Html Strip Tags

using System;

using System.IO;

using System.Collections.Generic;

using System.Text;

using System.Text.RegularExpressions;

namespace HtmlStripTags

{

class HtmlHelper

{

private static readonly string[][] htmlNamedEntities = new string[][] {

new string[] { """, "\"" },

new string[] { "<", "<" },

new string[] { ">", ">" },

new string[] { " ", " " },

new string[] { "¡", "¡" },

new string[] { "¢", "¢" },

new string[] { "£", "£" },

new string[] { "¤", "¤" },

new string[] { "¥", "¥" },

new string[] { "¦", "¦" },

new string[] { "§", "§" },

new string[] { "¨", "¨" },

new string[] { "©", "©" },

new string[] { "ª", "ª" },

new string[] { "«", "«" },

new string[] { "¬", "¬" },

new string[] { "", "" },

new string[] { "®", "®" },

new string[] { "¯", "¯" },

new string[] { "°", "°" },

new string[] { "±", "±" },

new string[] { "²", "²" },

new string[] { "³", "³" },

new string[] { "´", "´" },

new string[] { "µ", "µ" },

new string[] { "¶", "¶" },

new string[] { "·", "·" },

new string[] { "¸", "¸" },

new string[] { "¹", "¹" },

new string[] { "º", "º" },

new string[] { "»", " »" },

new string[] { "¼", "¼" },

new string[] { "½", "½" },

new string[] { "¾", "¾" },

new string[] { "¿", "¿" },

new string[] { "À", "À" },

new string[] { "Á", "Á" },

new string[] { "Â", "Â" },

new string[] { "Ã", "Ã" },

new string[] { "Ä", "Ä" },

new string[] { "Å", "Å" },

new string[] { "Æ", "Æ" },

new string[] { "Ç", "Ç" },

new string[] { "È", "È" },

new string[] { "É", "É" },

new string[] { "Ê", "Ê" },

new string[] { "Ë", "Ë" },

new string[] { "Ì", "Ì" },

new string[] { "Í", "Í" },

new string[] { "Î", "Î" },

new string[] { "Ï", "Ï" },

new string[] { "Ð", "Ð" },

new string[] { "Ñ", "Ñ" },

new string[] { "Ò", "Ò" },

new string[] { "Ó", "Ó" },

new string[] { "Ô", "Ô" },

new string[] { "Õ", "Õ" },

new string[] { "Ö", "Ö" },

new string[] { "×", "×" },

new string[] { "Ø", "Ø" },

new string[] { "Ù", "Ù" },

new string[] { "Ú", "Ú" },

new string[] { "Û", "Û" },

new string[] { "Ü", "Ü" },

new string[] { "Ý", "Ý" },

new string[] { "Þ", "Þ" },

new string[] { "ß", "ß" },

new string[] { "à", "à" },

new string[] { "á", "á" },

new string[] { "â", "â" },

new string[] { "ã", "ã" },

new string[] { "ä", "ä" },

new string[] { "å", "å" },

new string[] { "æ", "æ" },

new string[] { "ç", "ç" },

new string[] { "è", "è" },

new string[] { "é", "é" },

new string[] { "ê", "ê" },

new string[] { "ë", "ë" },

new string[] { "ì", "ì" },

new string[] { "í", "í" },

new string[] { "î", "î" },

new string[] { "ï", "ï" },

new string[] { "ð", "ð" },

new string[] { "ñ", "ñ" },

new string[] { "ò", "ò" },

new string[] { "ó", "ó" },

new string[] { "ô", "ô" },

new string[] { "õ", "õ" },

new string[] { "ö", "ö" },

new string[] { "÷", "÷" },

new string[] { "ø", "ø" },

new string[] { "ù", "ù" },

new string[] { "ú", "ú" },

new string[] { "û", "û" },

new string[] { "ü", "ü" },

new string[] { "ý", "ý" },

new string[] { "þ", "þ" },

new string[] { "ÿ", "ÿ" },

new string[] { "&OElig;", "Œ" },

new string[] { "&oelig;", "œ" },

new string[] { "&Scaron;", "Š" },

new string[] { "&scaron;", "š" },

new string[] { "&Yuml;", "Ÿ" },

new string[] { "&fnof;", "ƒ" },

new string[] { "&circ;", "ˆ" },

new string[] { "&tilde;", "˜" },

new string[] { "Α", "Α" },

new string[] { "Β", "Β" },

new string[] { "Γ", "Γ" },

new string[] { "Δ", "Δ" },

new string[] { "Ε", "Ε" },

new string[] { "Ζ", "Ζ" },

new string[] { "Η", "Η" },

new string[] { "Θ", "Θ" },

new string[] { "Ι", "Ι" },

new string[] { "Κ", "Κ" },

new string[] { "Λ", "Λ" },

new string[] { "Μ", "Μ" },

new string[] { "Ν", "Ν" },

new string[] { "Ξ", "Ξ" },

new string[] { "Ο", "Ο" },

new string[] { "Π", "Π" },

new string[] { "Ρ", "Ρ" },

new string[] { "Σ", "Σ" },

new string[] { "Τ", "Τ" },

new string[] { "Υ", "Υ" },

new string[] { "Φ", "Φ" },

new string[] { "Χ", "Χ" },

new string[] { "Ψ", "Ψ" },

new string[] { "Ω", "Ω" },

new string[] { "α", "α" },

new string[] { "β", "β" },

new string[] { "γ", "γ" },

new string[] { "δ", "δ" },

new string[] { "ε", "ε" },

new string[] { "ζ", "ζ" },

new string[] { "η", "η" },

new string[] { "θ", "θ" },

new string[] { "ι", "ι" },

new string[] { "κ", "κ" },

new string[] { "λ", "λ" },

new string[] { "μ", "μ" },

new string[] { "ν", "ν" },

new string[] { "ξ", "ξ" },

new string[] { "ο", "ο" },

new string[] { "π", "π" },

new string[] { "ρ", "ρ" },

new string[] { "&sigmaf;", "ς" },

new string[] { "σ", "σ" },

new string[] { "τ", "τ" },

new string[] { "υ", "υ" },

new string[] { "φ", "φ" },

new string[] { "χ", "χ" },

new string[] { "ψ", "ψ" },

new string[] { "ω", "ω" },

new string[] { "&thetasym;", "ϑ" },

new string[] { "&upsih;", "ϒ" },

new string[] { "ϖ", "ϖ" },

new string[] { "&ensp;", " " },

new string[] { "&emsp;", " " },

new string[] { " ", " " },

new string[] { "&zwnj;", "‌" },

new string[] { "&zwj;", "‍" },

new string[] { "&lrm;", "‎" },

new string[] { "&rlm;", "‏" },

new string[] { "–", "–" },

new string[] { "—", "—" },

new string[] { "‘", "‘" },

new string[] { "’", "’" },

new string[] { "&sbquo;", "‚" },

new string[] { "“", "“" },

new string[] { "”", "”" },

new string[] { "&bdquo;", "„" },

new string[] { "&dagger;", "†" },

new string[] { "&Dagger;", "‡" },

new string[] { "•", "•" },

new string[] { "…", "…" },

new string[] { "&permil;", "‰" },

new string[] { "′", "′" },

new string[] { "″", "″" },

new string[] { "&lsaquo;", "‹" },

new string[] { "&rsaquo;", "›" },

new string[] { "&oline;", "‾" },

new string[] { "&frasl;", "⁄" },

new string[] { "€", "€" },

new string[] { "&image;", "ℑ" },

new string[] { "&weierp;", "℘" },

new string[] { "&real;", "ℜ" },

new string[] { "™", "™" },

new string[] { "&alefsym;", "ℵ" },

new string[] { "←", "←" },

new string[] { "↑", "↑" },

new string[] { "→", "→" },

new string[] { "↓", "↓" },

new string[] { "↔", "↔" },

new string[] { "&crarr;", "↵" },

new string[] { "⇐", "⇐" },

new string[] { "&uArr;", "⇑" },

new string[] { "⇒", "⇒" },

new string[] { "&dArr;", "⇓" },

new string[] { "⇔", "⇔" },

new string[] { "∀", "∀" },

new string[] { "∂", "∂" },

new string[] { "∃", "∃" },

new string[] { "∅", "∅" },

new string[] { "∇", "∇" },

new string[] { "∈", "∈" },

new string[] { "∉", "∉" },

new string[] { "&ni;", "∋" },

new string[] { "∏", "∏" },

new string[] { "∑", "∑" },

new string[] { "−", "−" },

new string[] { "&lowast;", "∗" },

new string[] { "√", "√" },

new string[] { "&prop;", "∝" },

new string[] { "∞", "∞" },

new string[] { "&ang;", "∠" },

new string[] { "&and;", "∧" },

new string[] { "&or;", "∨" },

new string[] { "∩", "∩" },

new string[] { "∪", "∪" },

new string[] { "∫", "∫" },

new string[] { "&there4;", "∴" },

new string[] { "&sim;", "∼" },

new string[] { "&cong;", "≅" },

new string[] { "≈", "≈" },

new string[] { "≠", "≠" },

new string[] { "&equiv;", "≡" },

new string[] { "≤", "≤" },

new string[] { "≥", "≥" },

new string[] { "⊂", "⊂" },

new string[] { "⊃", "⊃" },

new string[] { "&nsub;", "⊄" },

new string[] { "&sube;", "⊆" },

new string[] { "&supe;", "⊇" },

new string[] { "&oplus;", "⊕" },

new string[] { "&otimes;", "⊗" },

new string[] { "&perp;", "⊥" },

new string[] { "⋅", "⋅" },

new string[] { "&lceil;", "⌈" },

new string[] { "&rceil;", "⌉" },

new string[] { "&lfloor;", "⌊" },

new string[] { "&rfloor;", "⌋" },

new string[] { "&lang;", "〈" },

new string[] { "&rang;", "〉" },

new string[] { "&loz;", "◊" },

new string[] { "&spades;", "♠" },

new string[] { "&clubs;", "♣" },

new string[] { "&hearts;", "♥" },

new string[] { "&diams;", "♦" },

new string[] { "&", "&" }

};

public static string HtmlStripTags(string htmlContent, bool replaceNamedEntities, bool replaceNumberedEntities)

{

if (htmlContent == null)

return null;

htmlContent = htmlContent.Trim();

if (htmlContent == string.Empty)

return string.Empty;

int bodyStartTagIdx = htmlContent.IndexOf("<body", StringComparison.CurrentCultureIgnoreCase);

int bodyEndTagIdx = htmlContent.IndexOf("</body>", StringComparison.CurrentCultureIgnoreCase);

int startIdx = 0, endIdx = htmlContent.Length - 1;

if (bodyStartTagIdx >= 0)

startIdx = bodyStartTagIdx;

if (bodyEndTagIdx >= 0)

endIdx = bodyEndTagIdx;

bool insideTag = false,

insideAttributeValue = false,

insideHtmlComment = false,

insideScriptBlock = false,

insideNoScriptBlock = false,

insideStyleBlock = false;

char attributeValueDelimiter = '"';

StringBuilder sb = new StringBuilder(htmlContent.Length);

for (int i = startIdx; i <= endIdx; i++)

{

if (i == 4163)

{

}

// html comment block

if (!insideHtmlComment)

{

if (i + 3 < htmlContent.Length &&

htmlContent[i] == '<' &&

htmlContent[i + 1] == '!' &&

htmlContent[i + 2] == '-' &&

htmlContent[i + 3] == '-')

{

i += 3;

insideHtmlComment = true;

continue;

}

else // inside html comment

{

if (i + 2 < htmlContent.Length &&

htmlContent[i] == '-' &&

htmlContent[i + 1] == '-' &&

htmlContent[i + 2] == '>')

{

i += 2;

insideHtmlComment = false;

continue;

}

else

continue;

}

// noscript block

if (!insideNoScriptBlock)

{

if (i + 9 < htmlContent.Length &&

htmlContent[i] == '<' &&

(htmlContent[i + 1] == 'n' || htmlContent[i + 1] == 'N') &&

(htmlContent[i + 2] == 'o' || htmlContent[i + 2] == 'O') &&

(htmlContent[i + 3] == 's' || htmlContent[i + 3] == 'S') &&

(htmlContent[i + 4] == 'c' || htmlContent[i + 4] == 'C') &&

(htmlContent[i + 5] == 'r' || htmlContent[i + 5] == 'R') &&

(htmlContent[i + 6] == 'i' || htmlContent[i + 6] == 'I') &&

(htmlContent[i + 7] == 'p' || htmlContent[i + 7] == 'P') &&

(htmlContent[i + 8] == 't' || htmlContent[i + 8] == 'T') &&

(char.IsWhiteSpace(htmlContent[i + 9]) || htmlContent[i + 9] == '>'))

{

i += 9;

insideNoScriptBlock = true;

continue;

}

else // inside noscript block

{

if (i + 10 < htmlContent.Length &&

htmlContent[i] == '<' &&

htmlContent[i + 1] == '/' &&

(htmlContent[i + 2] == 'n' || htmlContent[i + 2] == 'N') &&

(htmlContent[i + 3] == 'o' || htmlContent[i + 3] == 'O') &&

(htmlContent[i + 4] == 's' || htmlContent[i + 4] == 'S') &&

(htmlContent[i + 5] == 'c' || htmlContent[i + 5] == 'C') &&

(htmlContent[i + 6] == 'r' || htmlContent[i + 6] == 'R') &&

(htmlContent[i + 7] == 'i' || htmlContent[i + 7] == 'I') &&

(htmlContent[i + 8] == 'p' || htmlContent[i + 8] == 'P') &&

(htmlContent[i + 9] == 't' || htmlContent[i + 9] == 'T') &&

(char.IsWhiteSpace(htmlContent[i + 10]) || htmlContent[i + 10] == '>'))

{

if (htmlContent[i + 10] != '>')

{

i += 9;

while (i < htmlContent.Length && htmlContent[i] != '>')

i++;

}

else

i += 10;

insideNoScriptBlock = false;

}

continue;

}

// script block

if (!insideScriptBlock)

{

if (i + 7 < htmlContent.Length &&

htmlContent[i] == '<' &&

(htmlContent[i + 1] == 's' || htmlContent[i + 1] == 'S') &&

(htmlContent[i + 2] == 'c' || htmlContent[i + 2] == 'C') &&

(htmlContent[i + 3] == 'r' || htmlContent[i + 3] == 'R') &&

(htmlContent[i + 4] == 'i' || htmlContent[i + 4] == 'I') &&

(htmlContent[i + 5] == 'p' || htmlContent[i + 5] == 'P') &&

(htmlContent[i + 6] == 't' || htmlContent[i + 6] == 'T') &&

(char.IsWhiteSpace(htmlContent[i + 7]) || htmlContent[i + 7] == '>'))

{

i += 6;

insideScriptBlock = true;

continue;

}

else // inside script block

{

if (i + 8 < htmlContent.Length &&

htmlContent[i] == '<' &&

htmlContent[i + 1] == '/' &&

(htmlContent[i + 2] == 's' || htmlContent[i + 2] == 'S') &&

(htmlContent[i + 3] == 'c' || htmlContent[i + 3] == 'C') &&

(htmlContent[i + 4] == 'r' || htmlContent[i + 4] == 'R') &&

(htmlContent[i + 5] == 'i' || htmlContent[i + 5] == 'I') &&

(htmlContent[i + 6] == 'p' || htmlContent[i + 6] == 'P') &&

(htmlContent[i + 7] == 't' || htmlContent[i + 7] == 'T') &&

(char.IsWhiteSpace(htmlContent[i + 8]) || htmlContent[i + 8] == '>'))

{

if (htmlContent[i + 8] != '>')

{

i += 7;

while (i < htmlContent.Length && htmlContent[i] != '>')

i++;

}

else

i += 8;

insideScriptBlock = false;

}

continue;

}

// style block

if (!insideStyleBlock)

{

if (i + 7 < htmlContent.Length &&

htmlContent[i] == '<' &&

(htmlContent[i + 1] == 's' || htmlContent[i + 1] == 'S') &&

(htmlContent[i + 2] == 't' || htmlContent[i + 2] == 'T') &&

(htmlContent[i + 3] == 'y' || htmlContent[i + 3] == 'Y') &&

(htmlContent[i + 4] == 'l' || htmlContent[i + 4] == 'L') &&

(htmlContent[i + 5] == 'e' || htmlContent[i + 5] == 'E') &&

(char.IsWhiteSpace(htmlContent[i + 6]) || htmlContent[i + 6] == '>'))

{

i += 5;

insideStyleBlock = true;

continue;

}

else // inside script block

{

if (i + 8 < htmlContent.Length &&

htmlContent[i] == '<' &&

htmlContent[i + 1] == '/' &&

(htmlContent[i + 2] == 's' || htmlContent[i + 2] == 'S') &&

(htmlContent[i + 3] == 't' || htmlContent[i + 3] == 'C') &&

(htmlContent[i + 4] == 'y' || htmlContent[i + 4] == 'R') &&

(htmlContent[i + 5] == 'l' || htmlContent[i + 5] == 'I') &&

(htmlContent[i + 6] == 'e' || htmlContent[i + 6] == 'P') &&

(char.IsWhiteSpace(htmlContent[i + 7]) || htmlContent[i + 7] == '>'))

{

if (htmlContent[i + 7] != '>')

{

i += 7;

while (i < htmlContent.Length && htmlContent[i] != '>')

i++;

}

else

i += 7;

insideStyleBlock = false;

}

continue;

}

if (!insideTag)

{

if (i < htmlContent.Length &&

htmlContent[i] == '<')

{

insideTag = true;

continue;

}

else // inside tag

{

if (!insideAttributeValue)

{

if (htmlContent[i] == '"' || htmlContent[i] == '\'')

{

attributeValueDelimiter = htmlContent[i];

insideAttributeValue = true;

continue;

}

if (htmlContent[i] == '>')

{

insideTag = false;

sb.Append(' '); // prevent words from different tags (<td>s for example) from joining together

continue;

}

else // inside tag and inside attribute value

{

if (htmlContent[i] == attributeValueDelimiter)

{

insideAttributeValue = false;

continue;

}

continue;

}

sb.Append(htmlContent[i]);

}

if (replaceNamedEntities)

foreach (string[] htmlNamedEntity in htmlNamedEntities)

sb.Replace(htmlNamedEntity[0], htmlNamedEntity[1]);

if (replaceNumberedEntities)

for (int i = 0; i < 512; i++)

sb.Replace("&#" + i + ";", ((char)i).ToString());

return sb.ToString();

}

posted on 2007-09-24 23:28 迷你软件阅读(365) 评论(0) 收藏举报