _response)
+
+        _response = await self.downloader.fetch(request)
+        if _response is None:
+            return None
+        output = await _successful(_response)
+        return output
+
+    async def enqueue_request(self, start_request):
+        await self._schedule_request(start_request)
+
+    async def _schedule_request(self, request):
+        # TODO 去重
+        if await self.scheduler.enqueue_request(request):
+            asyncio.create_task(self.crawler.subscriber.notify(CrawlerEvent.REQUEST_SCHEDULED, request, self.crawler.spider))
+
+    async def _get_next_request(self):
+        return await self.scheduler.next_request()
+
+    async def _handle_spider_output(self, outputs):
+        async for spider_output in outputs:
+            if isinstance(spider_output, (Request, Item)):
+                await self.processor.enqueue(spider_output)
+            elif isinstance(spider_output, Exception):
+                asyncio.create_task(
+                    self.crawler.subscriber.notify(CrawlerEvent.SPIDER_ERROR, spider_output, self.spider)
+                )
+                raise spider_output
+            else:
+                raise OutputError(f'{type(self.spider)} must return `Request` or `Item`.')
+
+    async def _exit(self):
+        if self.scheduler.idle() and self.downloader.idle() and self.task_manager.all_done() and self.processor.idle():
+            return True
+        return False
+
+    async def _should_exit(self) -> bool:
+        """检查是否应该退出"""
+        self.logger.debug(f"检查退出条件: start_requests={self.start_requests is not None}")
+        # 没有启动请求，且所有队列都空闲
+        if self.start_requests is None:
+            self.logger.debug("start_requests 为 None，检查其他组件状态")
+            # 使用异步的idle检查方法以获得更精确的结果
+            scheduler_idle = await self.scheduler.async_idle() if hasattr(self.scheduler, 'async_idle') else self.scheduler.idle()
+            downloader_idle = self.downloader.idle()
+            task_manager_done = self.task_manager.all_done()
+            processor_idle = self.processor.idle()
+            
+            self.logger.debug(f"组件状态 - Scheduler: {scheduler_idle}, Downloader: {downloader_idle}, TaskManager: {task_manager_done}, Processor: {processor_idle}")
+            
+            if (scheduler_idle and 
+                downloader_idle and 
+                task_manager_done and 
+                processor_idle):
+                # 立即进行二次检查，不等待
+                scheduler_idle = await self.scheduler.async_idle() if hasattr(self.scheduler, 'async_idle') else self.scheduler.idle()
+                downloader_idle = self.downloader.idle()
+                task_manager_done = self.task_manager.all_done()
+                processor_idle = self.processor.idle()
+                
+                self.logger.debug(f"二次检查组件状态 - Scheduler: {scheduler_idle}, Downloader: {downloader_idle}, TaskManager: {task_manager_done}, Processor: {processor_idle}")
+                
+                if (scheduler_idle and 
+                    downloader_idle and 
+                    task_manager_done and 
+                    processor_idle):
+                    self.logger.info("所有组件都空闲，准备退出")
+                    return True
+        else:
+            self.logger.debug("start_requests 不为 None，不退出")
+        
+        return False
+
+    async def close_spider(self):
+        # 不再调用crawler.close()，避免重复清理
+        # 清理工作应该由crawler的_lifecycle_manager上下文管理器来处理
+        await asyncio.gather(*self.task_manager.current_task)
+        await self.scheduler.close()
+        await self.downloader.close()
+    
+    def get_generation_stats(self) -> dict:
+        """获取生成统计"""
+        return {
+            **self._generation_stats,
+            'queue_size': len(self.scheduler) if self.scheduler else 0,
+            'active_tasks': len(self.task_manager.current_task) if self.task_manager else 0
         }