using System;
using System.Text.RegularExpressions;
using System.Web;


/// <summary>
/// Parse html data to produce xml content. Result can be exported to DataSet
/// <para>Parse HtmlTables. Ignore tables that have nested tables</para>
/// </summary>
public class HtmlToXml
{
    //The next Regex comes from Steve (http://blog.stevenlevithan.com/). Very clever, and exactly what I need
    protected const string RegExBase = @"<{0}\b[^>]*>(?:(?>[^<]+)|<(?!{0}\b[^>]*>))*?</{0}>";
    Regex regexTable = null;
    Regex regexTr = null;
    Regex regexTd = null;

    public HtmlToXml()
    {
        regexTable = new Regex(String.Format(RegExBase, "table"), RegexOptions.IgnoreCase | RegexOptions.Multiline | RegexOptions.Compiled);
        regexTr = new Regex(String.Format(RegExBase, "tr"), RegexOptions.IgnoreCase | RegexOptions.Multiline | RegexOptions.Compiled);
        regexTd = new Regex(String.Format(RegExBase, "td"), RegexOptions.IgnoreCase | RegexOptions.Multiline | RegexOptions.Compiled);
    }

    protected string CleanCell(string source)
    {
        Regex regex = new Regex("<(.|\n)+?>");
        source = regex.Replace(source, "");
        source = HttpUtility.HtmlDecode(source);
        return source.Trim();
    }

    protected string ParseCells(string input)
    {
        string result = "";
        int i = 0;
        MatchCollection matches = regexTd.Matches(input);
        //for each cell found
        foreach (Match match in matches)
        {
            i++;
            string cellText = CleanCell(match.Groups[0].Value);
            result += string.Format("<c{0}>{1}</c{0}>", i, HttpUtility.HtmlEncode(cellText));
        }
        return result;
    }

    protected string ParseRows(string source)
    {
        string result = "";
        MatchCollection matches = regexTr.Matches(source);
        //for each row found
        foreach (Match match in matches)
        {
            string cells = ParseCells(match.Groups[0].Value);
            result += "<row>" + cells + "</row>" + System.Environment.NewLine;
        }
        return result;
    }

    public string ParseHtml(string source)
    {
        string result = "";
        int i = 0;
        MatchCollection matches = regexTable.Matches(source);
        //foreach table found, capture its content
        foreach (Match match in matches)
        {
            i++;
            string row = ParseRows(match.Groups[0].Value);
            if (row.Length > 0)
                result += string.Format("<table{0}>{1}{2}</table{0}>{1}", i, System.Environment.NewLine, row);
        }
        return "<DataSet>" + result + "</DataSet>";
    }
}