发一个解析HTML的代码.目前只能解析table与div....

zyug 2012-02-17 04:56:41

代码挺简单的..但是解析的代码一定要配对出现。否则出现错误我不官。至少用来解析baidu搜索结果是没问题的
有志于写解析器的同学可以拿去玩玩





    public class SimpleHtmlParser

    {

        /// <summary>

        /// 解析函数

        /// </summary>

        /// <param name="s">解析字符串</param>

        /// <param name="elements">解析后的控件列表</param>

        /// <returns>返回控件树</returns>

        public static Element ParseHtml(string s,out List<Element> elements)

        {

            elements = new List<Element>();

            elements.Clear();

            Stack<Element> es = new Stack<Element>();

            string pattern = @"(?=(</?table.*?>)|(</?div/?.*?>))";

            RegexOptions options = RegexOptions.None | RegexOptions.IgnoreCase | RegexOptions.Singleline;

            Regex regex = new Regex(pattern, options);

            MatchCollection matches = regex.Matches(s);

            var element = new Element();

            var lastElement = element;

            foreach (Match match in matches)

            {

                var wordindex = 0;

                var wordlength = 0;

                var word = "";

                for (int i = 0; i < match.Groups.Count; i++)

                {

                    var t = match.Groups[i];

                    if (t.Length > 0)

                    {

                        wordindex = t.Index;

                        wordlength = t.Length;

                        word = t.Value;

                        break;

                    }

                }

                if (wordlength <= 0) continue;

                if (word == "<div/>") continue;

                bool isTable = word.IndexOf("table") >= 0;

                bool isDiv = word.IndexOf("div") >= 0;

                bool isEnd = word.IndexOf("</") >= 0;

                if (!isEnd)

                {

                    //新标签

                    Element ee;

                    if (isDiv)

                    {

                        ee = new DivElement();

                    }

                    else if (isTable)

                    {

                        ee = new TableElement();

                    }

                    else

                    {

                        ee = new Element();

                    }

                    ee.StartTagIndex = wordindex;

                    ee.StartTagLength = wordlength;

                    ee.BegTag = word;

                    //设定父级

                    ee.Parent = lastElement;

                    lastElement = ee;

                    ee.Parent.Children.Add(ee);

                    //进栈

                    es.Push(ee);

                }

                else

                {

                    //闭合标签

                    var t = es.Pop();

                    t.EndTag = word;

                    t.EndIndex = wordindex;

                    t.EndTagLength = wordlength;

                    lastElement = t.Parent;

                    t.OuterHtml = s.Substring(t.StartTagIndex, (t.EndIndex - t.StartTagIndex) + t.EndTagLength);

                    t.InnerHtml = s.Substring(t.StartTagIndex + t.StartTagLength, (t.EndIndex - t.StartTagIndex-t.StartTagLength));

                    elements.Add(t);

                }

            }

            return element;

        }



        //去除代码中无用的标签

        public static string ReplaceFontSpan(string s)

        {

            Regex r = new Regex("<head>.*?</head>");

            s = r.Replace(s, "");

            r = new Regex("</?font.*?>");

            s = r.Replace(s, "");

            r = new Regex("</?span.*?>");

            s = r.Replace(s, "");

            r = new Regex("</?a.*?>");

            s = r.Replace(s, "");

            return s;

        }



        //下载网页源文件

        public static string  DownLoadHtml(string url)

        {

            try

            {

                HttpWebRequest r = (HttpWebRequest)WebRequest.Create(url);

                r.Method = "get";

                HttpWebResponse rep = (HttpWebResponse)r.GetResponse();

                Stream receiveStream = rep.GetResponseStream();

                StreamReader readStream = new StreamReader(receiveStream, System.Text.Encoding.Default);

                var result = readStream.ReadToEnd();

                return result.ToString();

            }

            catch

            {

                return "";

            }

        }

    }



    public class Element : StringElement

    {

        

        public int StartTagIndex { get; set; }

        public int StartTagLength {get;set;}



        public int EndIndex { get; set; }

        public int EndTagLength { get; set; }



        public string BegTag { get; set; }

        public string EndTag {get;set;}



        public List<Element> Children = new List<Element>();



        public Element Parent

        {

            get;

            set;

        }

    }



    public class DivElement : Element

    {

        

    }



    public class TableElement : Element

    {

    }



    public class TrElement : Element

    {

 

    }

    

    public class StringElement

    {

        public string OuterHtml

        {

            get;

            set;

        }

        public string InnerHtml

        {

            get;

            set;

        }

    }

调用代码



 private void Form1_Load(object sender, EventArgs e)

        {

            var url = "http://www.baidu.com/s?wd=惠阳妇科病医院&rsv_bp=0&rsv_spt=3&inputT=21000";

            var s = SimpleHtmlParser.DownLoadHtml(url);

            //去除无用标签

            s = SimpleHtmlParser.ReplaceFontSpan(s);

            List<Element> t = null;

            var element = SimpleHtmlParser.ParseHtml(s, out t);

            //查找没有子控件的结果

            foreach (var o in t)

            {

                if (o.Children.Count <= 0)

                {

                    MessageBox.Show(o.OuterHtml);

                }

            }

            //遍历控件树

            List(element);

        }





        public void List(Element e)

        {

         

            if (e.Children.Count > 0)

            {

                foreach (var t in e.Children)

                { 

                    List(t);

                }

            }

            MessageBox.Show(e.OuterHtml); 

        }