Jsoup学习

Jsoup

jsoup 是一款Java 的HTML解析器，可直接解析某个URL地址、HTML文本内容。它提供了一套非常省力的API，可通过DOM，CSS以及类似于jQuery的操作方法来取出和操作数据。

主要功能:
1. 从一个URL，文件或字符串中解析HTML；
2. 使用DOM或CSS选择器来查找、取出数据；
3. 可操作HTML元素、属性、文本；

        /**
         * 根据属性选择
         */
        public class JsoupDemo02 {
            public static void main(String[] args) throws Exception{
                //新建HttpClient
                CloseableHttpClient httpClient = HttpClients.createDefault();
                HttpGet httpGet = new HttpGet("https://www.cnblogs.com/");
                //返回响应
                CloseableHttpResponse response = null;
                //执行请求
                try {
                    response = httpClient.execute(httpGet);
                    //获取实体
                    HttpEntity entity = response.getEntity();
                    //获取内容
                    String content = EntityUtils.toString(entity,"utf-8");

                    //获取Dom元素
                    Document document = Jsoup.parse(content);
                    //获取tag为标题的元素
                    Elements elementsByTag = document.getElementsByTag("title");
                    //打印标题内容
                    System.out.println("网页第一个标题: "+elementsByTag.get(0).text());

                    //获取网页id为site_nav_top的文档对象
                    Element elementById = document.getElementById("site_nav_top");
                    //打印该id对应的内容
                    System.out.println("id为site_nav_top的内容:"+elementById.html());

                    //获取class为post_item的元素
                    Elements elementsByClass = document.getElementsByClass("post_item");
                    for (Element element : elementsByClass) {
                        System.out.println("class属性为post_item: " + element.html());
                    }

                    //获取属性是height的元素
                    Elements elementsByAttribute = document.getElementsByAttribute("height");
                    for (Element element : elementsByAttribute) {
                        System.out.println("属性为height: " + element.toString());
                    }

                    //获取属性target的值为_blank的元素
                    Elements elementsByAttributeValue = document.getElementsByAttributeValue("target", "_blank");
                    for (Element element : elementsByAttributeValue) {
                        System.out.println("属性target的值为_blank: "+ element.toString());
                    }

                } catch (ClientProtocolException e) {
                    e.printStackTrace();
                } catch (IOException e) {
                    e.printStackTrace();
                } finally {
                    //关闭资源
                    if (response != null) {
                        try {
                            response.close();
                        } catch (IOException e) {
                            e.printStackTrace();
                        }
                    }
                }
                //关闭资源
                httpClient.close();
            }
        }

        /**
         * 根据选择器选择
         */
        public class JsoupDemo03 {

            public static void main(String[] args) throws Exception{
                //新建HttpClient
                CloseableHttpClient httpClient = HttpClients.createDefault();
                HttpGet httpGet = new HttpGet("https://www.cnblogs.com/");
                //返回响应
                CloseableHttpResponse response = null;
                //执行请求
                try {
                    response = httpClient.execute(httpGet);
                    //获取实体
                    HttpEntity entity = response.getEntity();
                    //获取内容
                    String content = EntityUtils.toString(entity,"utf-8");

                    //获取Dom元素
                    Document document = Jsoup.parse(content);

                    //使用id选择器
                    Elements selectElements = document.select("#post_list .post_item .post_item_body h3 a");
                    for (Element element : selectElements) {
                        System.out.println("标题: "+element.text());
                        System.out.println("博客地址: " + element.attr("href"));
                        System.out.println("target: " + element.attr("target"));
                    }

                    System.out.println("------------");

                    //选择带有href属性的a标签元素
                    Elements hrefElements = document.select("a[href]");
                    for (Element element : hrefElements) {
                        System.out.println("a标签的链接: "+element.toString());
                    }

                    //选择扩展名为.png的图片
                    Elements imgElements = document.select("img[src$=.png]");
                    for (Element element : imgElements) {
                        System.out.println("图片: " + element.toString() );
                    }

                    //获取第一个标题
                    Element title = document.getElementsByTag("title").first();
                    System.out.println("网页标题: " + title.text());

                } catch (ClientProtocolException e) {
                    e.printStackTrace();
                } catch (IOException e) {
                    e.printStackTrace();
                } finally {
                    //关闭资源
                    if (response != null) {
                        try {
                            response.close();
                        } catch (IOException e) {
                            e.printStackTrace();
                        }
                    }
                }
                //关闭资源
                httpClient.close();
            }
        }