woodyHTML 解析/提取器

woody 是一款基于 Java 的 HTML 解析/提取器，用法非常類似 webmagic，是對其抽取模塊的完全重寫。

功能：

• 多種結(jié)果數(shù)據(jù)類型（String, char, byte, short int, long, double, float, string[], Set, List,Data）
• 支持用戶之定義腳本處理函數(shù)（目前支持 Javascript 函數(shù)配置處理）
• 支持 css、xpath 內(nèi)核替換
• 支持 filter 功能
• 對 css、xpath 內(nèi)核對象的緩存

一個完整的例子：

public class OsChinaBlog {

	public static void main(String[] args) throws Exception {
		Document doc = Jsoup.connect("http://www.oschina.net/news/43879/webmagic-0-3-0").timeout(60000)
				.userAgent("Mozilla/5.0 (Macintosh; Intel Mac OS X 10.8; rv:23.0) Gecko/20100101 Firefox/23.0").get();
		String html = doc.html();
		OsChinaBlogModel model = AnnotationExtractor.me().process(html, OsChinaBlogModel.class);
		System.out.println(model.toJson());
	}

	public static class OsChinaBlogModel extends Model {

		public OsChinaBlogModel() {
			//use to reflect
		}

		@Inject
		@ComboExtract(value = { @ExtractBy(value = "h1.OSCTitle", type = ExprType.CSS),
				@ExtractBy(value = "http://title/text()", type = ExprType.XPATH) }, op = OP.OR)
		public String title;

		@Inject
		@ExtractBy(value = "div.PubDate a[href~=http://my\\.oschina\\.net/]", type = ExprType.CSS)
		public String author;

		@Inject
		@ExtractBy(value = "發(fā)布于.\\s*(\\d+年\\d+月\\d+日)", type = ExprType.REGEX)
		public Date publishDate;

		@Inject
		@ComboExtract(value = {
				@ExtractBy(value = "div.PubDate", type = ExprType.CSS, setting = @Setting(outerHtml = true)),
				@ExtractBy(value = "(\\d+)評", type = ExprType.REGEX) }, op = OP.AND)
		public int commentNum;

		@Inject
		@ExtractBy(value = "span#p_favor_count", type = ExprType.CSS, setting = @Setting(function = @Function(value = "replace", args = {
				"+", "" })))
		public int collectNum;

		@Inject
		@ComboExtract(value = {
				@ExtractBy(value = "div[id=userComments]", type = ExprType.CSS, setting = @Setting(outerHtml = true)),
				@ExtractBy(value = "div.TextContent", type = ExprType.CSS) }, op = OP.AND, multi = true)
		public List commentContents;

		@Inject
		@ExtractBy(value = "div[id=toolbar_wrapper]", setting = @Setting(fliters = { "b", "span" }), type = ExprType.CSS, impl = Document.class)
		public String weibo;

	}
}