我们的应用程序类似，我们有一个富文本编辑器（TinyMCE），我们的输出是通过iText PDF生成PDF。我们希望HTML尽可能干净，理想情况下只使用iText HTMLWorker支持的HTML标签。 TinyMCE可以做到这一点，但仍然有一些情况，最终用户可能会提交HTML，这真的搞砸了，这可能会破坏iText生成PDF的能力。
我们使用jSoup和jTidy + CSSParser的组合过滤掉HTML中“style”属性中输入的不需要的CSS样式。输入到TinyMCE中的HTML会使用此服务进行清理，清理所有来自文字标记的粘贴（如果用户没有使用TinyMCE中的从Word粘贴按钮），并且为我们提供了可以翻译iTextPDF HTMLWorker的HTML。
如果表格宽度在style属性中，HTMLWorker会忽略它并将表格宽度设置为0，我还发现iText的HTMLWorker解析器（5.0.6）中的表格宽度问题，并将表格宽度设置为0，所以这是一些逻辑来解决以下问题。我们用下面的库：一个
com.itextpdf:itextpdf:5.0.6 // used to generate PDFs
org.jsoup:jsoup:1.5.2 // used for cleaning HTML, primary cleaner
net.sf.jtidy:jtidy:r938 // used for cleaning HTML, secondary cleaner
net.sourceforge.cssparser:cssparser:0.9.5 // used to parse out unwanted HTML "style" attribute values
下面是我们建立擦洗HTML只保留通过iText的+所支持的标记和样式属性Groovy的一些服务代码修复表的问题。代码中有一些特定于我们的应用程序的假设。目前这对我们非常有用。
import com.steadystate.css.parser.CSSOMParser
import org.htmlcleaner.CleanerProperties
import org.htmlcleaner.HtmlCleaner;
import org.htmlcleaner.PrettyHtmlSerializer
import org.htmlcleaner.SimpleHtmlSerializer
import org.htmlcleaner.TagNode
import org.jsoup.Jsoup
import org.jsoup.nodes.Document
import org.jsoup.safety.Cleaner
import org.jsoup.safety.Whitelist
import org.jsoup.select.Elements
import org.w3c.css.sac.InputSource
import org.w3c.dom.css.CSSRule
import org.w3c.dom.css.CSSRuleList
import org.w3c.dom.css.CSSStyleDeclaration
import org.w3c.dom.css.CSSStyleSheet
import org.w3c.tidy.Tidy
class HtmlCleanerService {
static transactional = true
def cleanHTML(def html) {
// clean with JSoup which should filter out most unwanted things and
// ensure good html syntax
html = soupClean(html);
// run through JTidy to remove repeated nested tags, clean anything JSoup left out
html = tidyClean(html);
return html;
}
def tidyClean(def html) {
Tidy tidy = new Tidy()
tidy.setAsciiChars(true)
tidy.setDropEmptyParas(true)
tidy.setDropProprietaryAttributes(true)
tidy.setPrintBodyOnly(true)
tidy.setEncloseText(true)
tidy.setJoinStyles(true)
tidy.setLogicalEmphasis(true)
tidy.setQuoteMarks(true)
tidy.setHideComments(true)
tidy.setWraplen(120)
// (makeClean || dropFontTags) = replaces presentational markup by style rules
tidy.setMakeClean(true) // remove presentational clutter.
tidy.setDropFontTags(true)
// word2000 = drop style & class attributes and empty p, span elements
// draconian cleaning for Word2000
tidy.setWord2000(true)
tidy.setMakeBare(true) // remove Microsoft cruft.
tidy.setRepeatedAttributes(org.w3c.tidy.Configuration.KEEP_FIRST) // keep first or last duplicate attribute
// TODO ? tidy.setForceOutput(true)
def reader = new StringReader(html);
def writer = new StringWriter();
// hide output from stderr
tidy.setShowWarnings(false)
tidy.setErrout(new PrintWriter(new StringWriter()))
tidy.parse(reader, writer); // run tidy, providing an input and output stream
return writer.toString()
}
def soupClean(def html) {
// clean the html
Document dirty = Jsoup.parseBodyFragment(html);
Cleaner cleaner = new Cleaner(createWhitelist());
Document clean = cleaner.clean(dirty);
// now hunt down all style attributes and ensure we only have those that render with iTextPDF
Elements styledNodes = clean.select("[style]"); // a with href
styledNodes.each { element ->
def style = element.attr("style");
def tag = element.tagName().toLowerCase()
def newstyle = ""
CSSOMParser parser = new CSSOMParser();
InputSource is = new InputSource(new StringReader(style))
CSSStyleDeclaration styledeclaration = parser.parseStyleDeclaration(is)
boolean hasProps = false
for (int i=0; i < styledeclaration.getLength(); i++) {
def propname = styledeclaration.item(i)
def propval = styledeclaration.getPropertyValue(propname)
propval = propval ? propval.trim() : ""
if (["padding-left", "text-decoration", "text-align", "font-weight", "font-style"].contains(propname)) {
newstyle = newstyle + propname + ": " + propval + ";"
hasProps = true
}
// standardize table widths, itextPDF won't render tables if there is only width in the
// style attribute. Here we ensure the width is in its own attribute, and change the value so
// it is in percentage and no larger than 100% to avoid end users from creating really goofy
// tables that they can't edit properly becuase they have made the width too large.
//
// width of the display area in the editor is about 740px, so let's ensure everything
// is relative to that
//
// TODO could get into trouble with nested tables and widths within as we assume
// one table (e.g. could have nested tables both with widths of 500)
if (tag.equals("table") && propname.equals("width")) {
if (propval.endsWith("%")) {
// ensure it is <= 100%
propval = propval.replaceAll(~"[^0-9]", "")
propval = Math.min(100, propval.toInteger())
}
else {
// else we have measurement in px or assumed px, clean up and
// get integer value, then calculate a percentage
propval = propval.replaceAll(~"[^0-9]", "")
propval = Math.min(100, (int) (propval.toInteger()/740)*100)
}
element.attr("width", propval + "%")
}
}
if (hasProps) {
element.attr("style", newstyle)
} else {
element.removeAttr("style")
}
}
return clean.body().html();
}
/**
* Returns a JSoup whitelist suitable for sane HTML output and iTextPDF
*/
def createWhitelist() {
Whitelist wl = new Whitelist();
// iText supported tags
wl.addTags(
"br", "div", "p", "pre", "span", "blockquote", "q", "hr",
"h1", "h2", "h3", "h4", "h5", "h6",
"u", "strike", "s", "strong", "sub", "sup", "em", "i", "b",
"ul", "ol", "li", "ol",
"table", "tbody", "td", "tfoot", "th", "thead", "tr",
);
// iText attributes recognized which we care about
// padding-left (div/p/span indentation)
// text-align (for table right/left align)
// text-decoration (for span/div/p underline, strikethrough)
// font-weight (for span/div/p bolder etc)
// font-style (for span/div/p italic etc)
// width (for tables)
// colspan/rowspan (for tables)
["span", "div", "p", "table", "ul", "ol", "pre", "td", "th"].each { tag ->
["style", "padding-left", "text-decoration", "text-align", "font-weight", "font-style"].each { attr ->
wl.addAttributes(tag, attr)
}
}
["td", "th"].each { tag ->
["colspan", "rowspan", "width"].each { attr ->
wl.addAttributes(tag, attr)
}
}
wl.addAttributes("table", "width", "style", "cellpadding")
// img support
// wl.addAttributes("img", "align", "alt", "height", "src", "title", "width")
return wl
}
}
嗨。谢谢回复。我不是从HTML文档中获取文本内容，而是从数据库中获取文本内容。当用户从RTE提交内容时，它首先进入数据库，然后从数据库中检索并用于生成PDF。 – ashishjmeshram 2011-04-20 05:17:22