/* * Copyright (c) 2010 Arc90 Inc * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License.
*/
/** * Public constructor. * @param {HTMLDocument} doc The document to parse. * @param {Object} options The options object.
*/ function Readability(doc, options) { // In some older versions, people passed a URI as the first argument. Cope: if (options && options.documentElement) {
doc = options;
options = arguments[2];
} elseif (!doc || !doc.documentElement) { thrownew Error("First argument to Readability constructor should be a document object.");
}
options = options || {};
// Max number of nodes supported by this parser. Default: 0 (no limit)
DEFAULT_MAX_ELEMS_TO_PARSE: 0,
// The number of top candidates to consider when analysing how // tight the competition is among candidates.
DEFAULT_N_TOP_CANDIDATES: 5,
// Element tags to score by default.
DEFAULT_TAGS_TO_SCORE: "section,h2,h3,h4,h5,h6,p,td,pre".toUpperCase().split(","),
// The default number of chars an article must have in order to return a result
DEFAULT_CHAR_THRESHOLD: 500,
// All of the regular expressions in use within readability. // Defined up here so we don't instantiate them repeatedly in loops.
REGEXPS: { // NOTE: These two regular expressions are duplicated in // Readability-readerable.js. Please keep both copies in sync.
unlikelyCandidates: /-ad-|ai2html|banner|breadcrumbs|combx|comment|community|cover-wrap|disqus|extra|footer|gdpr|header|legends|menu|related|remark|replies|rss|shoutbox|sidebar|skyscraper|social|sponsor|supplemental|ad-break|agegate|pagination|pager|popup|yom-remote/i,
okMaybeItsACandidate: /and|article|body|column|content|main|shadow/i,
positive: /article|body|content|entry|hentry|h-entry|main|page|pagination|post|text|blog|story/i,
negative: /-ad-|hidden|^hid$| hid$| hid |^hid |banner|combx|comment|com-|contact|foot|footer|footnote|gdpr|masthead|media|meta|outbrain|promo|related|scroll|share|shoutbox|sidebar|skyscraper|sponsor|shopping|tags|tool|widget/i,
extraneous: /print|archive|comment|discuss|e[\-]?mail|share|reply|all|login|sign|single|utility/i,
byline: /byline|author|dateline|writtenby|p-author/i,
replaceFonts: /<(\/?)font[^>]*>/gi,
normalize: /\s{2,}/g,
videos: /\/\/(www\.)?((dailymotion|youtube|youtube-nocookie|player\.vimeo|v\.qq)\.com|(archive|upload\.wikimedia)\.org|player\.twitch\.tv)/i,
shareElements: /(\b|_)(share|sharedaddy)(\b|_)/i,
nextLink: /(next|weiter|continue|>([^\|]|$)|»([^\|]|$))/i,
prevLink: /(prev|earl|old|new|<|«)/i,
tokenize: /\W+/g,
whitespace: /^\s*$/,
hasContent: /\S$/,
hashUrl: /^#.+/,
srcsetUrl: /(\S+)(\s+[\d.]+[xw])?(\s*(?:,|$))/g,
b64DataUrl: /^data:\s*([^\s;,]+)\s*;\s*base64\s*,/i, // Commas as used in Latin, Sindhi, Chinese and various other scripts. // see: https://en.wikipedia.org/wiki/Comma#Comma_variants
commas: /\u002C|\u060C|\uFE50|\uFE10|\uFE11|\u2E41|\u2E34|\u2E32|\uFF0C/g, // See: https://schema.org/Article
jsonLdArticleTypes: /^Article|AdvertiserContentArticle|NewsArticle|AnalysisNewsArticle|AskPublicNewsArticle|BackgroundNewsArticle|OpinionNewsArticle|ReportageNewsArticle|ReviewNewsArticle|Report|SatiricalArticle|ScholarlyArticle|MedicalScholarlyArticle|SocialMediaPosting|BlogPosting|LiveBlogPosting|DiscussionForumPosting|TechArticle|APIReference$/
},
// The commented out elements qualify as phrasing content but tend to be // removed by readability when put into paragraphs, so we ignore them here.
PHRASING_ELEMS: [ // "CANVAS", "IFRAME", "SVG", "VIDEO", "ABBR", "AUDIO", "B", "BDO", "BR", "BUTTON", "CITE", "CODE", "DATA", "DATALIST", "DFN", "EM", "EMBED", "I", "IMG", "INPUT", "KBD", "LABEL", "MARK", "MATH", "METER", "NOSCRIPT", "OBJECT", "OUTPUT", "PROGRESS", "Q", "RUBY", "SAMP", "SCRIPT", "SELECT", "SMALL", "SPAN", "STRONG", "SUB", "SUP", "TEXTAREA", "TIME", "VAR", "WBR"
],
// These are the classes that readability sets itself.
CLASSES_TO_PRESERVE: [ "page" ],
// These are the list of HTML entities that need to be escaped.
HTML_ESCAPE_MAP: { "lt": "<", "gt": ">", "amp": "&", "quot": '"', "apos": "'",
},
/** * Run any post-process modifications to article content as necessary. * * @param Element * @return void
**/
_postProcessContent: function(articleContent) { // Readability cannot open relative uris so we convert them to absolute uris. this._fixRelativeUris(articleContent);
this._simplifyNestedElements(articleContent);
if (!this._keepClasses) { // Remove classes. this._cleanClasses(articleContent);
}
},
/** * Iterates over a NodeList, calls `filterFn` for each node and removes node * if function returned `true`. * * If function is not passed, removes all the nodes in node list. * * @param NodeList nodeList The nodes to operate on * @param Function filterFn the function to use as a filter * @return void
*/
_removeNodes: function(nodeList, filterFn) { // Avoid ever operating on live node lists. if (this._docJSDOMParser && nodeList._isLiveNodeList) { thrownew Error("Do not pass live node lists to _removeNodes");
} for (var i = nodeList.length - 1; i >= 0; i--) { var node = nodeList[i]; var parentNode = node.parentNode; if (parentNode) { if (!filterFn || filterFn.call(this, node, i, nodeList)) {
parentNode.removeChild(node);
}
}
}
},
/** * Iterates over a NodeList, and calls _setNodeTag for each node. * * @param NodeList nodeList The nodes to operate on * @param String newTagName the new tag name to use * @return void
*/
_replaceNodeTags: function(nodeList, newTagName) { // Avoid ever operating on live node lists. if (this._docJSDOMParser && nodeList._isLiveNodeList) { thrownew Error("Do not pass live node lists to _replaceNodeTags");
} for (const node of nodeList) { this._setNodeTag(node, newTagName);
}
},
/** * Iterate over a NodeList, which doesn't natively fully implement the Array * interface. * * For convenience, the current object context is applied to the provided * iterate function. * * @param NodeList nodeList The NodeList. * @param Function fn The iterate function. * @return void
*/
_forEachNode: function(nodeList, fn) {
Array.prototype.forEach.call(nodeList, fn, this);
},
/** * Iterate over a NodeList, and return the first node that passes * the supplied test function * * For convenience, the current object context is applied to the provided * test function. * * @param NodeList nodeList The NodeList. * @param Function fn The test function. * @return void
*/
_findNode: function(nodeList, fn) { return Array.prototype.find.call(nodeList, fn, this);
},
/** * Iterate over a NodeList, return true if any of the provided iterate * function calls returns true, false otherwise. * * For convenience, the current object context is applied to the * provided iterate function. * * @param NodeList nodeList The NodeList. * @param Function fn The iterate function. * @return Boolean
*/
_someNode: function(nodeList, fn) { return Array.prototype.some.call(nodeList, fn, this);
},
/** * Iterate over a NodeList, return true if all of the provided iterate * function calls return true, false otherwise. * * For convenience, the current object context is applied to the * provided iterate function. * * @param NodeList nodeList The NodeList. * @param Function fn The iterate function. * @return Boolean
*/
_everyNode: function(nodeList, fn) { return Array.prototype.every.call(nodeList, fn, this);
},
/** * Concat all nodelists passed as arguments. * * @return ...NodeList * @return Array
*/
_concatNodeLists: function() { var slice = Array.prototype.slice; var args = slice.call(arguments); var nodeLists = args.map(function(list) { return slice.call(list);
}); return Array.prototype.concat.apply([], nodeLists);
},
/** * Removes the class="" attribute from every element in the given * subtree, except those that match CLASSES_TO_PRESERVE and * the classesToPreserve array from the options object. * * @param Element * @return void
*/
_cleanClasses: function(node) { var classesToPreserve = this._classesToPreserve; var className = (node.getAttribute("class") || "")
.split(/\s+/)
.filter(function(cls) { return classesToPreserve.indexOf(cls) != -1;
})
.join(" ");
if (className) {
node.setAttribute("class", className);
} else {
node.removeAttribute("class");
}
/** * Converts each <a> and <img> uri in the given element to an absolute URI, * ignoring #ref URIs. * * @param Element * @return void
*/
_fixRelativeUris: function(articleContent) { var baseURI = this._doc.baseURI; var documentURI = this._doc.documentURI; function toAbsoluteURI(uri) { // Leave hash links alone if the base URI matches the document URI: if (baseURI == documentURI && uri.charAt(0) == "#") { return uri;
}
// Otherwise, resolve against base URI: try { returnnew URL(uri, baseURI).href;
} catch (ex) { // Something went wrong, just return the original:
} return uri;
}
var links = this._getAllNodesWithTag(articleContent, ["a"]); this._forEachNode(links, function(link) { var href = link.getAttribute("href"); if (href) { // Remove links with javascript: URIs, since // they won't work after scripts have been removed from the page. if (href.indexOf("javascript:") === 0) { // if the link only contains simple text content, it can be converted to a text node if (link.childNodes.length === 1 && link.childNodes[0].nodeType === this.TEXT_NODE) { var text = this._doc.createTextNode(link.textContent);
link.parentNode.replaceChild(text, link);
} else { // if the link has multiple children, they should all be preserved var container = this._doc.createElement("span"); while (link.firstChild) {
container.appendChild(link.firstChild);
}
link.parentNode.replaceChild(container, link);
}
} else {
link.setAttribute("href", toAbsoluteURI(href));
}
}
});
this._forEachNode(medias, function(media) { var src = media.getAttribute("src"); var poster = media.getAttribute("poster"); var srcset = media.getAttribute("srcset");
if (src) {
media.setAttribute("src", toAbsoluteURI(src));
}
if (poster) {
media.setAttribute("poster", toAbsoluteURI(poster));
}
_simplifyNestedElements: function(articleContent) { var node = articleContent;
while (node) { if (node.parentNode && ["DIV", "SECTION"].includes(node.tagName) && !(node.id && node.id.startsWith("readability"))) { if (this._isElementWithoutContent(node)) {
node = this._removeAndGetNext(node); continue;
} elseif (this._hasSingleTagInsideElement(node, "DIV") || this._hasSingleTagInsideElement(node, "SECTION")) { var child = node.children[0]; for (var i = 0; i < node.attributes.length; i++) {
child.setAttribute(node.attributes[i].name, node.attributes[i].value);
}
node.parentNode.replaceChild(child, node);
node = child; continue;
}
}
node = this._getNextNode(node);
}
},
/** * Get the article title as an H1. * * @return string
**/
_getArticleTitle: function() { var doc = this._doc; var curTitle = ""; var origTitle = "";
try {
curTitle = origTitle = doc.title.trim();
// If they had an element with id "title" in their HTML if (typeof curTitle !== "string")
curTitle = origTitle = this._getInnerText(doc.getElementsByTagName("title")[0]);
} catch (e) {/* ignore exceptions setting the title. */}
var titleHadHierarchicalSeparators = false; function wordCount(str) { return str.split(/\s+/).length;
}
// If there's a separator in the title, first remove the final part if ((/ [\|\-\\\/>»] /).test(curTitle)) {
titleHadHierarchicalSeparators = / [\\\/>»] /.test(curTitle);
curTitle = origTitle.replace(/(.*)[\|\-\\\/>»] .*/gi, "$1");
// If the resulting title is too short (3 words or fewer), remove // the first part instead: if (wordCount(curTitle) < 3)
curTitle = origTitle.replace(/[^\|\-\\\/>»]*[\|\-\\\/>»](.*)/gi, "$1");
} elseif (curTitle.indexOf(": ") !== -1) { // Check if we have an heading containing this exact string, so we // could assume it's the full title. var headings = this._concatNodeLists(
doc.getElementsByTagName("h1"),
doc.getElementsByTagName("h2")
); var trimmedTitle = curTitle.trim(); var match = this._someNode(headings, function(heading) { return heading.textContent.trim() === trimmedTitle;
});
// If we don't, let's extract the title out of the original title string. if (!match) {
curTitle = origTitle.substring(origTitle.lastIndexOf(":") + 1);
// If the title is now too short, try the first colon instead: if (wordCount(curTitle) < 3) {
curTitle = origTitle.substring(origTitle.indexOf(":") + 1); // But if we have too many words before the colon there's something weird // with the titles and the H tags so let's just use the original title instead
} elseif (wordCount(origTitle.substr(0, origTitle.indexOf(":"))) > 5) {
curTitle = origTitle;
}
}
} elseif (curTitle.length > 150 || curTitle.length < 15) { var hOnes = doc.getElementsByTagName("h1");
if (hOnes.length === 1)
curTitle = this._getInnerText(hOnes[0]);
}
curTitle = curTitle.trim().replace(this.REGEXPS.normalize, " "); // If we now have 4 words or fewer as our title, and either no // 'hierarchical' separators (\, /, > or ») were found in the original // title or we decreased the number of words by more than 1 word, use // the original title. var curTitleWordCount = wordCount(curTitle); if (curTitleWordCount <= 4 &&
(!titleHadHierarchicalSeparators ||
curTitleWordCount != wordCount(origTitle.replace(/[\|\-\\\/>»]+/g, "")) - 1)) {
curTitle = origTitle;
}
return curTitle;
},
/** * Prepare the HTML document for readability to scrape it. * This includes things like stripping javascript, CSS, and handling terrible markup. * * @return void
**/
_prepDocument: function() { var doc = this._doc;
// Remove all style tags in head this._removeNodes(this._getAllNodesWithTag(doc, ["style"]));
/** * Finds the next node, starting from the given node, and ignoring * whitespace in between. If the given node is an element, the same node is * returned.
*/
_nextNode: function (node) { var next = node; while (next
&& (next.nodeType != this.ELEMENT_NODE)
&& this.REGEXPS.whitespace.test(next.textContent)) {
next = next.nextSibling;
} return next;
},
/** * Replaces 2 or more successive <br> elements with a single <p>. * Whitespace between <br> elements are ignored. For example: * <div>foo<br>bar<br> <br><br>abc</div> * will become: * <div>foo<br>bar<p>abc</p></div>
*/
_replaceBrs: function (elem) { this._forEachNode(this._getAllNodesWithTag(elem, ["br"]), function(br) { var next = br.nextSibling;
// Whether 2 or more <br> elements have been found and replaced with a // <p> block. var replaced = false;
// If we find a <br> chain, remove the <br>s until we hit another node // or non-whitespace. This leaves behind the first <br> in the chain // (which will be replaced with a <p> later). while ((next = this._nextNode(next)) && (next.tagName == "BR")) {
replaced = true; var brSibling = next.nextSibling;
next.parentNode.removeChild(next);
next = brSibling;
}
// If we removed a <br> chain, replace the remaining <br> with a <p>. Add // all sibling nodes as children of the <p> until we hit another <br> // chain. if (replaced) { var p = this._doc.createElement("p");
br.parentNode.replaceChild(p, br);
next = p.nextSibling; while (next) { // If we've hit another <br><br>, we're done adding children to this <p>. if (next.tagName == "BR") { var nextElem = this._nextNode(next.nextSibling); if (nextElem && nextElem.tagName == "BR") break;
}
if (!this._isPhrasingContent(next)) break;
// Otherwise, make this node a child of the new <p>. var sibling = next.nextSibling;
p.appendChild(next);
next = sibling;
}
while (p.lastChild && this._isWhitespace(p.lastChild)) {
p.removeChild(p.lastChild);
}
if (p.parentNode.tagName === "P") this._setNodeTag(p.parentNode, "DIV");
}
});
},
_setNodeTag: function (node, tag) { this.log("_setNodeTag", node, tag); if (this._docJSDOMParser) {
node.localName = tag.toLowerCase();
node.tagName = tag.toUpperCase(); return node;
}
var replacement = node.ownerDocument.createElement(tag); while (node.firstChild) {
replacement.appendChild(node.firstChild);
}
node.parentNode.replaceChild(replacement, node); if (node.readability)
replacement.readability = node.readability;
for (var i = 0; i < node.attributes.length; i++) { try {
replacement.setAttribute(node.attributes[i].name, node.attributes[i].value);
} catch (ex) { /* it's possible for setAttribute() to throw if the attribute name * isn't a valid XML Name. Such attributes can however be parsed from * source in HTML docs, see https://github.com/whatwg/html/issues/4275, * so we can hit them here and then throw. We don't care about such * attributes so we ignore them.
*/
}
} return replacement;
},
/** * Prepare the article node for display. Clean out any inline styles, * iframes, forms, strip extraneous <p> tags, etc. * * @param Element * @return void
**/
_prepArticle: function(articleContent) { this._cleanStyles(articleContent);
// Check for data tables before we continue, to avoid removing items in // those tables, which will often be isolated even though they're // visually linked to other content-ful elements (text, images, etc.). this._markDataTables(articleContent);
this._fixLazyImages(articleContent);
// Clean out junk from the article content this._cleanConditionally(articleContent, "form"); this._cleanConditionally(articleContent, "fieldset"); this._clean(articleContent, "object"); this._clean(articleContent, "embed"); this._clean(articleContent, "footer"); this._clean(articleContent, "link"); this._clean(articleContent, "aside");
// Clean out elements with little content that have "share" in their id/class combinations from final top candidates, // which means we don't remove the top candidates even they have "share".
var shareElementThreshold = this.DEFAULT_CHAR_THRESHOLD;
this._forEachNode(articleContent.children, function (topCandidate) { this._cleanMatchedNodes(topCandidate, function (node, matchString) { returnthis.REGEXPS.shareElements.test(matchString) && node.textContent.length < shareElementThreshold;
});
});
// Do these last as the previous stuff may have removed junk // that will affect these this._cleanConditionally(articleContent, "table"); this._cleanConditionally(articleContent, "ul"); this._cleanConditionally(articleContent, "div");
// replace H1 with H2 as H1 should be only title that is displayed separately this._replaceNodeTags(this._getAllNodesWithTag(articleContent, ["h1"]), "h2");
// Remove extra paragraphs this._removeNodes(this._getAllNodesWithTag(articleContent, ["p"]), function (paragraph) { var imgCount = paragraph.getElementsByTagName("img").length; var embedCount = paragraph.getElementsByTagName("embed").length; var objectCount = paragraph.getElementsByTagName("object").length; // At this point, nasty iframes have been removed, only remain embedded video ones. var iframeCount = paragraph.getElementsByTagName("iframe").length; var totalCount = imgCount + embedCount + objectCount + iframeCount;
this._forEachNode(this._getAllNodesWithTag(articleContent, ["br"]), function(br) { var next = this._nextNode(br.nextSibling); if (next && next.tagName == "P")
br.parentNode.removeChild(br);
});
// Remove single-cell tables this._forEachNode(this._getAllNodesWithTag(articleContent, ["table"]), function(table) { var tbody = this._hasSingleTagInsideElement(table, "TBODY") ? table.firstElementChild : table; if (this._hasSingleTagInsideElement(tbody, "TR")) { var row = tbody.firstElementChild; if (this._hasSingleTagInsideElement(row, "TD")) { var cell = row.firstElementChild;
cell = this._setNodeTag(cell, this._everyNode(cell.childNodes, this._isPhrasingContent) ? "P" : "DIV");
table.parentNode.replaceChild(cell, table);
}
}
});
},
/** * Initialize a node with the readability object. Also checks the * className/id for special names to add to its score. * * @param Element * @return void
**/
_initializeNode: function(node) {
node.readability = {"contentScore": 0};
/** * Traverse the DOM from node to node, starting at the node passed in. * Pass true for the second parameter to indicate this node itself * (and its kids) are going away, and we want the next node over. * * Calling this in a loop will traverse the DOM depth-first.
*/
_getNextNode: function(node, ignoreSelfAndKids) { // First check for kids if those aren't being ignored if (!ignoreSelfAndKids && node.firstElementChild) { return node.firstElementChild;
} // Then for siblings... if (node.nextElementSibling) { return node.nextElementSibling;
} // And finally, move up the parent chain *and* find a sibling // (because this is depth-first traversal, we will have already // seen the parent nodes themselves). do {
node = node.parentNode;
} while (node && !node.nextElementSibling); return node && node.nextElementSibling;
},
// compares second text to first one // 1 = same text, 0 = completely different text // works the way that it splits both texts into words and then finds words that are unique in second text // the result is given by the lower length of unique parts
_textSimilarity: function(textA, textB) { var tokensA = textA.toLowerCase().split(this.REGEXPS.tokenize).filter(Boolean); var tokensB = textB.toLowerCase().split(this.REGEXPS.tokenize).filter(Boolean); if (!tokensA.length || !tokensB.length) { return 0;
} var uniqTokensB = tokensB.filter(token => !tokensA.includes(token)); var distanceB = uniqTokensB.join(" ").length / tokensB.join(" ").length; return 1 - distanceB;
},
_checkByline: function(node, matchString) { if (this._articleByline) { returnfalse;
}
if (node.getAttribute !== undefined) { var rel = node.getAttribute("rel"); var itemprop = node.getAttribute("itemprop");
}
_getNodeAncestors: function(node, maxDepth) {
maxDepth = maxDepth || 0; var i = 0, ancestors = []; while (node.parentNode) {
ancestors.push(node.parentNode); if (maxDepth && ++i === maxDepth) break;
node = node.parentNode;
} return ancestors;
},
/*** * grabArticle - Using a variety of metrics (content score, classname, element types), find the content that is * most likely to be the stuff a user wants to read. Then return it wrapped up in a div. * * @param page a document to run upon. Needs to be a full document, complete with body. * @return Element
**/
_grabArticle: function (page) { this.log("**** grabArticle ****"); var doc = this._doc; var isPaging = page !== null;
page = page ? page : this._doc.body;
// We can't grab an article if we don't have a page! if (!page) { this.log("No body found in document. Abort."); returnnull;
}
var pageCacheHtml = page.innerHTML;
while (true) { this.log("Starting grabArticle loop"); var stripUnlikelyCandidates = this._flagIsActive(this.FLAG_STRIP_UNLIKELYS);
// First, node prepping. Trash nodes that look cruddy (like ones with the // class name "comment", etc), and turn divs into P tags where they have been // used inappropriately (as in, where they contain no other block level elements.) var elementsToScore = []; var node = this._doc.documentElement;
let shouldRemoveTitleHeader = true;
while (node) {
if (node.tagName === "HTML") { this._articleLang = node.getAttribute("lang");
}
// User is not able to see elements applied with both "aria-modal = true" and "role = dialog" if (node.getAttribute("aria-modal") == "true" && node.getAttribute("role") == "dialog") {
node = this._removeAndGetNext(node); continue;
}
// Check to see if this node is a byline, and remove it if it is. if (this._checkByline(node, matchString)) {
node = this._removeAndGetNext(node); continue;
}
if (this.DEFAULT_TAGS_TO_SCORE.indexOf(node.tagName) !== -1) {
elementsToScore.push(node);
}
// Turn all divs that don't have children block level elements into p's if (node.tagName === "DIV") { // Put phrasing content into paragraphs. var p = null; var childNode = node.firstChild; while (childNode) { var nextSibling = childNode.nextSibling; if (this._isPhrasingContent(childNode)) { if (p !== null) {
p.appendChild(childNode);
} elseif (!this._isWhitespace(childNode)) {
p = doc.createElement("p");
node.replaceChild(p, childNode);
p.appendChild(childNode);
}
} elseif (p !== null) { while (p.lastChild && this._isWhitespace(p.lastChild)) {
p.removeChild(p.lastChild);
}
p = null;
}
childNode = nextSibling;
}
// Sites like http://mobile.slate.com encloses each paragraph with a DIV // element. DIVs with only a P element inside and no text content can be // safely converted into plain P elements to avoid confusing the scoring // algorithm with DIVs with are, in practice, paragraphs. if (this._hasSingleTagInsideElement(node, "P") && this._getLinkDensity(node) < 0.25) { var newNode = node.children[0];
node.parentNode.replaceChild(newNode, node);
node = newNode;
elementsToScore.push(node);
} elseif (!this._hasChildBlockElement(node)) {
node = this._setNodeTag(node, "P");
elementsToScore.push(node);
}
}
node = this._getNextNode(node);
}
/** * Loop through all paragraphs, and assign a score to them based on how content-y they look. * Then add their score to their parent node. * * A score is determined by things like number of commas, class names, etc. Maybe eventually link density.
**/ var candidates = []; this._forEachNode(elementsToScore, function(elementToScore) { if (!elementToScore.parentNode || typeof(elementToScore.parentNode.tagName) === "undefined") return;
// If this paragraph is less than 25 characters, don't even count it. var innerText = this._getInnerText(elementToScore); if (innerText.length < 25) return;
// Exclude nodes with no ancestor. var ancestors = this._getNodeAncestors(elementToScore, 5); if (ancestors.length === 0) return;
var contentScore = 0;
// Add a point for the paragraph itself as a base.
contentScore += 1;
// Add points for any commas within this paragraph.
contentScore += innerText.split(this.REGEXPS.commas).length;
// For every 100 characters in this paragraph, add another point. Up to 3 points.
contentScore += Math.min(Math.floor(innerText.length / 100), 3);
// Initialize and score ancestors. this._forEachNode(ancestors, function(ancestor, level) { if (!ancestor.tagName || !ancestor.parentNode || typeof(ancestor.parentNode.tagName) === "undefined") return;
if (typeof(ancestor.readability) === "undefined") { this._initializeNode(ancestor);
candidates.push(ancestor);
}
// After we've calculated scores, loop through all of the possible // candidate nodes we found and find the one with the highest score. var topCandidates = []; for (var c = 0, cl = candidates.length; c < cl; c += 1) { var candidate = candidates[c];
// Scale the final candidates score based on link density. Good content // should have a relatively small link density (5% or less) and be mostly // unaffected by this operation. var candidateScore = candidate.readability.contentScore * (1 - this._getLinkDensity(candidate));
candidate.readability.contentScore = candidateScore;
for (var t = 0; t < this._nbTopCandidates; t++) { var aTopCandidate = topCandidates[t];
if (!aTopCandidate || candidateScore > aTopCandidate.readability.contentScore) {
topCandidates.splice(t, 0, candidate); if (topCandidates.length > this._nbTopCandidates)
topCandidates.pop(); break;
}
}
}
var topCandidate = topCandidates[0] || null; var neededToCreateTopCandidate = false; var parentOfTopCandidate;
// If we still have no top candidate, just use the body as a last resort. // We also have to copy the body node so it is something we can modify. if (topCandidate === null || topCandidate.tagName === "BODY") { // Move all of the page's children into topCandidate
topCandidate = doc.createElement("DIV");
neededToCreateTopCandidate = true; // Move everything (not just elements, also text nodes etc.) into the container // so we even include text directly in the body: while (page.firstChild) { this.log("Moving child out:", page.firstChild);
topCandidate.appendChild(page.firstChild);
}
page.appendChild(topCandidate);
this._initializeNode(topCandidate);
} elseif (topCandidate) { // Find a better top candidate node if it contains (at least three) nodes which belong to `topCandidates` array // and whose scores are quite closed with current `topCandidate` node. var alternativeCandidateAncestors = []; for (var i = 1; i < topCandidates.length; i++) { if (topCandidates[i].readability.contentScore / topCandidate.readability.contentScore >= 0.75) {
alternativeCandidateAncestors.push(this._getNodeAncestors(topCandidates[i]));
}
} var MINIMUM_TOPCANDIDATES = 3; if (alternativeCandidateAncestors.length >= MINIMUM_TOPCANDIDATES) {
parentOfTopCandidate = topCandidate.parentNode; while (parentOfTopCandidate.tagName !== "BODY") { var listsContainingThisAncestor = 0; for (var ancestorIndex = 0; ancestorIndex < alternativeCandidateAncestors.length && listsContainingThisAncestor < MINIMUM_TOPCANDIDATES; ancestorIndex++) {
listsContainingThisAncestor += Number(alternativeCandidateAncestors[ancestorIndex].includes(parentOfTopCandidate));
} if (listsContainingThisAncestor >= MINIMUM_TOPCANDIDATES) {
topCandidate = parentOfTopCandidate; break;
}
parentOfTopCandidate = parentOfTopCandidate.parentNode;
}
} if (!topCandidate.readability) { this._initializeNode(topCandidate);
}
// Because of our bonus system, parents of candidates might have scores // themselves. They get half of the node. There won't be nodes with higher // scores than our topCandidate, but if we see the score going *up* in the first // few steps up the tree, that's a decent sign that there might be more content // lurking in other places that we want to unify in. The sibling stuff // below does some of that - but only if we've looked high enough up the DOM // tree.
parentOfTopCandidate = topCandidate.parentNode; var lastScore = topCandidate.readability.contentScore; // The scores shouldn't get too low. var scoreThreshold = lastScore / 3; while (parentOfTopCandidate.tagName !== "BODY") { if (!parentOfTopCandidate.readability) {
parentOfTopCandidate = parentOfTopCandidate.parentNode; continue;
} var parentScore = parentOfTopCandidate.readability.contentScore; if (parentScore < scoreThreshold) break; if (parentScore > lastScore) { // Alright! We found a better parent to use.
topCandidate = parentOfTopCandidate; break;
}
lastScore = parentOfTopCandidate.readability.contentScore;
parentOfTopCandidate = parentOfTopCandidate.parentNode;
}
// If the top candidate is the only child, use parent instead. This will help sibling // joining logic when adjacent content is actually located in parent's sibling node.
parentOfTopCandidate = topCandidate.parentNode; while (parentOfTopCandidate.tagName != "BODY" && parentOfTopCandidate.children.length == 1) {
topCandidate = parentOfTopCandidate;
parentOfTopCandidate = topCandidate.parentNode;
} if (!topCandidate.readability) { this._initializeNode(topCandidate);
}
}
// Now that we have the top candidate, look through its siblings for content // that might also be related. Things like preambles, content split by ads // that we removed, etc. var articleContent = doc.createElement("DIV"); if (isPaging)
articleContent.id = "readability-content";
var siblingScoreThreshold = Math.max(10, topCandidate.readability.contentScore * 0.2); // Keep potential top candidate's parent node to try to get text direction of it later.
parentOfTopCandidate = topCandidate.parentNode; var siblings = parentOfTopCandidate.children;
for (var s = 0, sl = siblings.length; s < sl; s++) { var sibling = siblings[s]; var append = false;
if (sibling === topCandidate) {
append = true;
} else { var contentBonus = 0;
// Give a bonus if sibling nodes and top candidates have the example same classname if (sibling.className === topCandidate.className && topCandidate.className !== "")
contentBonus += topCandidate.readability.contentScore * 0.2;
if (sibling.readability &&
((sibling.readability.contentScore + contentBonus) >= siblingScoreThreshold)) {
append = true;
} elseif (sibling.nodeName === "P") { var linkDensity = this._getLinkDensity(sibling); var nodeContent = this._getInnerText(sibling); var nodeLength = nodeContent.length;
if (append) { this.log("Appending node:", sibling);
if (this.ALTER_TO_DIV_EXCEPTIONS.indexOf(sibling.nodeName) === -1) { // We have a node that isn't a common block level element, like a form or td tag. // Turn it into a div so it doesn't get filtered out later by accident. this.log("Altering sibling:", sibling, "to div.");
sibling = this._setNodeTag(sibling, "DIV");
}
articleContent.appendChild(sibling); // Fetch children again to make it compatible // with DOM parsers without live collection support.
siblings = parentOfTopCandidate.children; // siblings is a reference to the children array, and // sibling is removed from the array when we call appendChild(). // As a result, we must revisit this index since the nodes // have been shifted.
s -= 1;
sl -= 1;
}
}
if (this._debug) this.log("Article content pre-prep: " + articleContent.innerHTML); // So we have all of the content that we need. Now we clean it up for presentation. this._prepArticle(articleContent); if (this._debug) this.log("Article content post-prep: " + articleContent.innerHTML);
if (neededToCreateTopCandidate) { // We already created a fake div thing, and there wouldn't have been any siblings left // for the previous loop, so there's no point trying to create a new div, and then // move all the children over. Just assign IDs and class names here. No need to append // because that already happened anyway.
topCandidate.id = "readability-page-1";
topCandidate.className = "page";
} else { var div = doc.createElement("DIV");
div.id = "readability-page-1";
div.className = "page"; while (articleContent.firstChild) {
div.appendChild(articleContent.firstChild);
}
articleContent.appendChild(div);
}
if (this._debug) this.log("Article content after paging: " + articleContent.innerHTML);
var parseSuccessful = true;
// Now that we've gone through the full algorithm, check to see if // we got any meaningful content. If we didn't, we may need to re-run // grabArticle with different flags set. This gives us a higher likelihood of // finding the content, and the sieve approach gives us a higher likelihood of // finding the -right- content. var textLength = this._getInnerText(articleContent, true).length; if (textLength < this._charThreshold) {
parseSuccessful = false;
page.innerHTML = pageCacheHtml;
if (this._flagIsActive(this.FLAG_STRIP_UNLIKELYS)) { this._removeFlag(this.FLAG_STRIP_UNLIKELYS); this._attempts.push({articleContent: articleContent, textLength: textLength});
} elseif (this._flagIsActive(this.FLAG_WEIGHT_CLASSES)) { this._removeFlag(this.FLAG_WEIGHT_CLASSES); this._attempts.push({articleContent: articleContent, textLength: textLength});
} elseif (this._flagIsActive(this.FLAG_CLEAN_CONDITIONALLY)) { this._removeFlag(this.FLAG_CLEAN_CONDITIONALLY); this._attempts.push({articleContent: articleContent, textLength: textLength});
} else { this._attempts.push({articleContent: articleContent, textLength: textLength}); // No luck after removing flags, just return the longest text we found during the different loops this._attempts.sort(function (a, b) { return b.textLength - a.textLength;
});
// But first check if we actually have something if (!this._attempts[0].textLength) { returnnull;
}
if (parseSuccessful) { // Find out text direction from ancestors of final top candidate. var ancestors = [parentOfTopCandidate, topCandidate].concat(this._getNodeAncestors(parentOfTopCandidate)); this._someNode(ancestors, function(ancestor) { if (!ancestor.tagName) returnfalse; var articleDir = ancestor.getAttribute("dir"); if (articleDir) { this._articleDir = articleDir; returntrue;
} returnfalse;
}); return articleContent;
}
}
},
/** * Check whether the input string could be a byline. * This verifies that the input is a string, and that the length * is less than 100 chars. * * @param possibleByline {string} - a string to check whether its a byline. * @return Boolean - whether the input string is a byline.
*/
_isValidByline: function(byline) { if (typeof byline == "string" || byline instanceof String) {
byline = byline.trim(); return (byline.length > 0) && (byline.length < 100);
} returnfalse;
},
/** * Converts some of the common HTML entities in string to their corresponding characters. * * @param str {string} - a string to unescape. * @return string without HTML entity.
*/
_unescapeHtmlEntities: function(str) { if (!str) { return str;
}
/** * Try to extract metadata from JSON-LD object. * For now, only Schema.org objects of type Article or its subtypes are supported. * @return Object with any metadata that could be extracted (possibly none)
*/
_getJSONLD: function (doc) { var scripts = this._getAllNodesWithTag(doc, ["script"]);
var metadata;
this._forEachNode(scripts, function(jsonLdElement) { if (!metadata && jsonLdElement.getAttribute("type") === "application/ld+json") { try { // Strip CDATA markers if present var content = jsonLdElement.textContent.replace(/^\s*<!\[CDATA\[|\]\]>\s*$/g, ""); var parsed = JSON.parse(content); if (
!parsed["@context"] ||
!parsed["@context"].match(/^https?\:\/\/schema\.org$/)
) { return;
}
if (typeof parsed.name === "string" && typeof parsed.headline === "string" && parsed.name !== parsed.headline) { // we have both name and headline element in the JSON-LD. They should both be the same but some websites like aktualne.cz // put their own name into "name" and the article title to "headline" which confuses Readability. So we try to check if either // "name" or "headline" closely matches the html title, and if so, use that one. If not, then we use "name" by default.
var title = this._getArticleTitle(); var nameMatches = this._textSimilarity(parsed.name, title) > 0.75; var headlineMatches = this._textSimilarity(parsed.headline, title) > 0.75;
/** * Attempts to get excerpt and byline metadata for the article. * * @param {Object} jsonld — object containing any metadata that * could be extracted from JSON-LD object. * * @return Object with optional "excerpt" and "byline" properties
*/
_getArticleMetadata: function(jsonld) { var metadata = {}; var values = {}; var metaElements = this._doc.getElementsByTagName("meta");
// property is a space-separated list of values var propertyPattern = /\s*(article|dc|dcterm|og|twitter)\s*:\s*(author|creator|description|published_time|title|site_name)\s*/gi;
// name is a single value var namePattern = /^\s*(?:(dc|dcterm|og|twitter|weibo:(article|webpage))\s*[\.:]\s*)?(author|creator|description|title|site_name)\s*$/i;
// Find description tags. this._forEachNode(metaElements, function(element) { var elementName = element.getAttribute("name"); var elementProperty = element.getAttribute("property"); var content = element.getAttribute("content"); if (!content) { return;
} var matches = null; var name = null;
if (elementProperty) {
matches = elementProperty.match(propertyPattern); if (matches) { // Convert to lowercase, and remove any whitespace // so we can match below.
name = matches[0].toLowerCase().replace(/\s/g, ""); // multiple authors
values[name] = content.trim();
}
} if (!matches && elementName && namePattern.test(elementName)) {
name = elementName; if (content) { // Convert to lowercase, remove any whitespace, and convert dots // to colons so we can match below.
name = name.toLowerCase().replace(/\s/g, "").replace(/\./g, ":");
values[name] = content.trim();
}
}
});
// get title
metadata.title = jsonld.title ||
values["dc:title"] ||
values["dcterm:title"] ||
values["og:title"] ||
values["weibo:article:title"] ||
values["weibo:webpage:title"] ||
values["title"] ||
values["twitter:title"];
if (!metadata.title) {
metadata.title = this._getArticleTitle();
}
// get site name
metadata.siteName = jsonld.siteName ||
values["og:site_name"];
// get article published time
metadata.publishedTime = jsonld.datePublished ||
values["article:published_time"] || null;
// in many sites the meta value is escaped with HTML entities, // so here we need to unescape it
metadata.title = this._unescapeHtmlEntities(metadata.title);
metadata.byline = this._unescapeHtmlEntities(metadata.byline);
metadata.excerpt = this._unescapeHtmlEntities(metadata.excerpt);
metadata.siteName = this._unescapeHtmlEntities(metadata.siteName);
metadata.publishedTime = this._unescapeHtmlEntities(metadata.publishedTime);
return metadata;
},
/** * Check if node is image, or if node contains exactly only one image * whether as a direct child or as its descendants. * * @param Element
**/
_isSingleImage: function(node) { if (node.tagName === "IMG") { returntrue;
}
/** * Find all <noscript> that are located after <img> nodes, and which contain only one * <img> element. Replace the first image with the image from inside the <noscript> tag, * and remove the <noscript> tag. This improves the quality of the images we use on * some sites (e.g. Medium). * * @param Element
**/
_unwrapNoscriptImages: function(doc) { // Find img without source or attributes that might contains image, and remove it. // This is done to prevent a placeholder img is replaced by img from noscript in next step. var imgs = Array.from(doc.getElementsByTagName("img")); this._forEachNode(imgs, function(img) { for (var i = 0; i < img.attributes.length; i++) { var attr = img.attributes[i]; switch (attr.name) { case"src": case"srcset": case"data-src": case"data-srcset": return;
}
if (/\.(jpg|jpeg|png|webp)/i.test(attr.value)) { return;
}
}
img.parentNode.removeChild(img);
});
// Next find noscript and try to extract its image var noscripts = Array.from(doc.getElementsByTagName("noscript")); this._forEachNode(noscripts, function(noscript) { // Parse content of noscript and make sure it only contains image var tmp = doc.createElement("div");
tmp.innerHTML = noscript.innerHTML; if (!this._isSingleImage(tmp)) { return;
}
// If noscript has previous sibling and it only contains image, // replace it with noscript content. However we also keep old // attributes that might contains image. var prevElement = noscript.previousElementSibling; if (prevElement && this._isSingleImage(prevElement)) { var prevImg = prevElement; if (prevImg.tagName !== "IMG") {
prevImg = prevElement.getElementsByTagName("img")[0];
}
var newImg = tmp.getElementsByTagName("img")[0]; for (var i = 0; i < prevImg.attributes.length; i++) { var attr = prevImg.attributes[i]; if (attr.value === "") { continue;
}
if (attr.name === "src" || attr.name === "srcset" || /\.(jpg|jpeg|png|webp)/i.test(attr.value)) { if (newImg.getAttribute(attr.name) === attr.value) { continue;
}
var attrName = attr.name; if (newImg.hasAttribute(attrName)) {
attrName = "data-old-" + attrName;
}
/** * Removes script tags from the document. * * @param Element
**/
_removeScripts: function(doc) { this._removeNodes(this._getAllNodesWithTag(doc, ["script", "noscript"]));
},
/** * Check if this node has only whitespace and a single element with given tag * Returns false if the DIV node contains non-empty text nodes * or if it contains no element with given tag or more than 1 element. * * @param Element * @param string tag of child element
**/
_hasSingleTagInsideElement: function(element, tag) { // There should be exactly 1 element child with given tag if (element.children.length != 1 || element.children[0].tagName !== tag) { returnfalse;
}
// And there should be no text nodes with real content return !this._someNode(element.childNodes, function(node) { return node.nodeType === this.TEXT_NODE && this.REGEXPS.hasContent.test(node.textContent);
});
},
/** * Determine whether element has any children block level elements. * * @param Element
*/
_hasChildBlockElement: function (element) { returnthis._someNode(element.childNodes, function(node) { returnthis.DIV_TO_P_ELEMS.has(node.tagName) || this._hasChildBlockElement(node);
});
},
/*** * Determine if a node qualifies as phrasing content. * https://developer.mozilla.org/en-US/docs/Web/Guide/HTML/Content_categories#Phrasing_content
**/
_isPhrasingContent: function(node) { return node.nodeType === this.TEXT_NODE || this.PHRASING_ELEMS.indexOf(node.tagName) !== -1 ||
((node.tagName === "A" || node.tagName === "DEL" || node.tagName === "INS") && this._everyNode(node.childNodes, this._isPhrasingContent));
},
/** * Get the inner text of a node - cross browser compatibly. * This also strips out any excess whitespace to be found. * * @param Element * @param Boolean normalizeSpaces (default: true) * @return string
**/
_getInnerText: function(e, normalizeSpaces) {
normalizeSpaces = (typeof normalizeSpaces === "undefined") ? true : normalizeSpaces; var textContent = e.textContent.trim();
/** * Get the number of times a string s appears in the node e. * * @param Element * @param string - what to split on. Default is "," * @return number (integer)
**/
_getCharCount: function(e, s) {
s = s || ","; returnthis._getInnerText(e).split(s).length - 1;
},
/** * Remove the style attribute on every e and under. * TODO: Test if getElementsByTagName(*) is faster. * * @param Element * @return void
**/
_cleanStyles: function(e) { if (!e || e.tagName.toLowerCase() === "svg") return;
// Remove `style` and deprecated presentational attributes for (var i = 0; i < this.PRESENTATIONAL_ATTRIBUTES.length; i++) {
e.removeAttribute(this.PRESENTATIONAL_ATTRIBUTES[i]);
}
if (this.DEPRECATED_SIZE_ATTRIBUTE_ELEMS.indexOf(e.tagName) !== -1) {
e.removeAttribute("width");
--> --------------------
--> maximum size reached
--> --------------------
Messung V0.5
¤ Dauer der Verarbeitung: 0.21 Sekunden
(vorverarbeitet)
¤
Die Informationen auf dieser Webseite wurden
nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit,
noch Qualität der bereit gestellten Informationen zugesichert.
Bemerkung:
Die farbliche Syntaxdarstellung und die Messung sind noch experimentell.