and ![]()
uri in the given element to an absolute URI,\n * ignoring #ref URIs.\n *\n * @param Element\n * @return void\n */\n _fixRelativeUris: function(articleContent) {\n var baseURI = this._doc.baseURI;\n var documentURI = this._doc.documentURI;\n function toAbsoluteURI(uri) {\n // Leave hash links alone if the base URI matches the document URI:\n if (baseURI == documentURI && uri.charAt(0) == \"#\") {\n return uri;\n }\n\n // Otherwise, resolve against base URI:\n try {\n return new URL(uri, baseURI).href;\n } catch (ex) {\n // Something went wrong, just return the original:\n }\n return uri;\n }\n\n var links = this._getAllNodesWithTag(articleContent, [\"a\"]);\n this._forEachNode(links, function(link) {\n var href = link.getAttribute(\"href\");\n if (href) {\n // Remove links with javascript: URIs, since\n // they won't work after scripts have been removed from the page.\n if (href.indexOf(\"javascript:\") === 0) {\n // if the link only contains simple text content, it can be converted to a text node\n if (link.childNodes.length === 1 && link.childNodes[0].nodeType === this.TEXT_NODE) {\n var text = this._doc.createTextNode(link.textContent);\n link.parentNode.replaceChild(text, link);\n } else {\n // if the link has multiple children, they should all be preserved\n var container = this._doc.createElement(\"span\");\n while (link.firstChild) {\n container.appendChild(link.firstChild);\n }\n link.parentNode.replaceChild(container, link);\n }\n } else {\n link.setAttribute(\"href\", toAbsoluteURI(href));\n }\n }\n });\n\n var medias = this._getAllNodesWithTag(articleContent, [\n \"img\", \"picture\", \"figure\", \"video\", \"audio\", \"source\"\n ]);\n\n this._forEachNode(medias, function(media) {\n var src = media.getAttribute(\"src\");\n var poster = media.getAttribute(\"poster\");\n var srcset = media.getAttribute(\"srcset\");\n\n if (src) {\n media.setAttribute(\"src\", toAbsoluteURI(src));\n }\n\n if (poster) {\n media.setAttribute(\"poster\", toAbsoluteURI(poster));\n }\n\n if (srcset) {\n var newSrcset = srcset.replace(this.REGEXPS.srcsetUrl, function(_, p1, p2, p3) {\n return toAbsoluteURI(p1) + (p2 || \"\") + p3;\n });\n\n media.setAttribute(\"srcset\", newSrcset);\n }\n });\n },\n\n _simplifyNestedElements: function(articleContent) {\n var node = articleContent;\n\n while (node) {\n if (node.parentNode && [\"DIV\", \"SECTION\"].includes(node.tagName) && !(node.id && node.id.startsWith(\"readability\"))) {\n if (this._isElementWithoutContent(node)) {\n node = this._removeAndGetNext(node);\n continue;\n } else if (this._hasSingleTagInsideElement(node, \"DIV\") || this._hasSingleTagInsideElement(node, \"SECTION\")) {\n var child = node.children[0];\n for (var i = 0; i < node.attributes.length; i++) {\n child.setAttribute(node.attributes[i].name, node.attributes[i].value);\n }\n node.parentNode.replaceChild(child, node);\n node = child;\n continue;\n }\n }\n\n node = this._getNextNode(node);\n }\n },\n\n /**\n * Get the article title as an H1.\n *\n * @return string\n **/\n _getArticleTitle: function() {\n var doc = this._doc;\n var curTitle = \"\";\n var origTitle = \"\";\n\n try {\n curTitle = origTitle = doc.title.trim();\n\n // If they had an element with id \"title\" in their HTML\n if (typeof curTitle !== \"string\")\n curTitle = origTitle = this._getInnerText(doc.getElementsByTagName(\"title\")[0]);\n } catch (e) {/* ignore exceptions setting the title. */}\n\n var titleHadHierarchicalSeparators = false;\n function wordCount(str) {\n return str.split(/\\s+/).length;\n }\n\n // If there's a separator in the title, first remove the final part\n if ((/ [\\|\\-\\\\\\/>»] /).test(curTitle)) {\n titleHadHierarchicalSeparators = / [\\\\\\/>»] /.test(curTitle);\n curTitle = origTitle.replace(/(.*)[\\|\\-\\\\\\/>»] .*/gi, \"$1\");\n\n // If the resulting title is too short (3 words or fewer), remove\n // the first part instead:\n if (wordCount(curTitle) < 3)\n curTitle = origTitle.replace(/[^\\|\\-\\\\\\/>»]*[\\|\\-\\\\\\/>»](.*)/gi, \"$1\");\n } else if (curTitle.indexOf(\": \") !== -1) {\n // Check if we have an heading containing this exact string, so we\n // could assume it's the full title.\n var headings = this._concatNodeLists(\n doc.getElementsByTagName(\"h1\"),\n doc.getElementsByTagName(\"h2\")\n );\n var trimmedTitle = curTitle.trim();\n var match = this._someNode(headings, function(heading) {\n return heading.textContent.trim() === trimmedTitle;\n });\n\n // If we don't, let's extract the title out of the original title string.\n if (!match) {\n curTitle = origTitle.substring(origTitle.lastIndexOf(\":\") + 1);\n\n // If the title is now too short, try the first colon instead:\n if (wordCount(curTitle) < 3) {\n curTitle = origTitle.substring(origTitle.indexOf(\":\") + 1);\n // But if we have too many words before the colon there's something weird\n // with the titles and the H tags so let's just use the original title instead\n } else if (wordCount(origTitle.substr(0, origTitle.indexOf(\":\"))) > 5) {\n curTitle = origTitle;\n }\n }\n } else if (curTitle.length > 150 || curTitle.length < 15) {\n var hOnes = doc.getElementsByTagName(\"h1\");\n\n if (hOnes.length === 1)\n curTitle = this._getInnerText(hOnes[0]);\n }\n\n curTitle = curTitle.trim().replace(this.REGEXPS.normalize, \" \");\n // If we now have 4 words or fewer as our title, and either no\n // 'hierarchical' separators (\\, /, > or ») were found in the original\n // title or we decreased the number of words by more than 1 word, use\n // the original title.\n var curTitleWordCount = wordCount(curTitle);\n if (curTitleWordCount <= 4 &&\n (!titleHadHierarchicalSeparators ||\n curTitleWordCount != wordCount(origTitle.replace(/[\\|\\-\\\\\\/>»]+/g, \"\")) - 1)) {\n curTitle = origTitle;\n }\n\n return curTitle;\n },\n\n /**\n * Prepare the HTML document for readability to scrape it.\n * This includes things like stripping javascript, CSS, and handling terrible markup.\n *\n * @return void\n **/\n _prepDocument: function() {\n var doc = this._doc;\n\n // Remove all style tags in head\n this._removeNodes(this._getAllNodesWithTag(doc, [\"style\"]));\n\n if (doc.body) {\n this._replaceBrs(doc.body);\n }\n\n this._replaceNodeTags(this._getAllNodesWithTag(doc, [\"font\"]), \"SPAN\");\n },\n\n /**\n * Finds the next node, starting from the given node, and ignoring\n * whitespace in between. If the given node is an element, the same node is\n * returned.\n */\n _nextNode: function (node) {\n var next = node;\n while (next\n && (next.nodeType != this.ELEMENT_NODE)\n && this.REGEXPS.whitespace.test(next.textContent)) {\n next = next.nextSibling;\n }\n return next;\n },\n\n /**\n * Replaces 2 or more successive
elements with a single .\n * Whitespace between
elements are ignored. For example:\n *
foo
bar
abc
\n * will become:\n * \n */\n _replaceBrs: function (elem) {\n this._forEachNode(this._getAllNodesWithTag(elem, [\"br\"]), function(br) {\n var next = br.nextSibling;\n\n // Whether 2 or more
elements have been found and replaced with a\n // block.\n var replaced = false;\n\n // If we find a
chain, remove the
s until we hit another node\n // or non-whitespace. This leaves behind the first
in the chain\n // (which will be replaced with a
later).\n while ((next = this._nextNode(next)) && (next.tagName == \"BR\")) {\n replaced = true;\n var brSibling = next.nextSibling;\n next.parentNode.removeChild(next);\n next = brSibling;\n }\n\n // If we removed a
chain, replace the remaining
with a
. Add\n // all sibling nodes as children of the
until we hit another
\n // chain.\n if (replaced) {\n var p = this._doc.createElement(\"p\");\n br.parentNode.replaceChild(p, br);\n\n next = p.nextSibling;\n while (next) {\n // If we've hit another
, we're done adding children to this
.\n if (next.tagName == \"BR\") {\n var nextElem = this._nextNode(next.nextSibling);\n if (nextElem && nextElem.tagName == \"BR\")\n break;\n }\n\n if (!this._isPhrasingContent(next))\n break;\n\n // Otherwise, make this node a child of the new
.\n var sibling = next.nextSibling;\n p.appendChild(next);\n next = sibling;\n }\n\n while (p.lastChild && this._isWhitespace(p.lastChild)) {\n p.removeChild(p.lastChild);\n }\n\n if (p.parentNode.tagName === \"P\")\n this._setNodeTag(p.parentNode, \"DIV\");\n }\n });\n },\n\n _setNodeTag: function (node, tag) {\n this.log(\"_setNodeTag\", node, tag);\n if (this._docJSDOMParser) {\n node.localName = tag.toLowerCase();\n node.tagName = tag.toUpperCase();\n return node;\n }\n\n var replacement = node.ownerDocument.createElement(tag);\n while (node.firstChild) {\n replacement.appendChild(node.firstChild);\n }\n node.parentNode.replaceChild(replacement, node);\n if (node.readability)\n replacement.readability = node.readability;\n\n for (var i = 0; i < node.attributes.length; i++) {\n try {\n replacement.setAttribute(node.attributes[i].name, node.attributes[i].value);\n } catch (ex) {\n /* it's possible for setAttribute() to throw if the attribute name\n * isn't a valid XML Name. Such attributes can however be parsed from\n * source in HTML docs, see https://github.com/whatwg/html/issues/4275,\n * so we can hit them here and then throw. We don't care about such\n * attributes so we ignore them.\n */\n }\n }\n return replacement;\n },\n\n /**\n * Prepare the article node for display. Clean out any inline styles,\n * iframes, forms, strip extraneous
tags, etc.\n *\n * @param Element\n * @return void\n **/\n _prepArticle: function(articleContent) {\n this._cleanStyles(articleContent);\n\n // Check for data tables before we continue, to avoid removing items in\n // those tables, which will often be isolated even though they're\n // visually linked to other content-ful elements (text, images, etc.).\n this._markDataTables(articleContent);\n\n this._fixLazyImages(articleContent);\n\n // Clean out junk from the article content\n this._cleanConditionally(articleContent, \"form\");\n this._cleanConditionally(articleContent, \"fieldset\");\n this._clean(articleContent, \"object\");\n this._clean(articleContent, \"embed\");\n this._clean(articleContent, \"footer\");\n this._clean(articleContent, \"link\");\n this._clean(articleContent, \"aside\");\n\n // Clean out elements with little content that have \"share\" in their id/class combinations from final top candidates,\n // which means we don't remove the top candidates even they have \"share\".\n\n var shareElementThreshold = this.DEFAULT_CHAR_THRESHOLD;\n\n this._forEachNode(articleContent.children, function (topCandidate) {\n this._cleanMatchedNodes(topCandidate, function (node, matchString) {\n return this.REGEXPS.shareElements.test(matchString) && node.textContent.length < shareElementThreshold;\n });\n });\n\n this._clean(articleContent, \"iframe\");\n this._clean(articleContent, \"input\");\n this._clean(articleContent, \"textarea\");\n this._clean(articleContent, \"select\");\n this._clean(articleContent, \"button\");\n this._cleanHeaders(articleContent);\n\n // Do these last as the previous stuff may have removed junk\n // that will affect these\n this._cleanConditionally(articleContent, \"table\");\n this._cleanConditionally(articleContent, \"ul\");\n this._cleanConditionally(articleContent, \"div\");\n\n // replace H1 with H2 as H1 should be only title that is displayed separately\n this._replaceNodeTags(this._getAllNodesWithTag(articleContent, [\"h1\"]), \"h2\");\n\n // Remove extra paragraphs\n this._removeNodes(this._getAllNodesWithTag(articleContent, [\"p\"]), function (paragraph) {\n var imgCount = paragraph.getElementsByTagName(\"img\").length;\n var embedCount = paragraph.getElementsByTagName(\"embed\").length;\n var objectCount = paragraph.getElementsByTagName(\"object\").length;\n // At this point, nasty iframes have been removed, only remain embedded video ones.\n var iframeCount = paragraph.getElementsByTagName(\"iframe\").length;\n var totalCount = imgCount + embedCount + objectCount + iframeCount;\n\n return totalCount === 0 && !this._getInnerText(paragraph, false);\n });\n\n this._forEachNode(this._getAllNodesWithTag(articleContent, [\"br\"]), function(br) {\n var next = this._nextNode(br.nextSibling);\n if (next && next.tagName == \"P\")\n br.parentNode.removeChild(br);\n });\n\n // Remove single-cell tables\n this._forEachNode(this._getAllNodesWithTag(articleContent, [\"table\"]), function(table) {\n var tbody = this._hasSingleTagInsideElement(table, \"TBODY\") ? table.firstElementChild : table;\n if (this._hasSingleTagInsideElement(tbody, \"TR\")) {\n var row = tbody.firstElementChild;\n if (this._hasSingleTagInsideElement(row, \"TD\")) {\n var cell = row.firstElementChild;\n cell = this._setNodeTag(cell, this._everyNode(cell.childNodes, this._isPhrasingContent) ? \"P\" : \"DIV\");\n table.parentNode.replaceChild(cell, table);\n }\n }\n });\n },\n\n /**\n * Initialize a node with the readability object. Also checks the\n * className/id for special names to add to its score.\n *\n * @param Element\n * @return void\n **/\n _initializeNode: function(node) {\n node.readability = {\"contentScore\": 0};\n\n switch (node.tagName) {\n case \"DIV\":\n node.readability.contentScore += 5;\n break;\n\n case \"PRE\":\n case \"TD\":\n case \"BLOCKQUOTE\":\n node.readability.contentScore += 3;\n break;\n\n case \"ADDRESS\":\n case \"OL\":\n case \"UL\":\n case \"DL\":\n case \"DD\":\n case \"DT\":\n case \"LI\":\n case \"FORM\":\n node.readability.contentScore -= 3;\n break;\n\n case \"H1\":\n case \"H2\":\n case \"H3\":\n case \"H4\":\n case \"H5\":\n case \"H6\":\n case \"TH\":\n node.readability.contentScore -= 5;\n break;\n }\n\n node.readability.contentScore += this._getClassWeight(node);\n },\n\n _removeAndGetNext: function(node) {\n var nextNode = this._getNextNode(node, true);\n node.parentNode.removeChild(node);\n return nextNode;\n },\n\n /**\n * Traverse the DOM from node to node, starting at the node passed in.\n * Pass true for the second parameter to indicate this node itself\n * (and its kids) are going away, and we want the next node over.\n *\n * Calling this in a loop will traverse the DOM depth-first.\n */\n _getNextNode: function(node, ignoreSelfAndKids) {\n // First check for kids if those aren't being ignored\n if (!ignoreSelfAndKids && node.firstElementChild) {\n return node.firstElementChild;\n }\n // Then for siblings...\n if (node.nextElementSibling) {\n return node.nextElementSibling;\n }\n // And finally, move up the parent chain *and* find a sibling\n // (because this is depth-first traversal, we will have already\n // seen the parent nodes themselves).\n do {\n node = node.parentNode;\n } while (node && !node.nextElementSibling);\n return node && node.nextElementSibling;\n },\n\n // compares second text to first one\n // 1 = same text, 0 = completely different text\n // works the way that it splits both texts into words and then finds words that are unique in second text\n // the result is given by the lower length of unique parts\n _textSimilarity: function(textA, textB) {\n var tokensA = textA.toLowerCase().split(this.REGEXPS.tokenize).filter(Boolean);\n var tokensB = textB.toLowerCase().split(this.REGEXPS.tokenize).filter(Boolean);\n if (!tokensA.length || !tokensB.length) {\n return 0;\n }\n var uniqTokensB = tokensB.filter(token => !tokensA.includes(token));\n var distanceB = uniqTokensB.join(\" \").length / tokensB.join(\" \").length;\n return 1 - distanceB;\n },\n\n _checkByline: function(node, matchString) {\n if (this._articleByline) {\n return false;\n }\n\n if (node.getAttribute !== undefined) {\n var rel = node.getAttribute(\"rel\");\n var itemprop = node.getAttribute(\"itemprop\");\n }\n\n if ((rel === \"author\" || (itemprop && itemprop.indexOf(\"author\") !== -1) || this.REGEXPS.byline.test(matchString)) && this._isValidByline(node.textContent)) {\n this._articleByline = node.textContent.trim();\n return true;\n }\n\n return false;\n },\n\n _getNodeAncestors: function(node, maxDepth) {\n maxDepth = maxDepth || 0;\n var i = 0, ancestors = [];\n while (node.parentNode) {\n ancestors.push(node.parentNode);\n if (maxDepth && ++i === maxDepth)\n break;\n node = node.parentNode;\n }\n return ancestors;\n },\n\n /***\n * grabArticle - Using a variety of metrics (content score, classname, element types), find the content that is\n * most likely to be the stuff a user wants to read. Then return it wrapped up in a div.\n *\n * @param page a document to run upon. Needs to be a full document, complete with body.\n * @return Element\n **/\n _grabArticle: function (page) {\n this.log(\"**** grabArticle ****\");\n var doc = this._doc;\n var isPaging = page !== null;\n page = page ? page : this._doc.body;\n\n // We can't grab an article if we don't have a page!\n if (!page) {\n this.log(\"No body found in document. Abort.\");\n return null;\n }\n\n var pageCacheHtml = page.innerHTML;\n\n while (true) {\n this.log(\"Starting grabArticle loop\");\n var stripUnlikelyCandidates = this._flagIsActive(this.FLAG_STRIP_UNLIKELYS);\n\n // First, node prepping. Trash nodes that look cruddy (like ones with the\n // class name \"comment\", etc), and turn divs into P tags where they have been\n // used inappropriately (as in, where they contain no other block level elements.)\n var elementsToScore = [];\n var node = this._doc.documentElement;\n\n let shouldRemoveTitleHeader = true;\n\n while (node) {\n\n if (node.tagName === \"HTML\") {\n this._articleLang = node.getAttribute(\"lang\");\n }\n\n var matchString = node.className + \" \" + node.id;\n\n if (!this._isProbablyVisible(node)) {\n this.log(\"Removing hidden node - \" + matchString);\n node = this._removeAndGetNext(node);\n continue;\n }\n\n // User is not able to see elements applied with both \"aria-modal = true\" and \"role = dialog\"\n if (node.getAttribute(\"aria-modal\") == \"true\" && node.getAttribute(\"role\") == \"dialog\") {\n node = this._removeAndGetNext(node);\n continue;\n }\n\n // Check to see if this node is a byline, and remove it if it is.\n if (this._checkByline(node, matchString)) {\n node = this._removeAndGetNext(node);\n continue;\n }\n\n if (shouldRemoveTitleHeader && this._headerDuplicatesTitle(node)) {\n this.log(\"Removing header: \", node.textContent.trim(), this._articleTitle.trim());\n shouldRemoveTitleHeader = false;\n node = this._removeAndGetNext(node);\n continue;\n }\n\n // Remove unlikely candidates\n if (stripUnlikelyCandidates) {\n if (this.REGEXPS.unlikelyCandidates.test(matchString) &&\n !this.REGEXPS.okMaybeItsACandidate.test(matchString) &&\n !this._hasAncestorTag(node, \"table\") &&\n !this._hasAncestorTag(node, \"code\") &&\n node.tagName !== \"BODY\" &&\n node.tagName !== \"A\") {\n this.log(\"Removing unlikely candidate - \" + matchString);\n node = this._removeAndGetNext(node);\n continue;\n }\n\n if (this.UNLIKELY_ROLES.includes(node.getAttribute(\"role\"))) {\n this.log(\"Removing content with role \" + node.getAttribute(\"role\") + \" - \" + matchString);\n node = this._removeAndGetNext(node);\n continue;\n }\n }\n\n // Remove DIV, SECTION, and HEADER nodes without any content(e.g. text, image, video, or iframe).\n if ((node.tagName === \"DIV\" || node.tagName === \"SECTION\" || node.tagName === \"HEADER\" ||\n node.tagName === \"H1\" || node.tagName === \"H2\" || node.tagName === \"H3\" ||\n node.tagName === \"H4\" || node.tagName === \"H5\" || node.tagName === \"H6\") &&\n this._isElementWithoutContent(node)) {\n node = this._removeAndGetNext(node);\n continue;\n }\n\n if (this.DEFAULT_TAGS_TO_SCORE.indexOf(node.tagName) !== -1) {\n elementsToScore.push(node);\n }\n\n // Turn all divs that don't have children block level elements into p's\n if (node.tagName === \"DIV\") {\n // Put phrasing content into paragraphs.\n var p = null;\n var childNode = node.firstChild;\n while (childNode) {\n var nextSibling = childNode.nextSibling;\n if (this._isPhrasingContent(childNode)) {\n if (p !== null) {\n p.appendChild(childNode);\n } else if (!this._isWhitespace(childNode)) {\n p = doc.createElement(\"p\");\n node.replaceChild(p, childNode);\n p.appendChild(childNode);\n }\n } else if (p !== null) {\n while (p.lastChild && this._isWhitespace(p.lastChild)) {\n p.removeChild(p.lastChild);\n }\n p = null;\n }\n childNode = nextSibling;\n }\n\n // Sites like http://mobile.slate.com encloses each paragraph with a DIV\n // element. DIVs with only a P element inside and no text content can be\n // safely converted into plain P elements to avoid confusing the scoring\n // algorithm with DIVs with are, in practice, paragraphs.\n if (this._hasSingleTagInsideElement(node, \"P\") && this._getLinkDensity(node) < 0.25) {\n var newNode = node.children[0];\n node.parentNode.replaceChild(newNode, node);\n node = newNode;\n elementsToScore.push(node);\n } else if (!this._hasChildBlockElement(node)) {\n node = this._setNodeTag(node, \"P\");\n elementsToScore.push(node);\n }\n }\n node = this._getNextNode(node);\n }\n\n /**\n * Loop through all paragraphs, and assign a score to them based on how content-y they look.\n * Then add their score to their parent node.\n *\n * A score is determined by things like number of commas, class names, etc. Maybe eventually link density.\n **/\n var candidates = [];\n this._forEachNode(elementsToScore, function(elementToScore) {\n if (!elementToScore.parentNode || typeof(elementToScore.parentNode.tagName) === \"undefined\")\n return;\n\n // If this paragraph is less than 25 characters, don't even count it.\n var innerText = this._getInnerText(elementToScore);\n if (innerText.length < 25)\n return;\n\n // Exclude nodes with no ancestor.\n var ancestors = this._getNodeAncestors(elementToScore, 5);\n if (ancestors.length === 0)\n return;\n\n var contentScore = 0;\n\n // Add a point for the paragraph itself as a base.\n contentScore += 1;\n\n // Add points for any commas within this paragraph.\n contentScore += innerText.split(this.REGEXPS.commas).length;\n\n // For every 100 characters in this paragraph, add another point. Up to 3 points.\n contentScore += Math.min(Math.floor(innerText.length / 100), 3);\n\n // Initialize and score ancestors.\n this._forEachNode(ancestors, function(ancestor, level) {\n if (!ancestor.tagName || !ancestor.parentNode || typeof(ancestor.parentNode.tagName) === \"undefined\")\n return;\n\n if (typeof(ancestor.readability) === \"undefined\") {\n this._initializeNode(ancestor);\n candidates.push(ancestor);\n }\n\n // Node score divider:\n // - parent: 1 (no division)\n // - grandparent: 2\n // - great grandparent+: ancestor level * 3\n if (level === 0)\n var scoreDivider = 1;\n else if (level === 1)\n scoreDivider = 2;\n else\n scoreDivider = level * 3;\n ancestor.readability.contentScore += contentScore / scoreDivider;\n });\n });\n\n // After we've calculated scores, loop through all of the possible\n // candidate nodes we found and find the one with the highest score.\n var topCandidates = [];\n for (var c = 0, cl = candidates.length; c < cl; c += 1) {\n var candidate = candidates[c];\n\n // Scale the final candidates score based on link density. Good content\n // should have a relatively small link density (5% or less) and be mostly\n // unaffected by this operation.\n var candidateScore = candidate.readability.contentScore * (1 - this._getLinkDensity(candidate));\n candidate.readability.contentScore = candidateScore;\n\n this.log(\"Candidate:\", candidate, \"with score \" + candidateScore);\n\n for (var t = 0; t < this._nbTopCandidates; t++) {\n var aTopCandidate = topCandidates[t];\n\n if (!aTopCandidate || candidateScore > aTopCandidate.readability.contentScore) {\n topCandidates.splice(t, 0, candidate);\n if (topCandidates.length > this._nbTopCandidates)\n topCandidates.pop();\n break;\n }\n }\n }\n\n var topCandidate = topCandidates[0] || null;\n var neededToCreateTopCandidate = false;\n var parentOfTopCandidate;\n\n // If we still have no top candidate, just use the body as a last resort.\n // We also have to copy the body node so it is something we can modify.\n if (topCandidate === null || topCandidate.tagName === \"BODY\") {\n // Move all of the page's children into topCandidate\n topCandidate = doc.createElement(\"DIV\");\n neededToCreateTopCandidate = true;\n // Move everything (not just elements, also text nodes etc.) into the container\n // so we even include text directly in the body:\n while (page.firstChild) {\n this.log(\"Moving child out:\", page.firstChild);\n topCandidate.appendChild(page.firstChild);\n }\n\n page.appendChild(topCandidate);\n\n this._initializeNode(topCandidate);\n } else if (topCandidate) {\n // Find a better top candidate node if it contains (at least three) nodes which belong to `topCandidates` array\n // and whose scores are quite closed with current `topCandidate` node.\n var alternativeCandidateAncestors = [];\n for (var i = 1; i < topCandidates.length; i++) {\n if (topCandidates[i].readability.contentScore / topCandidate.readability.contentScore >= 0.75) {\n alternativeCandidateAncestors.push(this._getNodeAncestors(topCandidates[i]));\n }\n }\n var MINIMUM_TOPCANDIDATES = 3;\n if (alternativeCandidateAncestors.length >= MINIMUM_TOPCANDIDATES) {\n parentOfTopCandidate = topCandidate.parentNode;\n while (parentOfTopCandidate.tagName !== \"BODY\") {\n var listsContainingThisAncestor = 0;\n for (var ancestorIndex = 0; ancestorIndex < alternativeCandidateAncestors.length && listsContainingThisAncestor < MINIMUM_TOPCANDIDATES; ancestorIndex++) {\n listsContainingThisAncestor += Number(alternativeCandidateAncestors[ancestorIndex].includes(parentOfTopCandidate));\n }\n if (listsContainingThisAncestor >= MINIMUM_TOPCANDIDATES) {\n topCandidate = parentOfTopCandidate;\n break;\n }\n parentOfTopCandidate = parentOfTopCandidate.parentNode;\n }\n }\n if (!topCandidate.readability) {\n this._initializeNode(topCandidate);\n }\n\n // Because of our bonus system, parents of candidates might have scores\n // themselves. They get half of the node. There won't be nodes with higher\n // scores than our topCandidate, but if we see the score going *up* in the first\n // few steps up the tree, that's a decent sign that there might be more content\n // lurking in other places that we want to unify in. The sibling stuff\n // below does some of that - but only if we've looked high enough up the DOM\n // tree.\n parentOfTopCandidate = topCandidate.parentNode;\n var lastScore = topCandidate.readability.contentScore;\n // The scores shouldn't get too low.\n var scoreThreshold = lastScore / 3;\n while (parentOfTopCandidate.tagName !== \"BODY\") {\n if (!parentOfTopCandidate.readability) {\n parentOfTopCandidate = parentOfTopCandidate.parentNode;\n continue;\n }\n var parentScore = parentOfTopCandidate.readability.contentScore;\n if (parentScore < scoreThreshold)\n break;\n if (parentScore > lastScore) {\n // Alright! We found a better parent to use.\n topCandidate = parentOfTopCandidate;\n break;\n }\n lastScore = parentOfTopCandidate.readability.contentScore;\n parentOfTopCandidate = parentOfTopCandidate.parentNode;\n }\n\n // If the top candidate is the only child, use parent instead. This will help sibling\n // joining logic when adjacent content is actually located in parent's sibling node.\n parentOfTopCandidate = topCandidate.parentNode;\n while (parentOfTopCandidate.tagName != \"BODY\" && parentOfTopCandidate.children.length == 1) {\n topCandidate = parentOfTopCandidate;\n parentOfTopCandidate = topCandidate.parentNode;\n }\n if (!topCandidate.readability) {\n this._initializeNode(topCandidate);\n }\n }\n\n // Now that we have the top candidate, look through its siblings for content\n // that might also be related. Things like preambles, content split by ads\n // that we removed, etc.\n var articleContent = doc.createElement(\"DIV\");\n if (isPaging)\n articleContent.id = \"readability-content\";\n\n var siblingScoreThreshold = Math.max(10, topCandidate.readability.contentScore * 0.2);\n // Keep potential top candidate's parent node to try to get text direction of it later.\n parentOfTopCandidate = topCandidate.parentNode;\n var siblings = parentOfTopCandidate.children;\n\n for (var s = 0, sl = siblings.length; s < sl; s++) {\n var sibling = siblings[s];\n var append = false;\n\n this.log(\"Looking at sibling node:\", sibling, sibling.readability ? (\"with score \" + sibling.readability.contentScore) : \"\");\n this.log(\"Sibling has score\", sibling.readability ? sibling.readability.contentScore : \"Unknown\");\n\n if (sibling === topCandidate) {\n append = true;\n } else {\n var contentBonus = 0;\n\n // Give a bonus if sibling nodes and top candidates have the example same classname\n if (sibling.className === topCandidate.className && topCandidate.className !== \"\")\n contentBonus += topCandidate.readability.contentScore * 0.2;\n\n if (sibling.readability &&\n ((sibling.readability.contentScore + contentBonus) >= siblingScoreThreshold)) {\n append = true;\n } else if (sibling.nodeName === \"P\") {\n var linkDensity = this._getLinkDensity(sibling);\n var nodeContent = this._getInnerText(sibling);\n var nodeLength = nodeContent.length;\n\n if (nodeLength > 80 && linkDensity < 0.25) {\n append = true;\n } else if (nodeLength < 80 && nodeLength > 0 && linkDensity === 0 &&\n nodeContent.search(/\\.( |$)/) !== -1) {\n append = true;\n }\n }\n }\n\n if (append) {\n this.log(\"Appending node:\", sibling);\n\n if (this.ALTER_TO_DIV_EXCEPTIONS.indexOf(sibling.nodeName) === -1) {\n // We have a node that isn't a common block level element, like a form or td tag.\n // Turn it into a div so it doesn't get filtered out later by accident.\n this.log(\"Altering sibling:\", sibling, \"to div.\");\n\n sibling = this._setNodeTag(sibling, \"DIV\");\n }\n\n articleContent.appendChild(sibling);\n // Fetch children again to make it compatible\n // with DOM parsers without live collection support.\n siblings = parentOfTopCandidate.children;\n // siblings is a reference to the children array, and\n // sibling is removed from the array when we call appendChild().\n // As a result, we must revisit this index since the nodes\n // have been shifted.\n s -= 1;\n sl -= 1;\n }\n }\n\n if (this._debug)\n this.log(\"Article content pre-prep: \" + articleContent.innerHTML);\n // So we have all of the content that we need. Now we clean it up for presentation.\n this._prepArticle(articleContent);\n if (this._debug)\n this.log(\"Article content post-prep: \" + articleContent.innerHTML);\n\n if (neededToCreateTopCandidate) {\n // We already created a fake div thing, and there wouldn't have been any siblings left\n // for the previous loop, so there's no point trying to create a new div, and then\n // move all the children over. Just assign IDs and class names here. No need to append\n // because that already happened anyway.\n topCandidate.id = \"readability-page-1\";\n topCandidate.className = \"page\";\n } else {\n var div = doc.createElement(\"DIV\");\n div.id = \"readability-page-1\";\n div.className = \"page\";\n while (articleContent.firstChild) {\n div.appendChild(articleContent.firstChild);\n }\n articleContent.appendChild(div);\n }\n\n if (this._debug)\n this.log(\"Article content after paging: \" + articleContent.innerHTML);\n\n var parseSuccessful = true;\n\n // Now that we've gone through the full algorithm, check to see if\n // we got any meaningful content. If we didn't, we may need to re-run\n // grabArticle with different flags set. This gives us a higher likelihood of\n // finding the content, and the sieve approach gives us a higher likelihood of\n // finding the -right- content.\n var textLength = this._getInnerText(articleContent, true).length;\n if (textLength < this._charThreshold) {\n parseSuccessful = false;\n page.innerHTML = pageCacheHtml;\n\n if (this._flagIsActive(this.FLAG_STRIP_UNLIKELYS)) {\n this._removeFlag(this.FLAG_STRIP_UNLIKELYS);\n this._attempts.push({articleContent: articleContent, textLength: textLength});\n } else if (this._flagIsActive(this.FLAG_WEIGHT_CLASSES)) {\n this._removeFlag(this.FLAG_WEIGHT_CLASSES);\n this._attempts.push({articleContent: articleContent, textLength: textLength});\n } else if (this._flagIsActive(this.FLAG_CLEAN_CONDITIONALLY)) {\n this._removeFlag(this.FLAG_CLEAN_CONDITIONALLY);\n this._attempts.push({articleContent: articleContent, textLength: textLength});\n } else {\n this._attempts.push({articleContent: articleContent, textLength: textLength});\n // No luck after removing flags, just return the longest text we found during the different loops\n this._attempts.sort(function (a, b) {\n return b.textLength - a.textLength;\n });\n\n // But first check if we actually have something\n if (!this._attempts[0].textLength) {\n return null;\n }\n\n articleContent = this._attempts[0].articleContent;\n parseSuccessful = true;\n }\n }\n\n if (parseSuccessful) {\n // Find out text direction from ancestors of final top candidate.\n var ancestors = [parentOfTopCandidate, topCandidate].concat(this._getNodeAncestors(parentOfTopCandidate));\n this._someNode(ancestors, function(ancestor) {\n if (!ancestor.tagName)\n return false;\n var articleDir = ancestor.getAttribute(\"dir\");\n if (articleDir) {\n this._articleDir = articleDir;\n return true;\n }\n return false;\n });\n return articleContent;\n }\n }\n },\n\n /**\n * Check whether the input string could be a byline.\n * This verifies that the input is a string, and that the length\n * is less than 100 chars.\n *\n * @param possibleByline {string} - a string to check whether its a byline.\n * @return Boolean - whether the input string is a byline.\n */\n _isValidByline: function(byline) {\n if (typeof byline == \"string\" || byline instanceof String) {\n byline = byline.trim();\n return (byline.length > 0) && (byline.length < 100);\n }\n return false;\n },\n\n /**\n * Converts some of the common HTML entities in string to their corresponding characters.\n *\n * @param str {string} - a string to unescape.\n * @return string without HTML entity.\n */\n _unescapeHtmlEntities: function(str) {\n if (!str) {\n return str;\n }\n\n var htmlEscapeMap = this.HTML_ESCAPE_MAP;\n return str.replace(/&(quot|amp|apos|lt|gt);/g, function(_, tag) {\n return htmlEscapeMap[tag];\n }).replace(/&#(?:x([0-9a-z]{1,4})|([0-9]{1,4}));/gi, function(_, hex, numStr) {\n var num = parseInt(hex || numStr, hex ? 16 : 10);\n return String.fromCharCode(num);\n });\n },\n\n /**\n * Try to extract metadata from JSON-LD object.\n * For now, only Schema.org objects of type Article or its subtypes are supported.\n * @return Object with any metadata that could be extracted (possibly none)\n */\n _getJSONLD: function (doc) {\n var scripts = this._getAllNodesWithTag(doc, [\"script\"]);\n\n var metadata;\n\n this._forEachNode(scripts, function(jsonLdElement) {\n if (!metadata && jsonLdElement.getAttribute(\"type\") === \"application/ld+json\") {\n try {\n // Strip CDATA markers if present\n var content = jsonLdElement.textContent.replace(/^\\s*\\s*$/g, \"\");\n var parsed = JSON.parse(content);\n if (\n !parsed[\"@context\"] ||\n !parsed[\"@context\"].match(/^https?\\:\\/\\/schema\\.org$/)\n ) {\n return;\n }\n\n if (!parsed[\"@type\"] && Array.isArray(parsed[\"@graph\"])) {\n parsed = parsed[\"@graph\"].find(function(it) {\n return (it[\"@type\"] || \"\").match(\n this.REGEXPS.jsonLdArticleTypes\n );\n });\n }\n\n if (\n !parsed ||\n !parsed[\"@type\"] ||\n !parsed[\"@type\"].match(this.REGEXPS.jsonLdArticleTypes)\n ) {\n return;\n }\n\n metadata = {};\n\n if (typeof parsed.name === \"string\" && typeof parsed.headline === \"string\" && parsed.name !== parsed.headline) {\n // we have both name and headline element in the JSON-LD. They should both be the same but some websites like aktualne.cz\n // put their own name into \"name\" and the article title to \"headline\" which confuses Readability. So we try to check if either\n // \"name\" or \"headline\" closely matches the html title, and if so, use that one. If not, then we use \"name\" by default.\n\n var title = this._getArticleTitle();\n var nameMatches = this._textSimilarity(parsed.name, title) > 0.75;\n var headlineMatches = this._textSimilarity(parsed.headline, title) > 0.75;\n\n if (headlineMatches && !nameMatches) {\n metadata.title = parsed.headline;\n } else {\n metadata.title = parsed.name;\n }\n } else if (typeof parsed.name === \"string\") {\n metadata.title = parsed.name.trim();\n } else if (typeof parsed.headline === \"string\") {\n metadata.title = parsed.headline.trim();\n }\n if (parsed.author) {\n if (typeof parsed.author.name === \"string\") {\n metadata.byline = parsed.author.name.trim();\n } else if (Array.isArray(parsed.author) && parsed.author[0] && typeof parsed.author[0].name === \"string\") {\n metadata.byline = parsed.author\n .filter(function(author) {\n return author && typeof author.name === \"string\";\n })\n .map(function(author) {\n return author.name.trim();\n })\n .join(\", \");\n }\n }\n if (typeof parsed.description === \"string\") {\n metadata.excerpt = parsed.description.trim();\n }\n if (\n parsed.publisher &&\n typeof parsed.publisher.name === \"string\"\n ) {\n metadata.siteName = parsed.publisher.name.trim();\n }\n if (typeof parsed.datePublished === \"string\") {\n metadata.datePublished = parsed.datePublished.trim();\n }\n return;\n } catch (err) {\n this.log(err.message);\n }\n }\n });\n return metadata ? metadata : {};\n },\n\n /**\n * Attempts to get excerpt and byline metadata for the article.\n *\n * @param {Object} jsonld — object containing any metadata that\n * could be extracted from JSON-LD object.\n *\n * @return Object with optional \"excerpt\" and \"byline\" properties\n */\n _getArticleMetadata: function(jsonld) {\n var metadata = {};\n var values = {};\n var metaElements = this._doc.getElementsByTagName(\"meta\");\n\n // property is a space-separated list of values\n var propertyPattern = /\\s*(article|dc|dcterm|og|twitter)\\s*:\\s*(author|creator|description|published_time|title|site_name)\\s*/gi;\n\n // name is a single value\n var namePattern = /^\\s*(?:(dc|dcterm|og|twitter|weibo:(article|webpage))\\s*[\\.:]\\s*)?(author|creator|description|title|site_name)\\s*$/i;\n\n // Find description tags.\n this._forEachNode(metaElements, function(element) {\n var elementName = element.getAttribute(\"name\");\n var elementProperty = element.getAttribute(\"property\");\n var content = element.getAttribute(\"content\");\n if (!content) {\n return;\n }\n var matches = null;\n var name = null;\n\n if (elementProperty) {\n matches = elementProperty.match(propertyPattern);\n if (matches) {\n // Convert to lowercase, and remove any whitespace\n // so we can match below.\n name = matches[0].toLowerCase().replace(/\\s/g, \"\");\n // multiple authors\n values[name] = content.trim();\n }\n }\n if (!matches && elementName && namePattern.test(elementName)) {\n name = elementName;\n if (content) {\n // Convert to lowercase, remove any whitespace, and convert dots\n // to colons so we can match below.\n name = name.toLowerCase().replace(/\\s/g, \"\").replace(/\\./g, \":\");\n values[name] = content.trim();\n }\n }\n });\n\n // get title\n metadata.title = jsonld.title ||\n values[\"dc:title\"] ||\n values[\"dcterm:title\"] ||\n values[\"og:title\"] ||\n values[\"weibo:article:title\"] ||\n values[\"weibo:webpage:title\"] ||\n values[\"title\"] ||\n values[\"twitter:title\"];\n\n if (!metadata.title) {\n metadata.title = this._getArticleTitle();\n }\n\n // get author\n metadata.byline = jsonld.byline ||\n values[\"dc:creator\"] ||\n values[\"dcterm:creator\"] ||\n values[\"author\"];\n\n // get description\n metadata.excerpt = jsonld.excerpt ||\n values[\"dc:description\"] ||\n values[\"dcterm:description\"] ||\n values[\"og:description\"] ||\n values[\"weibo:article:description\"] ||\n values[\"weibo:webpage:description\"] ||\n values[\"description\"] ||\n values[\"twitter:description\"];\n\n // get site name\n metadata.siteName = jsonld.siteName ||\n values[\"og:site_name\"];\n\n // get article published time\n metadata.publishedTime = jsonld.datePublished ||\n values[\"article:published_time\"] || null;\n\n // in many sites the meta value is escaped with HTML entities,\n // so here we need to unescape it\n metadata.title = this._unescapeHtmlEntities(metadata.title);\n metadata.byline = this._unescapeHtmlEntities(metadata.byline);\n metadata.excerpt = this._unescapeHtmlEntities(metadata.excerpt);\n metadata.siteName = this._unescapeHtmlEntities(metadata.siteName);\n metadata.publishedTime = this._unescapeHtmlEntities(metadata.publishedTime);\n\n return metadata;\n },\n\n /**\n * Check if node is image, or if node contains exactly only one image\n * whether as a direct child or as its descendants.\n *\n * @param Element\n **/\n _isSingleImage: function(node) {\n if (node.tagName === \"IMG\") {\n return true;\n }\n\n if (node.children.length !== 1 || node.textContent.trim() !== \"\") {\n return false;\n }\n\n return this._isSingleImage(node.children[0]);\n },\n\n /**\n * Find all ![]()
![]()
![]()
![]()
![]()