Skip to content

S
source_forward
Commit 9fcfba2d by zhiwei
Options

各个采集验证添加休眠，避免数据过多导致程序阻塞

parent aa059934
Showing with 594 additions and 588 deletions
src/main/java/com/zhiwei/source_forward/crawler/ContentCrawler.java
package com.zhiwei.source_forward.crawler; package com.zhiwei.source_forward.crawler;
import java.util.Objects; import java.util.Objects;
import org.apache.logging.log4j.LogManager; import com.zhiwei.tools.tools.ZhiWeiTools;
import org.apache.logging.log4j.Logger; import org.apache.logging.log4j.LogManager;
import org.apache.logging.log4j.Logger;
import com.zhiwei.async.GroupSync;
import com.zhiwei.crawler.core.HttpBoot; import com.zhiwei.async.GroupSync;
import com.zhiwei.crawler.core.proxy.ProxyHolder; import com.zhiwei.crawler.core.HttpBoot;
import com.zhiwei.crawler.core.utils.RequestUtils; import com.zhiwei.crawler.core.proxy.ProxyHolder;
import com.zhiwei.source_forward.bean.ContentBean; import com.zhiwei.crawler.core.utils.RequestUtils;
import com.zhiwei.source_forward.bean.ContentBean.Attribution; import com.zhiwei.source_forward.bean.ContentBean;
import com.zhiwei.source_forward.util.ContentDataCallback; import com.zhiwei.source_forward.bean.ContentBean.Attribution;
import com.zhiwei.source_forward.util.MatchContent; import com.zhiwei.source_forward.util.ContentDataCallback;
import com.zhiwei.source_forward.util.MatchContent;
import okhttp3.Request;
import okhttp3.Request;
public class ContentCrawler {
public class ContentCrawler {
private static Logger logger = LogManager.getLogger(ContentCrawler.class);
private static HttpBoot httpBoot = new HttpBoot.Builder().retryTimes(2).build(); private static Logger logger = LogManager.getLogger(ContentCrawler.class);
private static HttpBoot httpBoot = new HttpBoot.Builder().retryTimes(2).build();
/**
* /**
* @Description 链接传入 并 返回采集完信号 *
* @param callback * @Description 链接传入 并 返回采集完信号
* @param urls * @param callback
* @return * @param urls
* @throws Exception * @return
*/ * @throws Exception
public GroupSync submitTask(ContentDataCallback callback, */
String... urls) { public GroupSync submitTask(ContentDataCallback callback,
GroupSync counter = new GroupSync(); String... urls) {
start(counter, callback, urls); GroupSync counter = new GroupSync();
return counter; start(counter, callback, urls);
} return counter;
}
/**
* /**
* @Description 提交链接 *
* @param counter * @Description 提交链接
* @param callback * @param counter
* @param urls * @param callback
*/ * @param urls
private void start(GroupSync counter, */
ContentDataCallback callback, String... urls) { private void start(GroupSync counter,
if (urls != null && urls.length > 0) { ContentDataCallback callback, String... urls) {
for (String url : urls) { if (urls != null && urls.length > 0) {
if (url != null) { for (String url : urls) {
try { ZhiWeiTools.sleep(100);
search(counter, url, Attribution.of(url), callback); if (url != null) {
} catch (Exception e) { try {
logger.error("搜索创建出错", e); search(counter, url, Attribution.of(url), callback);
} } catch (Exception e) {
} logger.error("搜索创建出错", e);
} }
} }
} }
}
/** }
*
* @Description 链接获取文章信息 /**
* @param counter *
* @param url * @Description 链接获取文章信息
* @param attr * @param counter
* @param callback * @param url
* @return * @param attr
*/ * @param callback
private GroupSync search(GroupSync counter, * @return
String url, Attribution attr, ContentDataCallback callback) { */
logger.info("当前处理 URL: {}", url); private GroupSync search(GroupSync counter,
Request request = RequestUtils.wrapGet(url); String url, Attribution attr, ContentDataCallback callback) {
counter.add(); logger.info("当前处理 URL: {}", url);
Request request = RequestUtils.wrapGet(url);
httpBoot.asyncCall(request, ProxyHolder.NAT_HEAVY_PROXY, true).whenComplete((rs,ex) -> { counter.add();
try {
if (Objects.isNull(ex)) { httpBoot.asyncCall(request, ProxyHolder.NAT_HEAVY_PROXY, true).whenComplete((rs,ex) -> {
parseHtml(rs.body().string(), attr, callback); try {
} else { if (Objects.isNull(ex)) {
logger.info("{} 搜索结果访问失败: {}", request.url().url(),ex); parseHtml(rs.body().string(), attr, callback);
} } else {
} catch (Exception e) { logger.info("{} 搜索结果访问失败: {}", request.url().url(),ex);
logger.info("搜索结果访问失败: {}", ex); }
} finally { } catch (Exception e) {
counter.done(); logger.info("搜索结果访问失败: {}", ex);
} } finally {
counter.done();
}); }
return counter; });
}
return counter;
/** }
*
* /**
* @Description 获取正文解析 *
* @param response *
* @param attr * @Description 获取正文解析
* @param callback * @param response
*/ * @param attr
private void parseHtml(String result, Attribution attr, * @param callback
ContentDataCallback callback) { */
try { private void parseHtml(String result, Attribution attr,
String content = MatchContent.matchContent(attr.get().toString(), ContentDataCallback callback) {
result); try {
ContentBean cb = new ContentBean(attr.get().toString(), content); String content = MatchContent.matchContent(attr.get().toString(),
if (callback == null) { result);
logger.warn("DataCallback 对象为 null，无法保存数据"); ContentBean cb = new ContentBean(attr.get().toString(), content);
} else { if (callback == null) {
callback.onData(cb, attr); logger.warn("DataCallback 对象为 null，无法保存数据");
} } else {
} catch (Exception e) { callback.onData(cb, attr);
logger.error("网页链接失效", e); }
} } catch (Exception e) {
logger.error("网页链接失效", e);
} }
} }
}
src/main/java/com/zhiwei/source_forward/crawler/MediaSelfSourceCrawler.java
...@@ -5,6 +5,7 @@ import java.util.List; ...@@ -5,6 +5,7 @@ import java.util.List;
import java.util.Map; import java.util.Map;
import java.util.Objects; import java.util.Objects;
import com.zhiwei.tools.tools.ZhiWeiTools;
import org.apache.logging.log4j.LogManager; import org.apache.logging.log4j.LogManager;
import org.apache.logging.log4j.Logger; import org.apache.logging.log4j.Logger;
import org.jsoup.Jsoup; import org.jsoup.Jsoup;
...@@ -64,6 +65,7 @@ public class MediaSelfSourceCrawler { ...@@ -64,6 +65,7 @@ public class MediaSelfSourceCrawler {
private void start(GroupSync counter,MediaSelfSourceDataCallBack callback, String... urls) { private void start(GroupSync counter,MediaSelfSourceDataCallBack callback, String... urls) {
if (urls != null && urls.length > 0) { if (urls != null && urls.length > 0) {
for (String url : urls) { for (String url : urls) {
ZhiWeiTools.sleep(100);
counter.add(); counter.add();
if (url != null) { if (url != null) {
try { try {
......
src/main/java/com/zhiwei/source_forward/crawler/SourceForwardCrawler.java
package com.zhiwei.source_forward.crawler; package com.zhiwei.source_forward.crawler;
import java.util.HashMap; import java.util.HashMap;
import java.util.List; import java.util.List;
import java.util.Map; import java.util.Map;
import java.util.Objects; import java.util.Objects;
import org.apache.logging.log4j.LogManager; import com.zhiwei.tools.tools.ZhiWeiTools;
import org.apache.logging.log4j.Logger; import org.apache.logging.log4j.LogManager;
import org.jsoup.Jsoup; import org.apache.logging.log4j.Logger;
import org.jsoup.nodes.Document; import org.jsoup.Jsoup;
import org.jsoup.nodes.Node; import org.jsoup.nodes.Document;
import org.jsoup.nodes.Node;
import com.zhiwei.async.GroupSync;
import com.zhiwei.crawler.core.HttpBoot; import com.zhiwei.async.GroupSync;
import com.zhiwei.crawler.core.proxy.ProxyHolder; import com.zhiwei.crawler.core.HttpBoot;
import com.zhiwei.crawler.core.utils.RequestUtils; import com.zhiwei.crawler.core.proxy.ProxyHolder;
import com.zhiwei.source_forward.bean.SourceForwardBean; import com.zhiwei.crawler.core.utils.RequestUtils;
import com.zhiwei.source_forward.bean.SourceForwardBean.Attribution; import com.zhiwei.source_forward.bean.SourceForwardBean;
import com.zhiwei.source_forward.util.MatchChannel; import com.zhiwei.source_forward.bean.SourceForwardBean.Attribution;
import com.zhiwei.source_forward.util.MatchSource; import com.zhiwei.source_forward.util.MatchChannel;
import com.zhiwei.source_forward.util.SourceData; import com.zhiwei.source_forward.util.MatchSource;
import com.zhiwei.source_forward.util.SourceForwardDataCallBack; import com.zhiwei.source_forward.util.SourceData;
import com.zhiwei.source_forward.util.SourceForwardDataCallBack;
import okhttp3.Request;
import okhttp3.Request;
public class SourceForwardCrawler {
public class SourceForwardCrawler {
private static final Logger logger = LogManager.getLogger(SourceForwardCrawler.class);
private static final Logger logger = LogManager.getLogger(SourceForwardCrawler.class);
private static HttpBoot httpBoot = new HttpBoot.Builder().retryTimes(2).build();
private static List<String> sourceList = SourceData.getSourceList(); private static HttpBoot httpBoot = new HttpBoot.Builder().retryTimes(2).build();
private static List<String> sourceList = SourceData.getSourceList();
public GroupSync submitTask(SourceForwardDataCallBack callback,String... urls) {
try { public GroupSync submitTask(SourceForwardDataCallBack callback,String... urls) {
GroupSync counter = new GroupSync(); try {
start(counter, callback, urls); GroupSync counter = new GroupSync();
return counter; start(counter, callback, urls);
} catch (Exception e) { return counter;
logger.error(" exception ", e); } catch (Exception e) {
return null; logger.error(" exception ", e);
} return null;
} }
}
private void start(GroupSync counter,SourceForwardDataCallBack callback, String... urls) {
if (urls != null && urls.length > 0) { private void start(GroupSync counter,SourceForwardDataCallBack callback, String... urls) {
for (String url : urls) { if (urls != null && urls.length > 0) {
counter.add(); for (String url : urls) {
if (url != null) { counter.add();
try { ZhiWeiTools.sleep(100);
search(counter, url, Attribution.of(url), callback); if (url != null) {
} catch (Exception e) { try {
logger.error("搜索创建出错", e); search(counter, url, Attribution.of(url), callback);
} } catch (Exception e) {
} logger.error("搜索创建出错", e);
counter.done(); }
} }
} counter.done();
} }
}
private GroupSync search(GroupSync counter, String url,Attribution attr, SourceForwardDataCallBack callback) { }
logger.info("当前处理 URL: {}", url);
Map<String,String> headers = new HashMap<>(); private GroupSync search(GroupSync counter, String url,Attribution attr, SourceForwardDataCallBack callback) {
// Map<String,String> headers = HeaderTool.getCommonHead(); logger.info("当前处理 URL: {}", url);
if(url.contains("www.toutiao.com")){ Map<String,String> headers = new HashMap<>();
headers.put("referer", url); // Map<String,String> headers = HeaderTool.getCommonHead();
} if(url.contains("www.toutiao.com")){
if(url.contains("china.prcfe.com")) { headers.put("referer", url);
url = "http://china.prcfe.com/e/extend/ShowSource/?id=" + url.split("/")[url.split("/").length-1].split("\\.")[0]; }
} if(url.contains("china.prcfe.com")) {
Request request = RequestUtils.wrapGet(url, headers); url = "http://china.prcfe.com/e/extend/ShowSource/?id=" + url.split("/")[url.split("/").length-1].split("\\.")[0];
counter.add(); }
httpBoot.asyncCall(request, ProxyHolder.NAT_HEAVY_PROXY, true).whenComplete((rs,ex) -> { Request request = RequestUtils.wrapGet(url, headers);
try { counter.add();
if (Objects.isNull(ex)) { httpBoot.asyncCall(request, ProxyHolder.NAT_HEAVY_PROXY, true).whenComplete((rs,ex) -> {
parseHtml(rs.body().string(), attr, callback); try {
} else { if (Objects.isNull(ex)) {
logger.info("{} 搜索结果访问失败: {}", request.url().url(), ex); parseHtml(rs.body().string(), attr, callback);
} } else {
} catch (Exception e1) { logger.info("{} 搜索结果访问失败: {}", request.url().url(), ex);
logger.error("解析出错",e1); }
} finally { } catch (Exception e1) {
counter.done(); logger.error("解析出错",e1);
} } finally {
}); counter.done();
return counter; }
} });
return counter;
private void parseHtml(String body, Attribution attr, }
SourceForwardDataCallBack callback) {
String source = null; private void parseHtml(String body, Attribution attr,
String channel = "新闻"; SourceForwardDataCallBack callback) {
String isforward = "未知"; String source = null;
try { String channel = "新闻";
Document document = Jsoup.parse(body); String isforward = "未知";
if(attr.get().toString().contains("mp.weixin.qq.com")){ try {
isforward = document.select("div#meta_content").select("span#copyright_logo").text(); Document document = Jsoup.parse(body);
if(isforward.contains("原创")){ if(attr.get().toString().contains("mp.weixin.qq.com")){
isforward = "原创"; isforward = document.select("div#meta_content").select("span#copyright_logo").text();
}else { if(isforward.contains("原创")){
isforward = "未知"; isforward = "原创";
} }else {
}else if(attr.get().toString().contains("www.toutiao.com")){ isforward = "未知";
if(body.contains("isOriginal") && body.contains("isOriginal: true")){ }
isforward = "原创"; }else if(attr.get().toString().contains("www.toutiao.com")){
} if(body.contains("isOriginal") && body.contains("isOriginal: true")){
}else{ isforward = "原创";
channel = MatchChannel.verifyChannel(attr.get().toString()); }
if(channel==null){ }else{
List<Node> nodeList = document.head().childNodes(); channel = MatchChannel.verifyChannel(attr.get().toString());
channel = MatchChannel.matchChannel(nodeList); if(channel==null){
} List<Node> nodeList = document.head().childNodes();
source = MatchSource.matchSource(attr.get().toString(),document.toString(), sourceList); channel = MatchChannel.matchChannel(nodeList);
} }
} catch (Exception e) { source = MatchSource.matchSource(attr.get().toString(),document.toString(), sourceList);
source = null; }
channel = "新闻"; } catch (Exception e) {
} source = null;
logger.info(attr.get().toString()+"======="+channel+"================="+source); channel = "新闻";
SourceForwardBean sfb = new SourceForwardBean(attr.get().toString(), channel, source,isforward); }
if (callback == null) { logger.info(attr.get().toString()+"======="+channel+"================="+source);
logger.warn("DataCallback 对象为 null，无法保存数据"); SourceForwardBean sfb = new SourceForwardBean(attr.get().toString(), channel, source,isforward);
} else { if (callback == null) {
callback.onData(sfb, attr); logger.warn("DataCallback 对象为 null，无法保存数据");
} } else {
} callback.onData(sfb, attr);
}
} }
}
src/main/java/com/zhiwei/source_forward/crawler/UrlLiveCrawler.java
package com.zhiwei.source_forward.crawler; package com.zhiwei.source_forward.crawler;
import static java.util.Objects.nonNull; import static java.util.Objects.nonNull;
import java.util.Arrays; import java.util.Arrays;
import java.util.HashMap; import java.util.HashMap;
import java.util.List; import java.util.List;
import java.util.Map; import java.util.Map;
import java.util.Objects; import java.util.Objects;
import java.util.regex.Matcher; import java.util.regex.Matcher;
import java.util.regex.Pattern; import java.util.regex.Pattern;
import org.apache.logging.log4j.LogManager; import org.apache.logging.log4j.LogManager;
import org.apache.logging.log4j.Logger; import org.apache.logging.log4j.Logger;
import org.jsoup.Jsoup; import org.jsoup.Jsoup;
import org.jsoup.nodes.Document; import org.jsoup.nodes.Document;
import com.alibaba.fastjson.JSONObject; import com.alibaba.fastjson.JSONObject;
import com.zhiwei.async.GroupSync; import com.zhiwei.async.GroupSync;
import com.zhiwei.crawler.core.HttpBoot; import com.zhiwei.crawler.core.HttpBoot;
import com.zhiwei.crawler.core.proxy.ProxyHolder; import com.zhiwei.crawler.core.proxy.ProxyHolder;
import com.zhiwei.crawler.core.utils.RequestUtils; import com.zhiwei.crawler.core.utils.RequestUtils;
import com.zhiwei.source_forward.bean.UrlLiveBean; import com.zhiwei.source_forward.bean.UrlLiveBean;
import com.zhiwei.source_forward.bean.UrlLiveBean.Attribution; import com.zhiwei.source_forward.bean.UrlLiveBean.Attribution;
import com.zhiwei.source_forward.util.UrlLiveDataCallback; import com.zhiwei.source_forward.util.UrlLiveDataCallback;
import com.zhiwei.tools.tools.ZhiWeiTools; import com.zhiwei.tools.tools.ZhiWeiTools;
import okhttp3.Request; import okhttp3.Request;
/** /**
* *
* @ClassName UrlLiveCrawler * @ClassName UrlLiveCrawler
* @Description 判断页面是否存在 * @Description 判断页面是否存在
* @author byte-zbs * @author byte-zbs
* @Date 2018年8月20日 下午3:34:57 * @Date 2018年8月20日 下午3:34:57
* @version 1.0.0 * @version 1.0.0
*/ */
public class UrlLiveCrawler { public class UrlLiveCrawler {
private static final Logger logger = LogManager.getLogger(UrlLiveCrawler.class); private static final Logger logger = LogManager.getLogger(UrlLiveCrawler.class);
private static HttpBoot httpBoot = new HttpBoot.Builder().retryTimes(2).throwException(true).build(); private static HttpBoot httpBoot = new HttpBoot.Builder().retryTimes(2).throwException(true).build();
public GroupSync submitTask(UrlLiveDataCallback callback,String... urls) { public GroupSync submitTask(UrlLiveDataCallback callback,String... urls) {
GroupSync counter = new GroupSync(); GroupSync counter = new GroupSync();
start(counter, callback, urls); start(counter, callback, urls);
return counter; return counter;
} }
private void start(GroupSync counter,UrlLiveDataCallback callback, String... urls) { private void start(GroupSync counter,UrlLiveDataCallback callback, String... urls) {
if (nonNull(urls) && urls.length > 0) { if (nonNull(urls) && urls.length > 0) {
for (String url : urls) { for (String url : urls) {
counter.add(); counter.add();
if (nonNull(url)) { ZhiWeiTools.sleep(100);
try { if (nonNull(url)) {
// ZhiWeiTools.sleep(3000); try {
search(counter, url, Attribution.of(url, 1), callback); // ZhiWeiTools.sleep(3000);
} catch (Exception e) { search(counter, url, Attribution.of(url, 1), callback);
logger.error("搜索创建出错:", e); } catch (Exception e) {
} logger.error("搜索创建出错:", e);
} }
counter.done(); }
} counter.done();
} }
} }
}
private GroupSync search(GroupSync counter, String url,
Attribution attr, UrlLiveDataCallback callback) { private GroupSync search(GroupSync counter, String url,
// System.out.println(url); Attribution attr, UrlLiveDataCallback callback) {
url = dealUrl(url); // System.out.println(url);
logger.info("当前处理 URL: {}", url); url = dealUrl(url);
Map<String,String> headers = new HashMap<>(); logger.info("当前处理 URL: {}", url);
ProxyHolder ph = null; Map<String,String> headers = new HashMap<>();
if(url.contains("toutiao.com")){ ProxyHolder ph = null;
// headers.put("referer", url); if(url.contains("toutiao.com")){
// headers.put("cookie", "csrftoken=6d0e5967684dbb57cea14dc12858d263; WEATHER_CITY=%E5%8C%97%E4%BA%AC; __utmz=24953151.1576744435.1.1.utmcsr=(direct)|utmccn=(direct)|utmcmd=(none); _ga=GA1.2.1570444666.1576744435; ttcid=d5fa0cec2bca426cbbddeb27a8bc72f629; UM_distinctid=170f5ad3a5b43e-04a6930812e3c5-f313f6d-240000-170f5ad3a5c40b; SLARDAR_WEB_ID=770be4fe-dd54-4701-a4ed-396b861a237d; CNZZDATA1259612802=2091325281-1587691681-%7C1589960650; s_v_web_id=verify_kaqh0jch_kZtSHRI9_GfBj_403X_8i5y_VOyWz4ATIx0d; __utmc=24953151; __utma=24953151.1570444666.1576744435.1591353562.1591579877.8; tt_webid=6837283963338622477; tt_webid=6837283963338622477; __tasessionId=dsei2aty41591951911851; tt_scid=M--AJ-FYwZ0qcYTzQCLyMeS5MlykLS6ktMkvqKKJmq-ghRxX4waEBhJ3YbheuNmi2b8a"); // headers.put("referer", url);
// headers.put("accept", "text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,image/apng,*/*;q=0.8,application/signed-exchange;v=b3;q=0.9"); // headers.put("cookie", "csrftoken=6d0e5967684dbb57cea14dc12858d263; WEATHER_CITY=%E5%8C%97%E4%BA%AC; __utmz=24953151.1576744435.1.1.utmcsr=(direct)|utmccn=(direct)|utmcmd=(none); _ga=GA1.2.1570444666.1576744435; ttcid=d5fa0cec2bca426cbbddeb27a8bc72f629; UM_distinctid=170f5ad3a5b43e-04a6930812e3c5-f313f6d-240000-170f5ad3a5c40b; SLARDAR_WEB_ID=770be4fe-dd54-4701-a4ed-396b861a237d; CNZZDATA1259612802=2091325281-1587691681-%7C1589960650; s_v_web_id=verify_kaqh0jch_kZtSHRI9_GfBj_403X_8i5y_VOyWz4ATIx0d; __utmc=24953151; __utma=24953151.1570444666.1576744435.1591353562.1591579877.8; tt_webid=6837283963338622477; tt_webid=6837283963338622477; __tasessionId=dsei2aty41591951911851; tt_scid=M--AJ-FYwZ0qcYTzQCLyMeS5MlykLS6ktMkvqKKJmq-ghRxX4waEBhJ3YbheuNmi2b8a");
// headers.put("accept-encoding", "gzip, deflate, br"); // headers.put("accept", "text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,image/apng,*/*;q=0.8,application/signed-exchange;v=b3;q=0.9");
// headers.put("accept-language", "zh-CN,zh;q=0.9"); // headers.put("accept-encoding", "gzip, deflate, br");
// headers.put("cache-control", "no-cache"); // headers.put("accept-language", "zh-CN,zh;q=0.9");
// headers.put("sec-fetch-dest", "document"); // headers.put("cache-control", "no-cache");
// headers.put("sec-fetch-mode", "navigate"); // headers.put("sec-fetch-dest", "document");
// headers.put("sec-fetch-site", "same-origin"); // headers.put("sec-fetch-mode", "navigate");
// headers.put("sec-fetch-user", "?1"); // headers.put("sec-fetch-site", "same-origin");
// headers.put("upgrade-insecure-requests", "1"); // headers.put("sec-fetch-user", "?1");
// headers.put("user-agent", "Mozilla/5.0 (Linux; Android 5.0; SM-G900P Build/LRX21T) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/83.0.4103.97 Mobile Safari/537.36"); // headers.put("upgrade-insecure-requests", "1");
ph = ProxyHolder.NAT_HEAVY_PROXY; // headers.put("user-agent", "Mozilla/5.0 (Linux; Android 5.0; SM-G900P Build/LRX21T) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/83.0.4103.97 Mobile Safari/537.36");
}else if(url.contains("zhihu.com")) { ph = ProxyHolder.NAT_HEAVY_PROXY;
url = treatZhihuUrl(url); }else if(url.contains("zhihu.com")) {
ph = ProxyHolder.NAT_HEAVY_PROXY; url = treatZhihuUrl(url);
} ph = ProxyHolder.NAT_HEAVY_PROXY;
try { }
Request request = RequestUtils.wrapGet(url, headers); try {
if(Objects.nonNull(request)) { Request request = RequestUtils.wrapGet(url, headers);
counter.add(); if(Objects.nonNull(request)) {
// , new Proxy(Type.HTTP, new InetSocketAddress("119.3.38.9", 31128)) counter.add();
httpBoot.asyncCall(request, ph).whenComplete((rs,ex) -> { // , new Proxy(Type.HTTP, new InetSocketAddress("119.3.38.9", 31128))
try { httpBoot.asyncCall(request, ph).whenComplete((rs,ex) -> {
if (Objects.isNull(ex)) { try {
if(rs.isSuccessful()) { if (Objects.isNull(ex)) {
parseHtml(rs.body().string(), attr, callback); if(rs.isSuccessful()) {
}else if(rs.code() == 404){ parseHtml(rs.body().string(), attr, callback);
callBack(callback, attr, 1, String.valueOf(rs.code())); }else if(rs.code() == 404){
}else { callBack(callback, attr, 1, String.valueOf(rs.code()));
callBack(callback, attr, -1, "程序无法判断"); }else {
} callBack(callback, attr, -1, "程序无法判断");
} else { }
logger.error("e", ex); } else {
callBack(callback, attr, -1, "程序无法判断"); logger.error("e", ex);
} callBack(callback, attr, -1, "程序无法判断");
} catch (Exception e) { }
logger.error(" 数据是否删除 采集出错 {} ",e); } catch (Exception e) {
}finally { logger.error(" 数据是否删除 采集出错 {} ",e);
counter.done(); }finally {
} counter.done();
}); }
return counter; });
} return counter;
} catch (Exception e2) { }
logger.error("数据出错 {}" ,e2); } catch (Exception e2) {
} logger.error("数据出错 {}" ,e2);
return counter; }
} return counter;
}
private void callBack(UrlLiveDataCallback callback,Attribution attr,int i,String title) {
UrlLiveBean ulb = null; private void callBack(UrlLiveDataCallback callback,Attribution attr,int i,String title) {
if(i == 1) { UrlLiveBean ulb = null;
ulb = new UrlLiveBean(attr.getAttr().toString(), true, title); if(i == 1) {
}else { ulb = new UrlLiveBean(attr.getAttr().toString(), true, title);
ulb = new UrlLiveBean(attr.getAttr().toString(), i, title); }else {
} ulb = new UrlLiveBean(attr.getAttr().toString(), i, title);
if (callback == null) { }
logger.warn("DataCallback 对象为 null，无法保存数据"); if (callback == null) {
} else { logger.warn("DataCallback 对象为 null，无法保存数据");
callback.onData(ulb, attr); } else {
} callback.onData(ulb, attr);
} }
}
private String dealUrl(String url) {
try { private String dealUrl(String url) {
if(url.contains("www.toutiao.com")) { try {
if(url.contains("www.toutiao.com")) { if(url.contains("www.toutiao.com")) {
if(url.contains("www.toutiao.com")) {
}else {
url = url.replace("toutiao.com", "www.toutiao.com"); }else {
} url = url.replace("toutiao.com", "www.toutiao.com");
if(url.contains("https")) { }
if(url.contains("https")) {
}else {
url = url.replace("http", "https"); }else {
} url = url.replace("http", "https");
if(url.contains("group")) { }
url = "https://www.toutiao.com/a" + url.split("/")[4] + "/"; if(url.contains("group")) {
} url = "https://www.toutiao.com/a" + url.split("/")[4] + "/";
}else if(url.contains("mp.weixin.qq.com")) { }
if(url.contains("https")) { }else if(url.contains("mp.weixin.qq.com")) {
if(url.contains("https")) {
}else {
url = url.replace("http", "https"); }else {
} url = url.replace("http", "https");
}else if(url.contains("a.mp.uc.cn") && url.contains("wm_aid")) { }
url = "http://ff.dayu.com/contents/origin/"+url.split("wm_aid=")[1].split("!!wm_id")[0]+"?biz_id=1002&_fetch_author=1&_fetch_incrs=1"; }else if(url.contains("a.mp.uc.cn") && url.contains("wm_aid")) {
}else if(url.contains("tznew.58.com/view") && url.contains("infoid=")) { url = "http://ff.dayu.com/contents/origin/"+url.split("wm_aid=")[1].split("!!wm_id")[0]+"?biz_id=1002&_fetch_author=1&_fetch_incrs=1";
// https://tznew.58.com/view/c/sharingDetailNew?infoid=117073473 }else if(url.contains("tznew.58.com/view") && url.contains("infoid=")) {
url = "https://tznew.58.com/tznew/c/info-detail?infoid=" + url.split("infoid=")[1].split("&")[0]; // https://tznew.58.com/view/c/sharingDetailNew?infoid=117073473
} url = "https://tznew.58.com/tznew/c/info-detail?infoid=" + url.split("infoid=")[1].split("&")[0];
return url; }
} catch (Exception e) { return url;
return url; } catch (Exception e) {
} return url;
} }
}
/**
* /**
* @Description 判断是否删除 *
* @param html * @Description 判断是否删除
* @param attr * @param html
* @param callback * @param attr
*/ * @param callback
private void parseHtml(String html, Attribution attr, */
UrlLiveDataCallback callback) { private void parseHtml(String html, Attribution attr,
if (callback == null) { UrlLiveDataCallback callback) {
logger.warn("DataCallback 对象为 null，无法保存数据"); if (callback == null) {
} else { logger.warn("DataCallback 对象为 null，无法保存数据");
UrlLiveBean ulb = matchDel(html, attr, attr.getAttr().toString()); } else {
if(Objects.nonNull(ulb)) { UrlLiveBean ulb = matchDel(html, attr, attr.getAttr().toString());
callback.onData(ulb, attr); if(Objects.nonNull(ulb)) {
}else { callback.onData(ulb, attr);
callBack(callback, attr, -1, "程序无法判断"); }else {
} callBack(callback, attr, -1, "程序无法判断");
} }
} }
}
/***
* @Title: matchDel /***
* @author hero * @Title: matchDel
* @Description: 验证链接是否有效 * @author hero
* @param @param page * @Description: 验证链接是否有效
* @param @return 设定文件 * @param @param page
* @return boolean 返回类型 * @param @return 设定文件
*/ * @return boolean 返回类型
public UrlLiveBean matchDel(String result,Attribution attr,String url){ */
try { public UrlLiveBean matchDel(String result,Attribution attr,String url){
Document doc = Jsoup.parse(result); try {
String title = null; Document doc = Jsoup.parse(result);
if(url.contains("mp.weixin.qq.com") || url.contains("post.mp.qq.com") || url.contains("weixin.sogou.com")){ String title = null;
title = doc.select("h2.rich_media_title").text().replaceAll(" ", ""); if(url.contains("mp.weixin.qq.com") || url.contains("post.mp.qq.com") || url.contains("weixin.sogou.com")){
if(Objects.isNull(title) || title.isEmpty()) { title = doc.select("h2.rich_media_title").text().replaceAll(" ", "");
title = doc.select("p.title").text(); if(Objects.isNull(title) || title.isEmpty()) {
} title = doc.select("p.title").text();
if(Objects.isNull(title) || title.isEmpty()) { }
title = doc.select("h3.msg-title").text(); if(Objects.isNull(title) || title.isEmpty()) {
} title = doc.select("h3.msg-title").text();
if(Objects.isNull(title) || title.isEmpty()) { }
title = doc.select("div.global_error_msg.warn").text(); if(Objects.isNull(title) || title.isEmpty()) {
} title = doc.select("div.global_error_msg.warn").text();
if(Objects.isNull(title) || title.isEmpty()) { }
title = doc.select("p.tips").text(); if(Objects.isNull(title) || title.isEmpty()) {
} title = doc.select("p.tips").text();
if(Objects.isNull(title) || title.isEmpty()) { }
title = doc.select("h2").text(); if(Objects.isNull(title) || title.isEmpty()) {
} title = doc.select("h2").text();
// 获取title }
Matcher ma5 = Pattern.compile("var msg_title = \'(.*)\'") // 获取title
.matcher(result); Matcher ma5 = Pattern.compile("var msg_title = \'(.*)\'")
if (ma5.find()) { .matcher(result);
title = ma5.group(1).replaceAll(" ", " ").trim(); if (ma5.find()) {
} title = ma5.group(1).replaceAll(" ", " ").trim();
}else if(url.contains("kuaibao")){ }
title = doc.select("p.title").text().replaceAll(" ", ""); }else if(url.contains("kuaibao")){
}else if(url.contains("chinadaily.com.cn")){ title = doc.select("p.title").text().replaceAll(" ", "");
title = doc.select("p.style1").text().replaceAll(" ", ""); }else if(url.contains("chinadaily.com.cn")){
}else if(url.contains("baidu.com") || url.contains("hao123.com")) { title = doc.select("p.style1").text().replaceAll(" ", "");
title = doc.select("p#contaniner").text(); }else if(url.contains("baidu.com") || url.contains("hao123.com")) {
}else if(url.contains("kanfanews.com")) { title = doc.select("p#contaniner").text();
title = doc.select("p#tit").text(); }else if(url.contains("kanfanews.com")) {
}else if(url.contains("ifeng.com") && result.contains("url=http://www.ifeng.com/")) { title = doc.select("p#tit").text();
title = "网页已删除"; }else if(url.contains("ifeng.com") && result.contains("url=http://www.ifeng.com/")) {
}else if(url.contains("a.mp.uc.cn")) { title = "网页已删除";
try { }else if(url.contains("a.mp.uc.cn")) {
JSONObject json = JSONObject.parseObject(result); try {
title = json.getJSONObject("data").getString("title"); JSONObject json = JSONObject.parseObject(result);
} catch (Exception e) { title = json.getJSONObject("data").getString("title");
logger.error(" uc 数据 json 转换失败", e); } catch (Exception e) {
} logger.error(" uc 数据 json 转换失败", e);
}else if(url.contains("huanqiu.com") && result.contains("www.huanqiu.com/404.html")) { }
title = "网页已删除"; }else if(url.contains("huanqiu.com") && result.contains("www.huanqiu.com/404.html")) {
}else if(url.contains("zhihu.com")) { title = "网页已删除";
JSONObject resultJson = JSONObject.parseObject(result); }else if(url.contains("zhihu.com")) {
if(url.contains("/answer/")) { JSONObject resultJson = JSONObject.parseObject(result);
title = resultJson.getJSONObject("question").getString("title"); if(url.contains("/answer/")) {
}else if(url.contains("/question/") && !url.contains("/answer/") || url.contains("/p/")) { title = resultJson.getJSONObject("question").getString("title");
title = resultJson.getString("title"); }else if(url.contains("/question/") && !url.contains("/answer/") || url.contains("/p/")) {
} title = resultJson.getString("title");
}else if(url.contains("360kuai.com") && result.contains("您访问的文章走失了")) { }
title = String.valueOf("404"); }else if(url.contains("360kuai.com") && result.contains("您访问的文章走失了")) {
}else if(result.contains("文章没有找到哦") && url.contains("yidianzixun.com")) { title = String.valueOf("404");
title = "文章未找到"; }else if(result.contains("文章没有找到哦") && url.contains("yidianzixun.com")) {
}else if(url.contains("tznew.58.com/view")) { title = "文章未找到";
try { }else if(url.contains("tznew.58.com/view")) {
JSONObject json = JSONObject.parseObject(result); try {
title = json.getJSONObject("result").getString("title"); JSONObject json = JSONObject.parseObject(result);
} catch (Exception e) { title = json.getJSONObject("result").getString("title");
logger.error(" uc 数据 json 转换失败", e); } catch (Exception e) {
} logger.error(" uc 数据 json 转换失败", e);
} }
}
//若title 为拿到 用 此方法
if(Objects.isNull(title) || title.length() < 1) { //若title 为拿到 用 此方法
title = doc.select("div.adiv > p > span").text().replaceAll(" ", ""); if(Objects.isNull(title) || title.length() < 1) {
} title = doc.select("div.adiv > p > span").text().replaceAll(" ", "");
}
//若title 为拿到 用 此方法
if(Objects.isNull(title) || title.length() < 1) { //若title 为拿到 用 此方法
title = doc.select("title").text().replaceAll(" ", ""); if(Objects.isNull(title) || title.length() < 1) {
} title = doc.select("title").text().replaceAll(" ", "");
}
//若title 为拿到 用 此方法
if(Objects.isNull(title) || title.length() < 1) { //若title 为拿到 用 此方法
title = doc.select("h1").text().replaceAll(" ", ""); if(Objects.isNull(title) || title.length() < 1) {
} title = doc.select("h1").text().replaceAll(" ", "");
}
//若title 为拿到 用 此方法 无法获取标题不进行程序迷惑性判断
// if(Objects.isNull(title) || title.length() < 1 || result.length() < 200) { //若title 为拿到 用 此方法 无法获取标题不进行程序迷惑性判断
// title = "网页已删除"; // if(Objects.isNull(title) || title.length() < 1 || result.length() < 200) {
// } // title = "网页已删除";
// }
if(Objects.nonNull(title) && title.length() > 1){
return new UrlLiveBean(attr.getAttr().toString(), isDelete(title),title); if(Objects.nonNull(title) && title.length() > 1){
} else { return new UrlLiveBean(attr.getAttr().toString(), isDelete(title),title);
return null; } else {
} return null;
} catch (Exception e) { }
return null; } catch (Exception e) {
} return null;
} }
}
/**
* /**
* @Description 标题判断 *
* @param title * @Description 标题判断
* @return * @param title
*/ * @return
private boolean isDelete(String title) { */
List<String> eList = Arrays.asList("系统出错","该内容已被发布者删除","网页已删除" private boolean isDelete(String title) {
,"此帐号已自主注销，内容无法查看","页面提示","正在维护中" List<String> eList = Arrays.asList("系统出错","该内容已被发布者删除","网页已删除"
,"此文章被第三方评估为不实信息","财经头条","知识100题","502BadGateway" ,"此帐号已自主注销，内容无法查看","页面提示","正在维护中"
,"提示信息","跳转页","跳转中...","此帐号在冻结期,内容无法查看","东北新闻网" ,"此文章被第三方评估为不实信息","财经头条","知识100题","502BadGateway"
,"百度一下，你就知道","帐号已迁移","手机百度","内容被删除","亚博国际|首页" ,"提示信息","跳转页","跳转中...","此帐号在冻结期,内容无法查看","东北新闻网"
,"中国软件网","云广网","新浪首页","文章暂时找不到了","-法易网" ,"百度一下，你就知道","帐号已迁移","手机百度","内容被删除","亚博国际|首页"
,"【一点资讯】www.yidianzixun.com","错误页面","网站暂停通知","【快资讯】你的专属资讯平台" ,"中国软件网","云广网","新浪首页","文章暂时找不到了","-法易网"
,"百度新闻——全球最大的中文新闻平台","以上文章由以下机构判定为不实信息","该公众号已迁移" ,"【一点资讯】www.yidianzixun.com","错误页面","网站暂停通知","【快资讯】你的专属资讯平台"
,"财经网-CAIJING.COM.CN","蚂蚁资讯","参数错误","时尚头条_YOKA时尚网","该文章已经被删除" ,"百度新闻——全球最大的中文新闻平台","以上文章由以下机构判定为不实信息","该公众号已迁移"
,"网易","链接已过期","找不到页面","今晚网","该文章已被删除", "该回答已被删除-知乎", "资源不存在","文章未找到" ,"财经网-CAIJING.COM.CN","蚂蚁资讯","参数错误","时尚头条_YOKA时尚网","该文章已经被删除"
, "UC头条"); ,"网易","链接已过期","找不到页面","今晚网","该文章已被删除", "该回答已被删除-知乎", "资源不存在","文章未找到"
, "UC头条");
List<String> cList = Arrays.asList("提示信息-","此内容因违规无法查看","微信公众号不存在"
,"此内容被投诉且经审核涉嫌侵权，无法查看","thepageyourequestedwasnotfound","未知错误" List<String> cList = Arrays.asList("提示信息-","此内容因违规无法查看","微信公众号不存在"
,"Objectmoved","404","页面没有找到","页面未找到","301MovedPermanently","加载异常", ,"此内容被投诉且经审核涉嫌侵权，无法查看","thepageyourequestedwasnotfound","未知错误"
"此帐号已被屏蔽, 内容无法查看","链接不存在", "新闻已删除"); ,"Objectmoved","404","页面没有找到","页面未找到","301MovedPermanently","加载异常",
"此帐号已被屏蔽, 内容无法查看","链接不存在", "新闻已删除");
return cList.stream().anyMatch(title::contains) || eList.stream().anyMatch(title::equals);
} return cList.stream().anyMatch(title::contains) || eList.stream().anyMatch(title::equals);
}
/**
* 处理知乎链接 /**
* * 处理知乎链接
* */ *
private static String treatZhihuUrl(String url) { * */
if(url.contains("/answer/")) { private static String treatZhihuUrl(String url) {
url = "https://api.zhihu.com/answers/" + url.replaceAll(".*/answer/", ""); if(url.contains("/answer/")) {
}else if(url.contains("/question/") && !url.contains("/answer/")) { url = "https://api.zhihu.com/answers/" + url.replaceAll(".*/answer/", "");
url = "https://api.zhihu.com/questions/" + url.replaceAll(".*/question/", ""); }else if(url.contains("/question/") && !url.contains("/answer/")) {
}else if(url.contains("/p/")) { url = "https://api.zhihu.com/questions/" + url.replaceAll(".*/question/", "");
url = "https://api.zhihu.com/articles/" + url.replaceAll(".*/p/", ""); }else if(url.contains("/p/")) {
} url = "https://api.zhihu.com/articles/" + url.replaceAll(".*/p/", "");
return url; }
} return url;
}
}
}
src/main/java/com/zhiwei/source_forward/util/MatchSource.java
...@@ -325,7 +325,6 @@ public class MatchSource { ...@@ -325,7 +325,6 @@ public class MatchSource {
source = source.replaceAll(".*来源：|）", ""); source = source.replaceAll(".*来源：|）", "");
} }
} }
if(Objects.nonNull(source) && source.length() != 0) { if(Objects.nonNull(source) && source.length() != 0) {
return source; return source;
} }
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment