//
------------
private
static
string
html2TextPattern
=
@"
(?<script><script[^>]*?>.*?</script>)|(?<style><style>.*?</style>)|(?<comment><!--.*?-->)
"
+
@"
|(?<html><[^>]+>)
"
+
//
HTML标记
@"
|(?<quot>&(quot|#34);)
"
+
//
符号: "
@"
|(?<amp>&(amp|#38);)
"
+
//
符号: &
@"
|(?<lt>&(lt|#60);)
"
+
//
符号: <
@"
|(?<gt>&(gt|#62);)
"
+
//
符号: >
@"
|(?<iexcl>&(iexcl|#161);)
"
+
//
符号: (char)161
@"
|(?<cent>&(cent|#162);)
"
+
//
符号: (char)162
@"
|(?<pound>&(pound|#163);)
"
+
//
符号: (char)163
@"
|(?<copy>&(copy|#169);)
"
+
//
符号: (char)169
@"
|(?<others>&(d+);)
"
+
//
符号: 其他
@"
|(?<space> | )
"
;
//
空格
/**/
/// <summary> 
/// 转换HTML为纯文本 /// </summary> /// <param name="html">HTML字符串</param> /// <param name="keepFormat">是否保留换行格式</param> 
/// <returns></returns>
public
static
string
Html2Text(
string
html,
bool
keepFormat)
...
{ string pattern = html2TextPattern; if (! keepFormat) pattern += "|(?<control>[ /s])"; // 换行字符 RegexOptions options = RegexOptions.IgnoreCase | RegexOptions.Singleline | RegexOptions.Compiled; string txt = Regex.Replace(html, pattern, new MatchEvaluator(Html2Text_Match), options); if (! keepFormat) return Regex.Replace(txt.Trim(), "[ ]+", " ", options); // 替换多个连续空格 else return txt; }
private
static
string
Html2Text_Match(System.Text.RegularExpressions.Match m)
...
{ if (m.Groups["quot"].Value != string.Empty) return """; else if (m.Groups["amp"].Value != string.Empty) return "&"; else if (m.Groups["lt"].Value != string.Empty) return "<"; else if (m.Groups["gt"].Value != string.Empty) return ">"; else if (m.Groups["iexcl"].Value != string.Empty) return "¡"; else if (m.Groups["cent"].Value != string.Empty) return "¢"; else if (m.Groups["pound"].Value != string.Empty) return "£"; else if (m.Groups["copy"].Value != string.Empty) return "(c)"; else if (m.Groups["space"].Value != string.Empty) return " "; else if (m.Groups["control"].Value != string.Empty) return " "; else return string.Empty; }
2412
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
