我們的應用程序類似,我們有一個富文本編輯器(TinyMCE),我們的輸出是通過iText PDF生成PDF。我們希望HTML儘可能乾淨,理想情況下只使用iText HTMLWorker支持的HTML標籤。 TinyMCE可以做到這一點,但仍然有一些情況,最終用戶可能會提交HTML,這真的搞砸了,這可能會破壞iText生成PDF的能力。
我們使用jSoup和jTidy + CSSParser的組合過濾掉HTML中「style」屬性中輸入的不需要的CSS樣式。輸入到TinyMCE中的HTML會使用此服務進行清理,清理所有來自文字標記的粘貼(如果用戶沒有使用TinyMCE中的從Word粘貼按鈕),並且爲我們提供了可以翻譯iTextPDF HTMLWorker的HTML。
如果表格寬度在style屬性中,HTMLWorker會忽略它並將表格寬度設置爲0,我還發現iText的HTMLWorker解析器(5.0.6)中的表格寬度問題,並將表格寬度設置爲0,所以這是一些邏輯來解決以下問題。我們用下面的庫:一個
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
下面是我們建立擦洗HTML只保留通過iText的+所支持的標記和樣式屬性Groovy的一些服務代碼修復表的問題。代碼中有一些特定於我們的應用程序的假設。目前這對我們非常有用。
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。謝謝回覆。我不是從HTML文檔中獲取文本內容,而是從數據庫中獲取文本內容。當用戶從RTE提交內容時,它首先進入數據庫,然後從數據庫中檢索並用於生成PDF。 – ashishjmeshram 2011-04-20 05:17:22