add retry

webmagic-core/src/main/java/us/codecraft/webmagic/Request.java

@@ -17,6 +17,8 @@ public class Request implements Serializable {
 
     private static final long serialVersionUID = 2062192774891352043L;
 
+    public static final String CYCLE_TRIED_TIMES = "_cycle_tried_times";
+
     private String url;
 
     /**

webmagic-core/src/main/java/us/codecraft/webmagic/Site.java

@@ -30,6 +30,8 @@ public class Site {
 
     private int retryTimes = 0;
 
+    private int cycleRetryTimes = 0;
+
     private static final Set<Integer> DEFAULT_STATUS_CODE_SET = new HashSet<Integer>();
 
     private Set<Integer> acceptStatCode = DEFAULT_STATUS_CODE_SET;
@@ -200,7 +202,7 @@ public class Site {
     }
 
     /**
-     * Get retry times when download fail, 0 by default.<br>
+     * Get retry times when download fail immediately, 0 by default.<br>
      *
      * @return retry times when download fail
      */
@@ -218,6 +220,25 @@ public class Site {
         return this;
     }
 
+    /**
+     * When cycleRetryTimes is more than 0, it will add back to scheduler and try download again. <br>
+     *
+     * @return retry times when download fail
+     */
+    public int getCycleRetryTimes() {
+        return cycleRetryTimes;
+    }
+
+    /**
+     * Set cycleRetryTimes times when download fail, 0 by default. Only work in RedisScheduler. <br>
+     *
+     * @return this
+     */
+    public Site setCycleRetryTimes(int cycleRetryTimes) {
+        this.cycleRetryTimes = cycleRetryTimes;
+        return this;
+    }
+
     @Override
     public boolean equals(Object o) {
         if (this == o) return true;

webmagic-core/src/main/java/us/codecraft/webmagic/downloader/HttpClientDownloader.java

@@ -52,7 +52,7 @@ public class HttpClientDownloader implements Downloader {
      * @param url
      * @return html
      */
-    public Html download(String url,String charset) {
+    public Html download(String url, String charset) {
         Page page = download(new Request(url), Site.me().setCharset(charset).toTask());
         return (Html) page.getHtml();
     }
@@ -90,6 +90,21 @@ public class HttpClientDownloader implements Downloader {
 
                     if (tried > retryTimes) {
                         logger.warn("download page " + request.getUrl() + " error", e);
+                        if (site.getCycleRetryTimes() > 0) {
+                            Page page = new Page();
+                            Object cycleTriedTimesObject = request.getExtra(Request.CYCLE_TRIED_TIMES);
+                            if (cycleTriedTimesObject == null) {
+                                page.addTargetRequest(request.setPriority(0).putExtra(Request.CYCLE_TRIED_TIMES, 1));
+                            } else {
+                                int cycleTriedTimes = (Integer) cycleTriedTimesObject;
+                                cycleTriedTimes++;
+                                if (cycleTriedTimes >= site.getCycleRetryTimes()) {
+                                    return null;
+                                }
+                                page.addTargetRequest(request.setPriority(0).putExtra(Request.CYCLE_TRIED_TIMES, 1));
+                            }
+                            return page;
+                        }
                         return null;
                     }
                     logger.info("download page " + request.getUrl() + " error, retry the " + tried + " time!");

webmagic-core/src/main/java/us/codecraft/webmagic/selector/Html.java

@@ -1,5 +1,6 @@
 package us.codecraft.webmagic.selector;
 
+import org.apache.log4j.Logger;
 import org.jsoup.Jsoup;
 import org.jsoup.nodes.Document;
 import us.codecraft.webmagic.utils.EnvironmentUtil;
@@ -15,6 +16,8 @@ import java.util.List;
  */
 public class Html extends PlainText {
 
+    private Logger logger = Logger.getLogger(getClass());
+
     /**
      * Store parsed document for better performance when only one text exist.
      */
@@ -26,7 +29,11 @@ public class Html extends PlainText {
 
     public Html(String text) {
         super(text);
-        this.document = Jsoup.parse(text);
+        try {
+            this.document = Jsoup.parse(text);
+        } catch (Exception e) {
+            logger.warn("parse document error ", e);
+        }
     }
 
     public Html(Document document) {
@@ -108,7 +115,7 @@ public class Html extends PlainText {
     }
 
     public String getText() {
-        if (strings!=null&&strings.size()>0){
+        if (strings != null && strings.size() > 0) {
             return strings.get(0);
         }
         return document.html();

webmagic-extension/src/main/java/us/codecraft/webmagic/scheduler/RedisScheduler.java

@@ -36,9 +36,11 @@ public class RedisScheduler implements Scheduler {
     public synchronized void push(Request request, Task task) {
         Jedis jedis = pool.getResource();
         try {
-            //使用Set进行url去重
-            if (!jedis.sismember(SET_PREFIX + task.getUUID(), request.getUrl())) {
-                //使用List保存队列
+            // if cycleRetriedTimes is set, allow duplicated.
+            Object cycleRetriedTimes = request.getExtra(Request.CYCLE_TRIED_TIMES);
+            // use set to remove duplicate url
+            if (cycleRetriedTimes != null || !jedis.sismember(SET_PREFIX + task.getUUID(), request.getUrl())) {
+                // use list to store queue
                 jedis.rpush(QUEUE_PREFIX + task.getUUID(), request.getUrl());
                 jedis.sadd(SET_PREFIX + task.getUUID(), request.getUrl());
                 if (request.getExtras() != null) {