protected String cleaner(String rs) {
Whitelist wlist = new Whitelist();
wlist.addTags("txt", "p");
wlist.addTags("table", "tbody", "tr", "td");
wlist.addTags("img").addAttributes("img", "src");
return Jsoup.clean(rs, wlist);
}
static {
// 增加可信标签到白名单
content_filter.addTags("embed", "object", "param", "div", "font", "del");
// 增加可信属性
content_filter.addAttributes(":all", "style", "class", "id", "name", "on");
content_filter.addAttributes("object", "width", "height", "classid", "codebase");
content_filter.addAttributes("param", "name", "value");
content_filter.addAttributes(
"embed",
"src",
"quality",
"width",
"height",
"allowFullScreen",
"allowScriptAccess",
"flashvars",
"name",
"type",
"pluginspage");
}
@Test
public void shouldCleanHTMLContentReadability() throws IOException {
// http://trib.al/KsvH2JE
// http://tcrn.ch/1NxNAZJ
String urll = "http://tcrn.ch/1NxNAZJ";
URL url = new URL(urll);
final Readability extractor = new Readability(url, 10000);
extractor.init();
String content = extractor.html();
Whitelist wl = Whitelist.relaxed();
// add additional tags here as necessary
wl.addTags("figure");
String clean = Jsoup.clean(content, wl);
Document mDocument = new Document("");
Element html = mDocument.createElement("html");
Element head = mDocument.createElement("head");
Element body = mDocument.createElement("body");
html.appendChild(head);
body = body.append(clean).select("div").first();
html.appendChild(body);
mDocument.appendChild(html);
// Document document = Jsoup.parse(content);
// Element head = document.head();
// String style = "<STYLE type=\"text/css\">"+
// "blockquote{"+
// " display:block;"+
// " background: #fff;"+
// " padding: 15px 20px 15px 45px;"+
// " margin: 0 0 20px;"+
// " position: relative;"+
// " "+
// " /*Font*/"+
// " font-family: Georgia, serif;"+
// " font-size: 16px;"+
// " line-height: 1.2;"+
// " color: #666;"+
// " text-align: justify;"+
// " "+
// " /*Borders - (Optional)*/"+
// " border-left: 15px solid #76AABA;"+
// " border-right: 2px solid #76AABA;"+
// " "+
// " /*Box Shadow - (Optional)*/"+
// " -moz-box-shadow: 2px 2px 15px #ccc;"+
// " -webkit-box-shadow: 2px 2px 15px #ccc;"+
// " box-shadow: 2px 2px 15px #ccc;"+
// "}"+
// ""+
// "blockquote::before{"+
// " content: \"\\201C\"; /*Unicode for Left Double Quote*/"+
// " "+
// " /*Font*/"+
// " font-family: Georgia, serif;"+
// " font-size: 60px;"+
// " font-weight: bold;"+
// " color: #999;"+
// " "+
// " /*Positioning*/"+
// " position: absolute;"+
// " left: 10px;"+
// " top:5px;"+
// "}"+
// ""+
// "blockquote::after{"+
// " /*Reset to make sure"+
// " content: \"\";*/"+
// " "+
// " content: \"\\201D\"; /*Unicode for Left Double Quote*/"+
// " "+
// " /*Font*/"+
// " font-family: Georgia, serif;"+
// " font-size: 60px;"+
// " font-weight: bold;"+
// " color: #999;"+
// " "+
// " /*Positioning*/"+
// " position: absolute;"+
// " right: 10px;"+
// " bottom:5px;"+
// "}"+
// ""+
// "blockquote a{"+
// " text-decoration: none;"+
// " background: #eee;"+
// " cursor: pointer;"+
// " padding: 0 3px;"+
// " color: #76AABA;"+
// "}"+
// ""+
// "blockquote a:hover{"+
// " color: #666;"+
// "}"+
// ""+
// "blockquote em{"+
// " font-style: italic;"+
// "}"+
// "</STYLE>";
// head.append(style);
//
// content = document.html();
LOG.debug("Article content : {}", mDocument.html());
// LOG.debug("Article content outer: {}", contentOuter);
}
