public void testReplaceHtml()
{
String html = @"<p>
Hello. <font size=""2"">I am size 2</font>
<font color=""red"">and I am red</font>
</p>";
String result = @"<p>
Hello. I am size 2
and I am red
</p>";
Assert.AreEqual(result, RegPattern.ReplaceHtml(html, "font", false));
html = @"<p><span style=""color:red;"">kkk</span> aaaaaaaa</p>";
result = @"<p>kkk aaaaaaaa</p>";
Assert.AreEqual(result, RegPattern.ReplaceHtml(html, "span", false));
html = @"<p>zzz <a href=""ccc.html"">aaa name</a> pppp</p>";
result = @"<p>zzz aaa name pppp</p>";
Assert.AreEqual(result, RegPattern.ReplaceHtml(html, "a", false));
//---------------------------------------------------------------------------------------------
html = @"<p>
<script> alert('sss'); </script>
<style>
p {color:red;}
</style>
Hello. <font size=""2"">I am size 2</font>
<font color=""red"">and I am red</font>
</p>";
result = @"<p>
Hello. <font size=""2"">I am size 2</font>
<font color=""red"">and I am red</font>
</p>";
String x = RegPattern.ReplaceHtml(html, "script", true);
x = RegPattern.ReplaceHtml(x, "style", true);
Assert.AreEqual(result, x);
html = @" <p> pic1 <img src=""eeeeeeee.jpg"" /> </p> pic2<IMG src=""xxxxxx.gif"" > ";
result = @" <p> pic1 </p> pic2 ";
x = RegPattern.ReplaceHtml(html, "img", true);
Assert.AreEqual(result, x);
html = @" <p> pic1 <br/> </p> pic2<BR> <div><br /></div> <br /> ";
result = @" <p> pic1 </p> pic2 <div></div> ";
x = RegPattern.ReplaceHtml(html, "br", true);
Assert.AreEqual(result, x);
}
private string filterPage(string input, SpiderTemplate spiderTemplate)
{
if (strUtil.IsNullOrEmpty(spiderTemplate.DetailClearTag))
{
return(input);
}
String[] arrTag = spiderTemplate.DetailClearTag.ToLower().Split(',');
if (arrTag.Length == 0)
{
return(input);
}
List <String> rTag = new List <String>();
logger.Info("filterTag, input=" + input);
// 过滤标签，以及标签内部的内容
foreach (String tag in arrTag)
{
// font/span/a 只过滤tag，不过滤内容；其他都过滤内容
if (tag == "font" || tag == "span" || tag == "a")
{
rTag.Add(tag);
continue;
}
logger.Info("tag=" + tag);
input = RegPattern.ReplaceHtml(input, tag, true);
}
logger.Info("filterTag, clear tag1=" + input);
// 只过滤标签，不过滤标签的内容
foreach (String tag in rTag)
{
logger.Info("tag=" + tag);
input = RegPattern.ReplaceHtml(input, tag, false);
}
logger.Info("filterTag, clear tag2=" + input);
return(input);
}