extractors

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

@@ -43,6 +43,8 @@ public class CssSelector implements Selector {
     private String getValue(Element element) {
         if (attrName == null) {
             return element.outerHtml();
+        } else if ("innerHtml".equalsIgnoreCase(attrName)) {
+            return element.html();
         } else {
             return element.attr(attrName);
         }

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

@@ -26,9 +26,9 @@ public class OrSelector implements Selector {
     @Override
     public String select(String text) {
         for (Selector selector : selectors) {
-            text = selector.select(text);
-            if (text != null) {
-                return text;
+            String result = selector.select(text);
+            if (result != null) {
+                return result;
             }
         }
         return null;

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

@@ -20,7 +20,9 @@ public class RegexSelector implements Selector {
 
     private Pattern regex;
 
-    public RegexSelector(String regexStr) {
+    private int group = 1;
+
+    public RegexSelector(String regexStr, int group) {
         if (StringUtils.isBlank(regexStr)) {
             throw new IllegalArgumentException("regex must not be empty");
         }
@@ -36,11 +38,16 @@ public class RegexSelector implements Selector {
         } catch (PatternSyntaxException e) {
             throw new IllegalArgumentException("invalid regex", e);
         }
+        this.group = group;
+    }
+
+    public RegexSelector(String regexStr) {
+        this(regexStr, 1);
     }
 
     @Override
     public String select(String text) {
-        return selectGroup(text).get(1);
+        return selectGroup(text).get(group);
     }
 
     @Override
@@ -48,7 +55,7 @@ public class RegexSelector implements Selector {
         List<String> strings = new ArrayList<String>();
         List<RegexResult> results = selectGroupList(text);
         for (RegexResult result : results) {
-            strings.add(result.get(1));
+            strings.add(result.get(group));
         }
         return strings;
     }

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

@@ -27,7 +27,11 @@ public class SelectorFactory {
     }
 
     public RegexSelector newRegexSelector(String regex, int group) {
-        return newSelector(RegexSelector.class, regex, String.valueOf(group));
+        String cacheKey = getCacheKey(RegexSelector.class, regex, String.valueOf(group));
+        if (innerCache.get(cacheKey) != null) {
+            return (RegexSelector) innerCache.get(cacheKey);
+        }
+        return new RegexSelector(regex, group);
     }
 
     public ReplaceSelector newReplaceSelector(String regex, String replacement) {

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

@@ -0,0 +1,44 @@
+package us.codecraft.webmagic.selector;
+
+/**
+ * Convenient methods for selectors.<br>
+ *
+ * @author code4crafter@gmail.com <br>
+ * @since 0.2.1
+ */
+public abstract class Selectors {
+
+    public static RegexSelector regex(String expr) {
+        return SelectorFactory.getInstatnce().newRegexSelector(expr);
+    }
+
+    public static RegexSelector regex(String expr, int group) {
+        return SelectorFactory.getInstatnce().newRegexSelector(expr, group);
+    }
+
+    public static CssSelector $(String expr) {
+        return new CssSelector(expr);
+    }
+
+    public static CssSelector $(String expr, String attrName) {
+        return new CssSelector(expr, attrName);
+    }
+
+    public static XpathSelector xpath(String expr) {
+        return SelectorFactory.getInstatnce().newXpathSelector(expr);
+    }
+
+    public static AndSelector and(Selector... selectors) {
+        return new AndSelector(selectors);
+    }
+
+    public static OrSelector or(Selector... selectors) {
+        return new OrSelector(selectors);
+    }
+
+    public static void main(String[] args) {
+        String s = "a";
+        or(regex("<title>(.*)</title>"), xpath("//title"), $("title")).select(s);
+    }
+
+}

webmagic-core/src/test/java/us/codecraft/webmagic/selector/ExtractorsTest.java

@@ -0,0 +1,34 @@
+package us.codecraft.webmagic.selector;
+
+import junit.framework.Assert;
+import org.junit.Test;
+
+import static us.codecraft.webmagic.selector.Selectors.*;
+
+/**
+ * @author code4crafter@gmail.com <br>
+ */
+public class ExtractorsTest {
+
+    String html = "<div><h1>test<a href=\"xxx\">aabbcc</a></h1></div>";
+
+    String html2 = "<title>aabbcc</title>";
+
+    @Test
+    public void testEach() {
+        Assert.assertEquals("<a href=\"xxx\">aabbcc</a>", $("div h1 a").select(html));
+        Assert.assertEquals("xxx", $("div h1 a", "href").select(html));
+        Assert.assertEquals("aabbcc", $("div h1 a", "innerHtml").select(html));
+        Assert.assertEquals("xxx", xpath("//a/@href").select(html));
+        Assert.assertEquals("xxx", regex("a href=\"(.*)\"").select(html));
+        Assert.assertEquals("xxx", regex("(a href)=\"(.*)\"", 2).select(html));
+    }
+
+    @Test
+    public void testCombo() {
+        Assert.assertEquals("bb", and($("title"), regex("aa(bb)cc")).select(html2));
+        OrSelector or = or($("div h1 a", "innerHtml"), xpath("//title"));
+        Assert.assertEquals("aabbcc", or.select(html));
+        Assert.assertEquals("aabbcc", or.select(html2));
+    }
+}

webmagic-core/src/test/java/us/codecraft/webmagic/selector/RegexSelectorTest.java

@@ -5,8 +5,6 @@ import org.junit.Test;
 
 /**
  * @author code4crafter@gmail.com <br>
- * Date: 13-4-21
- * Time: 上午7:13
  */
 public class RegexSelectorTest {