Scrapy 源碼分析之 RetryMiddleware 模塊
↑ 關(guān)注 + 星標(biāo) ，每天學(xué)Python新技能
后臺回復(fù)【大禮包】送你Python自學(xué)大禮包
時隔一個多月，scrapy 章節(jié)又迎來了重大更新，今天分享的主題是 RetryMiddleware 中間件。文中若有錯誤內(nèi)容，歡迎各位讀者多多指正。在閱讀的同時不要忘記點(diǎn)贊+關(guān)注哦??
一、問題思考
二、文檔查尋
三、源碼分析
四、源碼重寫
五、總結(jié)分享
趣味模塊
娜娜是一名爬蟲工程師，最近小娜在采集數(shù)據(jù)過程中遇到了難題。原因是因?yàn)槿蝿?wù)積壓代理超時了，所有的 request 全部無法下載了。娜娜很是苦惱，不知道如何解決這類型問題。后來小娜看了 TheWeiJun 發(fā)表的文章，存在的問題立馬迎刃而解，接下來，讓我們一起去看看他們是怎么做的吧。
一、問題思考
Question
①使用 scrapy 框架時，如果請求失敗，如何保證該請求成功率？
Question
②scrapy 的重試機(jī)制是否了解，默認(rèn)是幾次？在什么樣的情況下觸發(fā)？
Question
③scrapy 重試機(jī)制，重試狀態(tài)碼有哪些，我們是否可以動態(tài)定義？
Question
④scrapy 在重試過程中，如何實(shí)時更換代理？如何清除失效的代理？
前言：那么帶著這些問題，接下來我們對 Scrapy 源碼進(jìn)行分析探索吧，我相信這篇文章會讓大家受益匪淺！
二、文檔查尋
1、查看官網(wǎng)文檔，搜索指定的模塊 RetryMiddleware，搜索結(jié)果如下：
說明：觀察搜索結(jié)果，我們發(fā)現(xiàn)官方文檔中存在對 RetryMiddleware 模塊的解釋，接下來讓我們點(diǎn)進(jìn)去，一起去看看官方說明吧。
2、點(diǎn)擊搜索結(jié)果，查看官方對當(dāng)前模塊的說明解釋截圖如下：
說明：觀察上面的截圖，我們發(fā)現(xiàn)上面提到的問題大家應(yīng)該已經(jīng)知道了部分答案吧。但是還是不夠清晰，接下來，讓我?guī)Т蠹疫M(jìn)入源碼分析環(huán)節(jié)一探究竟吧！
三、源碼分析
RetryMiddleware 模塊源碼如下：
def get_retry_request(request: Request,*,spider: Spider,reason: Union[str, Exception] = 'unspecified',max_retry_times: Optional[int] = None,priority_adjust: Optional[int] = None,logger: Logger = retry_logger,stats_base_key: str = 'retry',):settings = spider.crawler.settingsstats = spider.crawler.statsretry_times = request.meta.get('retry_times', 0) + 1if max_retry_times is None:max_retry_times = request.meta.get('max_retry_times')if max_retry_times is None:max_retry_times = settings.getint('RETRY_TIMES')if retry_times <= max_retry_times:logger.debug("Retrying %(request)s (failed %(retry_times)d times): %(reason)s",{'request': request, 'retry_times': retry_times, 'reason': reason},extra={'spider': spider})new_request: Request = request.copy()new_request.meta['retry_times'] = retry_timesnew_request.dont_filter = Trueif priority_adjust is None:priority_adjust = settings.getint('RETRY_PRIORITY_ADJUST')new_request.priority = request.priority + priority_adjustif callable(reason):reason = reason()if isinstance(reason, Exception):reason = global_object_name(reason.__class__)stats.inc_value(f'{stats_base_key}/count')stats.inc_value(f'{stats_base_key}/reason_count/{reason}')return new_requestelse:stats.inc_value(f'{stats_base_key}/max_reached')logger.error("Gave up retrying %(request)s (failed %(retry_times)d times): ""%(reason)s",{'request': request, 'retry_times': retry_times, 'reason': reason},extra={'spider': spider},)return Noneclass RetryMiddleware:# IOError is raised by the HttpCompression middleware when trying to# decompress an empty responseEXCEPTIONS_TO_RETRY = (defer.TimeoutError, TimeoutError, DNSLookupError,ConnectionRefusedError, ConnectionDone, ConnectError,ConnectionLost, TCPTimedOutError, ResponseFailed,IOError, TunnelError)def __init__(self, settings):if not settings.getbool('RETRY_ENABLED'):raise NotConfiguredself.max_retry_times = settings.getint('RETRY_TIMES')self.retry_http_codes = set(int(x) for x in settings.getlist('RETRY_HTTP_CODES'))self.priority_adjust = settings.getint('RETRY_PRIORITY_ADJUST')def from_crawler(cls, crawler):return cls(crawler.settings)def process_response(self, request, response, spider):if request.meta.get('dont_retry', False):return responseif response.status in self.retry_http_codes:reason = response_status_message(response.status)return self._retry(request, reason, spider) or responsereturn responsedef process_exception(self, request, exception, spider):if (isinstance(exception, self.EXCEPTIONS_TO_RETRY)and not request.meta.get('dont_retry', False)):return self._retry(request, exception, spider)def _retry(self, request, reason, spider):max_retry_times = request.meta.get('max_retry_times', self.max_retry_times)priority_adjust = request.meta.get('priority_adjust', self.priority_adjust)return get_retry_request(request,reason=reason,spider=spider,max_retry_times=max_retry_times,priority_adjust=priority_adjust,)
環(huán)節(jié)說明：代碼一共也就 94 行，但是卻能實(shí)現(xiàn)多個功能。在好奇心的驅(qū)使下，我們還是對源碼進(jìn)行一一講解分析吧。
from_crawler 函數(shù)
# 類方法,創(chuàng)建當(dāng)前class的實(shí)例對象，參數(shù)：當(dāng)前spider settings對象def from_crawler(cls, crawler):return cls(crawler.settings)
__init__ 函數(shù)
"""這里涉及到了settings.py配置文件中定義的一些參數(shù)。RETRY_ENABLED: 用于開啟中間件，默認(rèn)為TrueRETRY_TIMES: 重試次數(shù), 默認(rèn)為2RETRY_HTTP_CODES: 遇到哪些返回狀態(tài)碼需要重試, 一個列表，默認(rèn)為[500, 503, 504, 400, 408]RETRY_PRIORITY_ADJUST：調(diào)整相對于原始請求的重試請求優(yōu)先級，默認(rèn)為-1"""def __init__(self, settings):if not settings.getbool('RETRY_ENABLED'):raise NotConfiguredself.max_retry_times = settings.getint('RETRY_TIMES')self.retry_http_codes = set(int(x) for x in settings.getlist('RETRY_HTTP_CODES'))self.priority_adjust = settings.getint('RETRY_PRIORITY_ADJUST')
process_response 函數(shù)
process_exception 函數(shù)
EXCEPTIONS_TO_RETRY = (defer.TimeoutError, TimeoutError, DNSLookupError,ConnectionRefusedError, ConnectionDone, ConnectError,ConnectionLost, TCPTimedOutError, ResponseFailed,IOError, TunnelError)def process_response(self, request, response, spider):# 處理request請求，確定是否需要請求重試，重試觸發(fā)機(jī)制，前面提到的問題.if request.meta.get('dont_retry', False):return response# 檢查response狀態(tài)碼是否在重試機(jī)制list中,如果存在就要調(diào)用_retry方法進(jìn)行重試if response.status in self.retry_http_codes:reason = response_status_message(response.status)return self._retry(request, reason, spider) or response# 不存在會返回response，但會被spider parse方法是過濾掉，只處理200狀態(tài)碼return responsedef process_exception(self, request, exception, spider):# 如果產(chǎn)生了EXCEPTIONS_TO_RETRY列表中的異常錯誤并且重試機(jī)制為開啟狀態(tài),則會調(diào)用_retry方法進(jìn)行重試。if (isinstance(exception, self.EXCEPTIONS_TO_RETRY)and not request.meta.get('dont_retry', False)):return self._retry(request, exception, spider)
_retry 函數(shù)
get_retry_request 函數(shù)
# 該方法獲取最大重試次數(shù)，和請求重試優(yōu)先級，然后調(diào)用get_retry_request方法def _retry(self, request, reason, spider):max_retry_times = request.meta.get('max_retry_times', self.max_retry_times)priority_adjust = request.meta.get('priority_adjust', self.priority_adjust)return get_retry_request(request,reason=reason,spider=spider,max_retry_times=max_retry_times,priority_adjust=priority_adjust,)"""讀取當(dāng)前重試次數(shù)和最大重試次數(shù)進(jìn)行比較，如果小于等于最大重試次數(shù)：利用copy方法在原來的request上復(fù)制一個新request，并更新其retry_times，并將dont_filter設(shè)為True來防止因url重復(fù)而被過濾。如果超出最大重試次數(shù)：記錄重試失敗請求量，并放棄該請求記錄到logger日志中,logger級別為：error"""def get_retry_request(request: Request,*,spider: Spider,reason: Union[str, Exception] = 'unspecified',max_retry_times: Optional[int] = None,priority_adjust: Optional[int] = None,logger: Logger = retry_logger,stats_base_key: str = 'retry',):settings = spider.crawler.settingsstats = spider.crawler.statsretry_times = request.meta.get('retry_times', 0) + 1if max_retry_times is None:max_retry_times = request.meta.get('max_retry_times')if max_retry_times is None:max_retry_times = settings.getint('RETRY_TIMES')if retry_times <= max_retry_times:logger.debug("Retrying %(request)s (failed %(retry_times)d times): %(reason)s",{'request': request, 'retry_times': retry_times, 'reason': reason},extra={'spider': spider})new_request: Request = request.copy()new_request.meta['retry_times'] = retry_timesnew_request.dont_filter = Trueif priority_adjust is None:priority_adjust = settings.getint('RETRY_PRIORITY_ADJUST')new_request.priority = request.priority + priority_adjustif callable(reason):reason = reason()if isinstance(reason, Exception):reason = global_object_name(reason.__class__)stats.inc_value(f'{stats_base_key}/count')stats.inc_value(f'{stats_base_key}/reason_count/{reason}')return new_requestelse:stats.inc_value(f'{stats_base_key}/max_reached')logger.error("Gave up retrying %(request)s (failed %(retry_times)d times): ""%(reason)s",{'request': request, 'retry_times': retry_times, 'reason': reason},extra={'spider': spider},)return None
環(huán)節(jié)總結(jié)：整個源碼分析流程到這里就結(jié)束了，接下來我們一起進(jìn)入源碼重寫環(huán)節(jié)來解決下娜娜遇到的問題吧，我相信大家會豁然開朗的。
四、源碼重寫
重寫 RetryMiddleware 源碼后完整代碼如下：
class RetryMiddleware:EXCEPTIONS_TO_RETRY = (defer.TimeoutError, TimeoutError, DNSLookupError,ConnectionRefusedError, ConnectionDone, ConnectError,ConnectionLost, TCPTimedOutError, ResponseFailed,IOError, TunnelError)def __init__(self, settings):if not settings.getbool('RETRY_ENABLED'):raise NotConfiguredself.max_retry_times = settings.getint('RETRY_TIMES')self.retry_http_codes = set(int(x) for x in settings.getlist('RETRY_HTTP_CODES'))self.priority_adjust = settings.getint('RETRY_PRIORITY_ADJUST')@classmethoddef from_crawler(cls, crawler):return cls(crawler.settings)def process_response(self, request, response, spider):if request.meta.get('dont_retry', False):return responseif response.status in self.retry_http_codes: # 可以自定義重試狀態(tài)碼reason = response_status_message(response.status)response.last_content = request.metareturn self._retry(request, reason, spider) or responsereturn responsedef process_exception(self, request, exception, spider):if (isinstance(exception, self.EXCEPTIONS_TO_RETRY)and not request.meta.get('dont_retry', False)):return self._retry(request, exception, spider)def _retry(self, request, reason, spider):max_retry_times = request.meta.get('max_retry_times', self.max_retry_times)priority_adjust = request.meta.get('priority_adjust', self.priority_adjust)request.meta['proxy'] = "xxx:xxxx"request.headers['Proxy-Authorization'] = "proxyauth"return get_retry_request(request,reason=reason,spider=spider,max_retry_times=max_retry_times,priority_adjust=priority_adjust,)
重寫總結(jié)：我們只需要在 _retry 函數(shù)中實(shí)時更換代理即可，如果涉及到代理池需要剔除失敗代理的問題，同樣在 _retry 函數(shù)中刪除代理池中指定代理即可。我們還可以自定義重試機(jī)制狀態(tài)碼，大家可自行添加即可！
五、總結(jié)分享
通過本次案例分析，上面的幾個問題我們都已經(jīng)得到了答案。今天分享到這里就結(jié)束了，歡迎大家關(guān)注下期文章，我們不見不散??。最后希望大家多多轉(zhuǎn)發(fā)、點(diǎn)贊、在看支持一波
評論
圖片
表情