add paged support

12 years ago · 619a12b303
parent a5c85c3c8b
commit 619a12b303
8 changed files with 337 additions and 14 deletions
--- a/webmagic-core/src/main/java/us/codecraft/webmagic/Spider.java
+++ b/webmagic-core/src/main/java/us/codecraft/webmagic/Spider.java
@ -283,6 +283,11 @@ public class Spider implements Runnable, Task {
        return this;
    }

+    public Spider clearPipeline(){
+        pipelines=new ArrayList<Pipeline>();
+        return this;
+    }
+
    @Override
    public String getUUID() {
        if (uuid != null) {
--- a/webmagic-core/src/main/java/us/codecraft/webmagic/pipeline/ConsolePipeline.java
+++ b/webmagic-core/src/main/java/us/codecraft/webmagic/pipeline/ConsolePipeline.java
@ -29,7 +29,6 @@ public class ConsolePipeline implements  Pipeline{
            } else {
                System.out.println(entry.getKey() + ":\t" + entry.getValue());
            }
-            System.out.println(entry.getKey()+":\t"+entry.getValue());
        }
    }
 }
--- a/webmagic-core/src/main/java/us/codecraft/webmagic/utils/DoubleKeyMap.java
+++ b/webmagic-core/src/main/java/us/codecraft/webmagic/utils/DoubleKeyMap.java
@ -0,0 +1,111 @@
+package us.codecraft.webmagic.utils;
+
+import java.util.Map;
+
+/**
+ * @author yihua.huang@dianping.com
+ * @date Dec 14, 2012
+ */
+public class DoubleKeyMap<K1, K2, V> extends MultiKeyMapBase {
+    private Map<K1, Map<K2, V>> map;
+
+    public DoubleKeyMap() {
+        init();
+    }
+
+    public DoubleKeyMap(Map<K1, Map<K2, V>> map) {
+        this(map,DEFAULT_CLAZZ);
+    }
+
+    public DoubleKeyMap(Class<? extends Map> protoMapClass) {
+        super(protoMapClass);
+        init();
+    }
+
+    private void init() {
+        if (map == null) {
+            map = this.<K1, Map<K2, V>>newMap();
+        }
+    }
+
+    /**
+     * init map with protoMapClass
+     *
+     * @param protoMapClass
+     */
+    @SuppressWarnings("rawtypes")
+    public DoubleKeyMap(Map<K1, Map<K2, V>> map, Class<? extends Map> protoMapClass) {
+        super(protoMapClass);
+        this.map = map;
+        init();
+    }
+
+    /**
+     * @param key
+     * @return
+     */
+    public Map<K2, V> get(K1 key) {
+        return map.get(key);
+    }
+
+    /**
+     * @param key1
+     * @param key2
+     * @return
+     */
+    public V get(K1 key1, K2 key2) {
+        if (get(key1) == null) {
+            return null;
+        }
+        return get(key1).get(key2);
+    }
+
+
+    /**
+     * @param key1
+     * @param submap
+     * @return
+     */
+    public V put(K1 key1, Map<K2, V> submap) {
+        return put(key1, submap);
+    }
+
+    /**
+     * @param key1
+     * @param key2
+     * @param value
+     * @return
+     */
+    public V put(K1 key1, K2 key2, V value) {
+        if (map.get(key1) == null) {
+            map.put(key1, this.<K2, V>newMap());
+        }
+        return get(key1).put(key2, value);
+    }
+
+    /**
+     * @param key1
+     * @param key2
+     * @return
+     */
+    public V remove(K1 key1, K2 key2) {
+        if (get(key1) == null) {
+            return null;
+        }
+        V remove = get(key1).remove(key2);
+        // 如果上一级map为空，把它也回收掉
+        if (get(key1).size() == 0) {
+            remove(key1);
+        }
+        return remove;
+    }
+
+    /**
+     * @param key1
+     * @return
+     */
+    public Map<K2, V> remove(K1 key1) {
+        Map<K2, V> remove = map.remove(key1);
+        return remove;
+    }
+}
--- a/webmagic-core/src/main/java/us/codecraft/webmagic/utils/MultiKeyMapBase.java
+++ b/webmagic-core/src/main/java/us/codecraft/webmagic/utils/MultiKeyMapBase.java
@ -0,0 +1,42 @@
+package us.codecraft.webmagic.utils;
+
+/**
+ * @author yihua.huang@dianping.com
+ * @date Dec 14, 2012
+ */
+
+import java.util.HashMap;
+import java.util.Map;
+
+/**
+ * multikey map, some basic objects *
+ *
+ * @author yihua.huang
+ */
+public abstract class MultiKeyMapBase {
+
+    protected static final Class<? extends Map> DEFAULT_CLAZZ = HashMap.class;
+    @SuppressWarnings("rawtypes")
+    private Class<? extends Map> protoMapClass = DEFAULT_CLAZZ;
+
+    public MultiKeyMapBase() {
+    }
+
+    @SuppressWarnings("rawtypes")
+    public MultiKeyMapBase(Class<? extends Map> protoMapClass) {
+        this.protoMapClass = protoMapClass;
+    }
+
+    @SuppressWarnings("unchecked")
+    protected <K, V2> Map<K, V2> newMap() {
+        try {
+            return (Map<K, V2>) protoMapClass.newInstance();
+        } catch (InstantiationException e) {
+            throw new IllegalArgumentException("wrong proto type map "
+                    + protoMapClass);
+        } catch (IllegalAccessException e) {
+            throw new IllegalArgumentException("wrong proto type map "
+                    + protoMapClass);
+        }
+    }
+}
--- a/webmagic-core/src/test/java/us/codecraft/webmagic/model/OschinaBlogComment.java
+++ b/webmagic-core/src/test/java/us/codecraft/webmagic/model/OschinaBlogComment.java
@ -1,13 +0,0 @@
-package us.codecraft.webmagic.model;
-
-/**
- * @author code4crafter@gmail.com <br>
- * @date: 13-8-1 <br>
- * Time: 下午10:18 <br>
- */
-@TargetUrl("http://my.oschina.net/flashsword/blog/*")
-public class OschinaBlogComment {
-
-
-
-}
--- a/webmagic-plugin/webmagic-misc/src/main/java/us/codecraft/webmagic/PagedModel.java
+++ b/webmagic-plugin/webmagic-misc/src/main/java/us/codecraft/webmagic/PagedModel.java
@ -0,0 +1,20 @@
+package us.codecraft.webmagic;
+
+import java.util.Collection;
+
+/**
+ * @author yihua.huang@dianping.com <br>
+ * @date: 13-8-4 <br>
+ * Time: 下午5:18 <br>
+ */
+public interface PagedModel {
+
+    public String getPageKey();
+
+    public Collection<String> getOtherPages();
+
+    public String getPage();
+
+    public PagedModel combine(PagedModel pagedModel);
+
+}
--- a/webmagic-plugin/webmagic-misc/src/main/java/us/codecraft/webmagic/pipeline/PagedPipeline.java
+++ b/webmagic-plugin/webmagic-misc/src/main/java/us/codecraft/webmagic/pipeline/PagedPipeline.java
@ -0,0 +1,78 @@
+package us.codecraft.webmagic.pipeline;
+
+import us.codecraft.webmagic.PagedModel;
+import us.codecraft.webmagic.ResultItems;
+import us.codecraft.webmagic.Task;
+import us.codecraft.webmagic.utils.DoubleKeyMap;
+
+import java.util.*;
+import java.util.concurrent.ConcurrentHashMap;
+
+/**
+ * @author yihua.huang@dianping.com <br>
+ * @date: 13-8-4 <br>
+ * Time: 下午5:15 <br>
+ */
+public class PagedPipeline implements Pipeline {
+
+    private DoubleKeyMap<String, String, Boolean> pageMap = new DoubleKeyMap<String, String, Boolean>(ConcurrentHashMap.class);
+
+    private DoubleKeyMap<String, String, PagedModel> objectMap = new DoubleKeyMap<String, String, PagedModel>(ConcurrentHashMap.class);
+
+    @Override
+    public void process(ResultItems resultItems, Task task) {
+        Map<String, Object> resultItemsAll = resultItems.getAll();
+        Iterator<Map.Entry<String, Object>> iterator = resultItemsAll.entrySet().iterator();
+        while (iterator.hasNext()) {
+            handleObject(iterator);
+        }
+    }
+
+    private void handleObject(Iterator<Map.Entry<String, Object>> iterator) {
+        Map.Entry<String, Object> objectEntry = iterator.next();
+        Object o = objectEntry.getValue();
+        if (o instanceof PagedModel) {
+            PagedModel pagedModel = (PagedModel) o;
+            for (String otherPage : pagedModel.getOtherPages()) {
+                Boolean aBoolean = pageMap.get(pagedModel.getPageKey(), otherPage);
+                if (aBoolean == null) {
+                    pageMap.put(pagedModel.getPageKey(), otherPage, Boolean.FALSE);
+                }
+            }
+            //check if all pages are processed
+            Map<String, Boolean> booleanMap = pageMap.get(pagedModel.getPageKey());
+            objectMap.put(pagedModel.getPageKey(), pagedModel.getPage(), pagedModel);
+            if (booleanMap == null) {
+                return;
+            }
+            for (Map.Entry<String, Boolean> stringBooleanEntry : booleanMap.entrySet()) {
+                if (!stringBooleanEntry.getValue()) {
+                    iterator.remove();
+                    return;
+                }
+            }
+            List<Map.Entry<String, PagedModel>> entryList = new ArrayList<Map.Entry<String, PagedModel>>();
+            entryList.addAll(objectMap.get(pagedModel.getPageKey()).entrySet());
+            if (entryList.size() != 0) {
+                Collections.sort(entryList, new Comparator<Map.Entry<String, PagedModel>>() {
+                    @Override
+                    public int compare(Map.Entry<String, PagedModel> o1, Map.Entry<String, PagedModel> o2) {
+                        try {
+                            int i1 = Integer.parseInt(o1.getKey());
+                            int i2 = Integer.parseInt(o2.getKey());
+                            return i1 - i2;
+                        } catch (NumberFormatException e) {
+                            return o1.getKey().compareTo(o2.getKey());
+                        }
+                    }
+                });
+                PagedModel value = entryList.get(0).getValue();
+                for (int i=1;i<entryList.size();i++){
+                    value=value.combine(entryList.get(i).getValue());
+                }
+                objectEntry.setValue(value);
+            }
+        }
+    }
+
+}
--- a/webmagic-samples/src/main/java/us/codecraft/webmagic/model/samples/News163.java
+++ b/webmagic-samples/src/main/java/us/codecraft/webmagic/model/samples/News163.java
@ -0,0 +1,81 @@
+package us.codecraft.webmagic.model.samples;
+
+import us.codecraft.webmagic.Page;
+import us.codecraft.webmagic.PagedModel;
+import us.codecraft.webmagic.Site;
+import us.codecraft.webmagic.model.*;
+import us.codecraft.webmagic.pipeline.ConsolePipeline;
+import us.codecraft.webmagic.pipeline.PagedPipeline;
+import us.codecraft.webmagic.selector.Selectable;
+
+import java.util.Collection;
+import java.util.List;
+
+/**
+ * @author yihua.huang@dianping.com <br>
+ * @date: 13-8-4 <br>
+ * Time: 下午8:17 <br>
+ */
+@TargetUrl("http://news.163.com/\\d+/\\d+/\\d+/\\w+*.html")
+public class News163 implements PagedModel, AfterExtractor {
+
+    @ExtractByUrl("http://news\\.163\\.com/\\d+/\\d+/\\d+/(\\w+)*\\.html")
+    private String pageKey;
+
+    @ExtractByUrl(value = "http://news\\.163\\.com/\\d+/\\d+/\\d+/\\w+_(\\d+)\\.html", notNull = false)
+    private String page;
+
+    private List<String> otherPage;
+
+    @ExtractBy("//h1[@id=\"h1title\"]/text()")
+    private String title;
+
+    @ExtractBy("//div[@id=\"epContentLeft\"]")
+    private String content;
+
+    @Override
+    public String getPageKey() {
+        return pageKey;
+    }
+
+    @Override
+    public Collection<String> getOtherPages() {
+        return otherPage;
+    }
+
+    @Override
+    public String getPage() {
+        if (page == null) {
+            return "0";
+        }
+        return page;
+    }
+
+    @Override
+    public PagedModel combine(PagedModel pagedModel) {
+        News163 news163 = new News163();
+        News163 pagedModel1 = (News163) pagedModel;
+        news163.content = this.content + pagedModel1.content;
+        return news163;
+    }
+
+    @Override
+    public String toString() {
+        return "News163{" +
+                "content='" + content + '\'' +
+                ", title='" + title + '\'' +
+                ", otherPage=" + otherPage +
+                '}';
+    }
+
+    public static void main(String[] args) {
+        OOSpider.create(Site.me().addStartUrl("http://news.163.com/13/0802/05/958I1E330001124J_2.html"), News163.class)
+                .clearPipeline().pipeline(new PagedPipeline()).pipeline(new ConsolePipeline()).run();
+    }
+
+    @Override
+    public void afterProcess(Page page) {
+        Selectable xpath = page.getHtml().xpath("//div[@class=\"ep-pages\"]//a/@href");
+        otherPage = xpath.regex("http://news\\.163\\.com/\\d+/\\d+/\\d+/\\w+_(\\d+)\\.html").all();
+    }
+}