| 1 | /* | | |
| 2 | * Copyright (c) 2010 Arc90 Inc | | |
| 3 | * | | |
| 4 | * Licensed under the Apache License, Version 2.0 (the "License"); | | |
| 5 | * you may not use this file except in compliance with the License. | | |
| 6 | * You may obtain a copy of the License at | | |
| 7 | * | | |
| 8 | * http://www.apache.org/licenses/LICENSE-2.0 | | |
| 9 | * | | |
| 10 | * Unless required by applicable law or agreed to in writing, software | | |
| 11 | * distributed under the License is distributed on an "AS IS" BASIS, | | |
| 12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | | |
| 13 | * See the License for the specific language governing permissions and | | |
| 14 | * limitations under the License. | | |
| 15 | */ | | |
| 16 | | | |
| 17 | /* | | |
| 18 | * This code is heavily based on Arc90's readability.js (1.7.1) script | | |
| 19 | * available at: http://code.google.com/p/arc90labs-readability | | |
| 20 | */ | | |
| 21 | | | |
| 22 | /** | | |
| 23 | * Public constructor. | | |
| 24 | * @param {HTMLDocument} doc The document to parse. | | |
| 25 | * @param {Object} options The options object. | | |
| 26 | */ | | |
| 27 | function Readability(doc, options) { | | |
| 28 | // In some older versions, people passed a URI as the first argument. Cope: | | |
| 29 | if (options && options.documentElement) { | | |
| 30 | doc = options; | | |
| 31 | options = arguments[2]; | | |
| 32 | } else if (!doc || !doc.documentElement) { | | |
| 33 | throw new Error("First argument to Readability constructor should be a document object."); | | |
| 34 | } | | |
| 35 | options = options || {}; | | |
| 36 | | | |
| 37 | this._doc = doc; | | |
| 38 | this._docJSDOMParser = this._doc.firstChild.__JSDOMParser__; | | |
| 39 | this._articleTitle = null; | | |
| 40 | this._articleByline = null; | | |
| 41 | this._articleDir = null; | | |
| 42 | this._articleSiteName = null; | | |
| 43 | this._attempts = []; | | |
| 44 | | | |
| 45 | // Configurable options | | |
| 46 | this._debug = !!options.debug; | | |
| 47 | this._maxElemsToParse = options.maxElemsToParse || this.DEFAULT_MAX_ELEMS_TO_PARSE; | | |
| 48 | this._nbTopCandidates = options.nbTopCandidates || this.DEFAULT_N_TOP_CANDIDATES; | | |
| 49 | this._charThreshold = options.charThreshold || this.DEFAULT_CHAR_THRESHOLD; | | |
| 50 | this._classesToPreserve = this.CLASSES_TO_PRESERVE.concat(options.classesToPreserve || []); | | |
| 51 | this._keepClasses = !!options.keepClasses; | | |
| 52 | this._serializer = options.serializer || function(el) { | | |
| 53 | return el.innerHTML; | | |
| 54 | }; | | |
| 55 | this._disableJSONLD = !!options.disableJSONLD; | | |
| 56 | this._allowedVideoRegex = options.allowedVideoRegex || this.REGEXPS.videos; | | |
| 57 | | | |
| 58 | // Start with all flags set | | |
| 59 | this._flags = this.FLAG_STRIP_UNLIKELYS | | | |
| 60 | this.FLAG_WEIGHT_CLASSES | | | |
| 61 | this.FLAG_CLEAN_CONDITIONALLY; | | |
| 62 | | | |
| 63 | | | |
| 64 | // Control whether log messages are sent to the console | | |
| 65 | if (this._debug) { | | |
| 66 | let logNode = function(node) { | | |
| 67 | if (node.nodeType == node.TEXT_NODE) { | | |
| 68 | return `${node.nodeName} ("${node.textContent}")`; | | |
| 69 | } | | |
| 70 | let attrPairs = Array.from(node.attributes || [], function(attr) { | | |
| 71 | return `${attr.name}="${attr.value}"`; | | |
| 72 | }).join(" "); | | |
| 73 | return `<${node.localName} ${attrPairs}>`; | | |
| 74 | }; | | |
| 75 | this.log = function () { | | |
| 76 | if (typeof console !== "undefined") { | | |
| 77 | let args = Array.from(arguments, arg => { | | |
| 78 | if (arg && arg.nodeType == this.ELEMENT_NODE) { | | |
| 79 | return logNode(arg); | | |
| 80 | } | | |
| 81 | return arg; | | |
| 82 | }); | | |
| 83 | args.unshift("Reader: (Readability)"); | | |
| 84 | console.log.apply(console, args); | | |
| 85 | } else if (typeof dump !== "undefined") { | | |
| 86 | /* global dump */ | | |
| 87 | var msg = Array.prototype.map.call(arguments, function(x) { | | |
| 88 | return (x && x.nodeName) ? logNode(x) : x; | | |
| 89 | }).join(" "); | | |
| 90 | dump("Reader: (Readability) " + msg + "\n"); | | |
| 91 | } | | |
| 92 | }; | | |
| 93 | } else { | | |
| 94 | this.log = function () {}; | | |
| 95 | } | | |
| 96 | } | | |
| 97 | | | |
| 98 | Readability.prototype = { | | |
| 99 | FLAG_STRIP_UNLIKELYS: 0x1, | | |
| 100 | FLAG_WEIGHT_CLASSES: 0x2, | | |
| 101 | FLAG_CLEAN_CONDITIONALLY: 0x4, | | |
| 102 | | | |
| 103 | // https://developer.mozilla.org/en-US/docs/Web/API/Node/nodeType | | |
| 104 | ELEMENT_NODE: 1, | | |
| 105 | TEXT_NODE: 3, | | |
| 106 | | | |
| 107 | // Max number of nodes supported by this parser. Default: 0 (no limit) | | |
| 108 | DEFAULT_MAX_ELEMS_TO_PARSE: 0, | | |
| 109 | | | |
| 110 | // The number of top candidates to consider when analysing how | | |
| 111 | // tight the competition is among candidates. | | |
| 112 | DEFAULT_N_TOP_CANDIDATES: 5, | | |
| 113 | | | |
| 114 | // Element tags to score by default. | | |
| 115 | DEFAULT_TAGS_TO_SCORE: "section,h2,h3,h4,h5,h6,p,td,pre".toUpperCase().split(","), | | |
| 116 | | | |
| 117 | // The default number of chars an article must have in order to return a result | | |
| 118 | DEFAULT_CHAR_THRESHOLD: 500, | | |
| 119 | | | |
| 120 | // All of the regular expressions in use within readability. | | |
| 121 | // Defined up here so we don't instantiate them repeatedly in loops. | | |
| 122 | REGEXPS: { | | |
| 123 | // NOTE: These two regular expressions are duplicated in | | |
| 124 | // Readability-readerable.js. Please keep both copies in sync. | | |
| 125 | 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, | | |
| 126 | okMaybeItsACandidate: /and|article|body|column|content|main|shadow/i, | | |
| 127 | | | |
| 128 | positive: /article|body|content|entry|hentry|h-entry|main|page|pagination|post|text|blog|story/i, | | |
| 129 | negative: /-ad-|hidden|^hid$| hid$| hid |^hid |banner|combx|comment|com-|contact|foot|footer|footnote|gdpr|masthead|media|meta|outbrain|promo|related|scroll|share|shoutbox|sidebar|skyscraper|sponsor|shopping|tags|tool|widget/i, | | |
| 130 | extraneous: /print|archive|comment|discuss|e[\-]?mail|share|reply|all|login|sign|single|utility/i, | | |
| 131 | byline: /byline|author|dateline|writtenby|p-author/i, | | |
| 132 | replaceFonts: /<(\/?)font[^>]*>/gi, | | |
| 133 | normalize: /\s{2,}/g, | | |
| 134 | videos: /\/\/(www\.)?((dailymotion|youtube|youtube-nocookie|player\.vimeo|v\.qq)\.com|(archive|upload\.wikimedia)\.org|player\.twitch\.tv)/i, | | |
| 135 | shareElements: /(\b|_)(share|sharedaddy)(\b|_)/i, | | |
| 136 | nextLink: /(next|weiter|continue|>([^\|]|$)|»([^\|]|$))/i, | | |
| 137 | prevLink: /(prev|earl|old|new|<|«)/i, | | |
| 138 | tokenize: /\W+/g, | | |
| 139 | whitespace: /^\s*$/, | | |
| 140 | hasContent: /\S$/, | | |
| 141 | hashUrl: /^#.+/, | | |
| 142 | srcsetUrl: /(\S+)(\s+[\d.]+[xw])?(\s*(?:,|$))/g, | | |
| 143 | b64DataUrl: /^data:\s*([^\s;,]+)\s*;\s*base64\s*,/i, | | |
| 144 | // Commas as used in Latin, Sindhi, Chinese and various other scripts. | | |
| 145 | // see: https://en.wikipedia.org/wiki/Comma#Comma_variants | | |
| 146 | commas: /\u002C|\u060C|\uFE50|\uFE10|\uFE11|\u2E41|\u2E34|\u2E32|\uFF0C/g, | | |
| 147 | // See: https://schema.org/Article | | |
| 148 | jsonLdArticleTypes: /^Article|AdvertiserContentArticle|NewsArticle|AnalysisNewsArticle|AskPublicNewsArticle|BackgroundNewsArticle|OpinionNewsArticle|ReportageNewsArticle|ReviewNewsArticle|Report|SatiricalArticle|ScholarlyArticle|MedicalScholarlyArticle|SocialMediaPosting|BlogPosting|LiveBlogPosting|DiscussionForumPosting|TechArticle|APIReference$/ | | |
| 149 | }, | | |
| 150 | | | |
| 151 | UNLIKELY_ROLES: [ "menu", "menubar", "complementary", "navigation", "alert", "alertdialog", "dialog" ], | | |
| 152 | | | |
| 153 | DIV_TO_P_ELEMS: new Set([ "BLOCKQUOTE", "DL", "DIV", "IMG", "OL", "P", "PRE", "TABLE", "UL" ]), | | |
| 154 | | | |
| 155 | ALTER_TO_DIV_EXCEPTIONS: ["DIV", "ARTICLE", "SECTION", "P"], | | |
| 156 | | | |
| 157 | PRESENTATIONAL_ATTRIBUTES: [ "align", "background", "bgcolor", "border", "cellpadding", "cellspacing", "frame", "hspace", "rules", "style", "valign", "vspace" ], | | |
| 158 | | | |
| 159 | DEPRECATED_SIZE_ATTRIBUTE_ELEMS: [ "TABLE", "TH", "TD", "HR", "PRE" ], | | |
| 160 | | | |
| 161 | // The commented out elements qualify as phrasing content but tend to be | | |
| 162 | // removed by readability when put into paragraphs, so we ignore them here. | | |
| 163 | PHRASING_ELEMS: [ | | |
| 164 | // "CANVAS", "IFRAME", "SVG", "VIDEO", | | |
| 165 | "ABBR", "AUDIO", "B", "BDO", "BR", "BUTTON", "CITE", "CODE", "DATA", | | |
| 166 | "DATALIST", "DFN", "EM", "EMBED", "I", "IMG", "INPUT", "KBD", "LABEL", | | |
| 167 | "MARK", "MATH", "METER", "NOSCRIPT", "OBJECT", "OUTPUT", "PROGRESS", "Q", | | |
| 168 | "RUBY", "SAMP", "SCRIPT", "SELECT", "SMALL", "SPAN", "STRONG", "SUB", | | |
| 169 | "SUP", "TEXTAREA", "TIME", "VAR", "WBR" | | |
| 170 | ], | | |
| 171 | | | |
| 172 | // These are the classes that readability sets itself. | | |
| 173 | CLASSES_TO_PRESERVE: [ "page" ], | | |
| 174 | | | |
| 175 | // These are the list of HTML entities that need to be escaped. | | |
| 176 | HTML_ESCAPE_MAP: { | | |
| 177 | "lt": "<", | | |
| 178 | "gt": ">", | | |
| 179 | "amp": "&", | | |
| 180 | "quot": '"', | | |
| 181 | "apos": "'", | | |
| 182 | }, | | |
| 183 | | | |
| 184 | /** | | |
| 185 | * Run any post-process modifications to article content as necessary. | | |
| 186 | * | | |
| 187 | * @param Element | | |
| 188 | * @return void | | |
| 189 | **/ | | |
| 190 | _postProcessContent: function(articleContent) { | | |
| 191 | // Readability cannot open relative uris so we convert them to absolute uris. | | |
| 192 | this._fixRelativeUris(articleContent); | | |
| 193 | | | |
| 194 | this._simplifyNestedElements(articleContent); | | |
| 195 | | | |
| 196 | if (!this._keepClasses) { | | |
| 197 | // Remove classes. | | |
| 198 | this._cleanClasses(articleContent); | | |
| 199 | } | | |
| 200 | }, | | |
| 201 | | | |
| 202 | /** | | |
| 203 | * Iterates over a NodeList, calls `filterFn` for each node and removes node | | |
| 204 | * if function returned `true`. | | |
| 205 | * | | |
| 206 | * If function is not passed, removes all the nodes in node list. | | |
| 207 | * | | |
| 208 | * @param NodeList nodeList The nodes to operate on | | |
| 209 | * @param Function filterFn the function to use as a filter | | |
| 210 | * @return void | | |
| 211 | */ | | |
| 212 | _removeNodes: function(nodeList, filterFn) { | | |
| 213 | // Avoid ever operating on live node lists. | | |
| 214 | if (this._docJSDOMParser && nodeList._isLiveNodeList) { | | |
| 215 | throw new Error("Do not pass live node lists to _removeNodes"); | | |
| 216 | } | | |
| 217 | for (var i = nodeList.length - 1; i >= 0; i--) { | | |
| 218 | var node = nodeList[i]; | | |
| 219 | var parentNode = node.parentNode; | | |
| 220 | if (parentNode) { | | |
| 221 | if (!filterFn || filterFn.call(this, node, i, nodeList)) { | | |
| 222 | parentNode.removeChild(node); | | |
| 223 | } | | |
| 224 | } | | |
| 225 | } | | |
| 226 | }, | | |
| 227 | | | |
| 228 | /** | | |
| 229 | * Iterates over a NodeList, and calls _setNodeTag for each node. | | |
| 230 | * | | |
| 231 | * @param NodeList nodeList The nodes to operate on | | |
| 232 | * @param String newTagName the new tag name to use | | |
| 233 | * @return void | | |
| 234 | */ | | |
| 235 | _replaceNodeTags: function(nodeList, newTagName) { | | |
| 236 | // Avoid ever operating on live node lists. | | |
| 237 | if (this._docJSDOMParser && nodeList._isLiveNodeList) { | | |
| 238 | throw new Error("Do not pass live node lists to _replaceNodeTags"); | | |
| 239 | } | | |
| 240 | for (const node of nodeList) { | | |
| 241 | this._setNodeTag(node, newTagName); | | |
| 242 | } | | |
| 243 | }, | | |
| 244 | | | |
| 245 | /** | | |
| 246 | * Iterate over a NodeList, which doesn't natively fully implement the Array | | |
| 247 | * interface. | | |
| 248 | * | | |
| 249 | * For convenience, the current object context is applied to the provided | | |
| 250 | * iterate function. | | |
| 251 | * | | |
| 252 | * @param NodeList nodeList The NodeList. | | |
| 253 | * @param Function fn The iterate function. | | |
| 254 | * @return void | | |
| 255 | */ | | |
| 256 | _forEachNode: function(nodeList, fn) { | | |
| 257 | Array.prototype.forEach.call(nodeList, fn, this); | | |
| 258 | }, | | |
| 259 | | | |
| 260 | /** | | |
| 261 | * Iterate over a NodeList, and return the first node that passes | | |
| 262 | * the supplied test function | | |
| 263 | * | | |
| 264 | * For convenience, the current object context is applied to the provided | | |
| 265 | * test function. | | |
| 266 | * | | |
| 267 | * @param NodeList nodeList The NodeList. | | |
| 268 | * @param Function fn The test function. | | |
| 269 | * @return void | | |
| 270 | */ | | |
| 271 | _findNode: function(nodeList, fn) { | | |
| 272 | return Array.prototype.find.call(nodeList, fn, this); | | |
| 273 | }, | | |
| 274 | | | |
| 275 | /** | | |
| 276 | * Iterate over a NodeList, return true if any of the provided iterate | | |
| 277 | * function calls returns true, false otherwise. | | |
| 278 | * | | |
| 279 | * For convenience, the current object context is applied to the | | |
| 280 | * provided iterate function. | | |
| 281 | * | | |
| 282 | * @param NodeList nodeList The NodeList. | | |
| 283 | * @param Function fn The iterate function. | | |
| 284 | * @return Boolean | | |
| 285 | */ | | |
| 286 | _someNode: function(nodeList, fn) { | | |
| 287 | return Array.prototype.some.call(nodeList, fn, this); | | |
| 288 | }, | | |
| 289 | | | |
| 290 | /** | | |
| 291 | * Iterate over a NodeList, return true if all of the provided iterate | | |
| 292 | * function calls return true, false otherwise. | | |
| 293 | * | | |
| 294 | * For convenience, the current object context is applied to the | | |
| 295 | * provided iterate function. | | |
| 296 | * | | |
| 297 | * @param NodeList nodeList The NodeList. | | |
| 298 | * @param Function fn The iterate function. | | |
| 299 | * @return Boolean | | |
| 300 | */ | | |
| 301 | _everyNode: function(nodeList, fn) { | | |
| 302 | return Array.prototype.every.call(nodeList, fn, this); | | |
| 303 | }, | | |
| 304 | | | |
| 305 | /** | | |
| 306 | * Concat all nodelists passed as arguments. | | |
| 307 | * | | |
| 308 | * @return ...NodeList | | |
| 309 | * @return Array | | |
| 310 | */ | | |
| 311 | _concatNodeLists: function() { | | |
| 312 | var slice = Array.prototype.slice; | | |
| 313 | var args = slice.call(arguments); | | |
| 314 | var nodeLists = args.map(function(list) { | | |
| 315 | return slice.call(list); | | |
| 316 | }); | | |
| 317 | return Array.prototype.concat.apply([], nodeLists); | | |
| 318 | }, | | |
| 319 | | | |
| 320 | _getAllNodesWithTag: function(node, tagNames) { | | |
| 321 | if (node.querySelectorAll) { | | |
| 322 | return node.querySelectorAll(tagNames.join(",")); | | |
| 323 | } | | |
| 324 | return [].concat.apply([], tagNames.map(function(tag) { | | |
| 325 | var collection = node.getElementsByTagName(tag); | | |
| 326 | return Array.isArray(collection) ? collection : Array.from(collection); | | |
| 327 | })); | | |
| 328 | }, | | |
| 329 | | | |
| 330 | /** | | |
| 331 | * Removes the class="" attribute from every element in the given | | |
| 332 | * subtree, except those that match CLASSES_TO_PRESERVE and | | |
| 333 | * the classesToPreserve array from the options object. | | |
| 334 | * | | |
| 335 | * @param Element | | |
| 336 | * @return void | | |
| 337 | */ | | |
| 338 | _cleanClasses: function(node) { | | |
| 339 | var classesToPreserve = this._classesToPreserve; | | |
| 340 | var className = (node.getAttribute("class") || "") | | |
| 341 | .split(/\s+/) | | |
| 342 | .filter(function(cls) { | | |
| 343 | return classesToPreserve.indexOf(cls) != -1; | | |
| 344 | }) | | |
| 345 | .join(" "); | | |
| 346 | | | |
| 347 | if (className) { | | |
| 348 | node.setAttribute("class", className); | | |
| 349 | } else { | | |
| 350 | node.removeAttribute("class"); | | |
| 351 | } | | |
| 352 | | | |
| 353 | for (node = node.firstElementChild; node; node = node.nextElementSibling) { | | |
| 354 | this._cleanClasses(node); | | |
| 355 | } | | |
| 356 | }, | | |
| 357 | | | |
| 358 | /** | | |
| 359 | * Converts each <a> and <img> uri in the given element to an absolute URI, | | |
| 360 | * ignoring #ref URIs. | | |
| 361 | * | | |
| 362 | * @param Element | | |
| 363 | * @return void | | |
| 364 | */ | | |
| 365 | _fixRelativeUris: function(articleContent) { | | |
| 366 | var baseURI = this._doc.baseURI; | | |
| 367 | var documentURI = this._doc.documentURI; | | |
| 368 | function toAbsoluteURI(uri) { | | |
| 369 | // Leave hash links alone if the base URI matches the document URI: | | |
| 370 | if (baseURI == documentURI && uri.charAt(0) == "#") { | | |
| 371 | return uri; | | |
| 372 | } | | |
| 373 | | | |
| 374 | // Otherwise, resolve against base URI: | | |
| 375 | try { | | |
| 376 | return new URL(uri, baseURI).href; | | |
| 377 | } catch (ex) { | | |
| 378 | // Something went wrong, just return the original: | | |
| 379 | } | | |
| 380 | return uri; | | |
| 381 | } | | |
| 382 | | | |
| 383 | var links = this._getAllNodesWithTag(articleContent, ["a"]); | | |
| 384 | this._forEachNode(links, function(link) { | | |
| 385 | var href = link.getAttribute("href"); | | |
| 386 | if (href) { | | |
| 387 | // Remove links with javascript: URIs, since | | |
| 388 | // they won't work after scripts have been removed from the page. | | |
| 389 | if (href.indexOf("javascript:") === 0) { | | |
| 390 | // if the link only contains simple text content, it can be converted to a text node | | |
| 391 | if (link.childNodes.length === 1 && link.childNodes[0].nodeType === this.TEXT_NODE) { | | |
| 392 | var text = this._doc.createTextNode(link.textContent); | | |
| 393 | link.parentNode.replaceChild(text, link); | | |
| 394 | } else { | | |
| 395 | // if the link has multiple children, they should all be preserved | | |
| 396 | var container = this._doc.createElement("span"); | | |
| 397 | while (link.firstChild) { | | |
| 398 | container.appendChild(link.firstChild); | | |
| 399 | } | | |
| 400 | link.parentNode.replaceChild(container, link); | | |
| 401 | } | | |
| 402 | } else { | | |
| 403 | link.setAttribute("href", toAbsoluteURI(href)); | | |
| 404 | } | | |
| 405 | } | | |
| 406 | }); | | |
| 407 | | | |
| 408 | var medias = this._getAllNodesWithTag(articleContent, [ | | |
| 409 | "img", "picture", "figure", "video", "audio", "source" | | |
| 410 | ]); | | |
| 411 | | | |
| 412 | this._forEachNode(medias, function(media) { | | |
| 413 | var src = media.getAttribute("src"); | | |
| 414 | var poster = media.getAttribute("poster"); | | |
| 415 | var srcset = media.getAttribute("srcset"); | | |
| 416 | | | |
| 417 | if (src) { | | |
| 418 | media.setAttribute("src", toAbsoluteURI(src)); | | |
| 419 | } | | |
| 420 | | | |
| 421 | if (poster) { | | |
| 422 | media.setAttribute("poster", toAbsoluteURI(poster)); | | |
| 423 | } | | |
| 424 | | | |
| 425 | if (srcset) { | | |
| 426 | var newSrcset = srcset.replace(this.REGEXPS.srcsetUrl, function(_, p1, p2, p3) { | | |
| 427 | return toAbsoluteURI(p1) + (p2 || "") + p3; | | |
| 428 | }); | | |
| 429 | | | |
| 430 | media.setAttribute("srcset", newSrcset); | | |
| 431 | } | | |
| 432 | }); | | |
| 433 | }, | | |
| 434 | | | |
| 435 | _simplifyNestedElements: function(articleContent) { | | |
| 436 | var node = articleContent; | | |
| 437 | | | |
| 438 | while (node) { | | |
| 439 | if (node.parentNode && ["DIV", "SECTION"].includes(node.tagName) && !(node.id && node.id.startsWith("readability"))) { | | |
| 440 | if (this._isElementWithoutContent(node)) { | | |
| 441 | node = this._removeAndGetNext(node); | | |
| 442 | continue; | | |
| 443 | } else if (this._hasSingleTagInsideElement(node, "DIV") || this._hasSingleTagInsideElement(node, "SECTION")) { | | |
| 444 | var child = node.children[0]; | | |
| 445 | for (var i = 0; i < node.attributes.length; i++) { | | |
| 446 | child.setAttribute(node.attributes[i].name, node.attributes[i].value); | | |
| 447 | } | | |
| 448 | node.parentNode.replaceChild(child, node); | | |
| 449 | node = child; | | |
| 450 | continue; | | |
| 451 | } | | |
| 452 | } | | |
| 453 | | | |
| 454 | node = this._getNextNode(node); | | |
| 455 | } | | |
| 456 | }, | | |
| 457 | | | |
| 458 | /** | | |
| 459 | * Get the article title as an H1. | | |
| 460 | * | | |
| 461 | * @return string | | |
| 462 | **/ | | |
| 463 | _getArticleTitle: function() { | | |
| 464 | var doc = this._doc; | | |
| 465 | var curTitle = ""; | | |
| 466 | var origTitle = ""; | | |
| 467 | | | |
| 468 | try { | | |
| 469 | curTitle = origTitle = doc.title.trim(); | | |
| 470 | | | |
| 471 | // If they had an element with id "title" in their HTML | | |
| 472 | if (typeof curTitle !== "string") | | |
| 473 | curTitle = origTitle = this._getInnerText(doc.getElementsByTagName("title")[0]); | | |
| 474 | } catch (e) {/* ignore exceptions setting the title. */} | | |
| 475 | | | |
| 476 | var titleHadHierarchicalSeparators = false; | | |
| 477 | function wordCount(str) { | | |
| 478 | return str.split(/\s+/).length; | | |
| 479 | } | | |
| 480 | | | |
| 481 | // If there's a separator in the title, first remove the final part | | |
| 482 | if ((/ [\|\-\\\/>»] /).test(curTitle)) { | | |
| 483 | titleHadHierarchicalSeparators = / [\\\/>»] /.test(curTitle); | | |
| 484 | curTitle = origTitle.replace(/(.*)[\|\-\\\/>»] .*/gi, "$1"); | | |
| 485 | | | |
| 486 | // If the resulting title is too short (3 words or fewer), remove | | |
| 487 | // the first part instead: | | |
| 488 | if (wordCount(curTitle) < 3) | | |
| 489 | curTitle = origTitle.replace(/[^\|\-\\\/>»]*[\|\-\\\/>»](.*)/gi, "$1"); | | |
| 490 | } else if (curTitle.indexOf(": ") !== -1) { | | |
| 491 | // Check if we have an heading containing this exact string, so we | | |
| 492 | // could assume it's the full title. | | |
| 493 | var headings = this._concatNodeLists( | | |
| 494 | doc.getElementsByTagName("h1"), | | |
| 495 | doc.getElementsByTagName("h2") | | |
| 496 | ); | | |
| 497 | var trimmedTitle = curTitle.trim(); | | |
| 498 | var match = this._someNode(headings, function(heading) { | | |
| 499 | return heading.textContent.trim() === trimmedTitle; | | |
| 500 | }); | | |
| 501 | | | |
| 502 | // If we don't, let's extract the title out of the original title string. | | |
| 503 | if (!match) { | | |
| 504 | curTitle = origTitle.substring(origTitle.lastIndexOf(":") + 1); | | |
| 505 | | | |
| 506 | // If the title is now too short, try the first colon instead: | | |
| 507 | if (wordCount(curTitle) < 3) { | | |
| 508 | curTitle = origTitle.substring(origTitle.indexOf(":") + 1); | | |
| 509 | // But if we have too many words before the colon there's something weird | | |
| 510 | // with the titles and the H tags so let's just use the original title instead | | |
| 511 | } else if (wordCount(origTitle.substr(0, origTitle.indexOf(":"))) > 5) { | | |
| 512 | curTitle = origTitle; | | |
| 513 | } | | |
| 514 | } | | |
| 515 | } else if (curTitle.length > 150 || curTitle.length < 15) { | | |
| 516 | var hOnes = doc.getElementsByTagName("h1"); | | |
| 517 | | | |
| 518 | if (hOnes.length === 1) | | |
| 519 | curTitle = this._getInnerText(hOnes[0]); | | |
| 520 | } | | |
| 521 | | | |
| 522 | curTitle = curTitle.trim().replace(this.REGEXPS.normalize, " "); | | |
| 523 | // If we now have 4 words or fewer as our title, and either no | | |
| 524 | // 'hierarchical' separators (\, /, > or ») were found in the original | | |
| 525 | // title or we decreased the number of words by more than 1 word, use | | |
| 526 | // the original title. | | |
| 527 | var curTitleWordCount = wordCount(curTitle); | | |
| 528 | if (curTitleWordCount <= 4 && | | |
| 529 | (!titleHadHierarchicalSeparators || | | |
| 530 | curTitleWordCount != wordCount(origTitle.replace(/[\|\-\\\/>»]+/g, "")) - 1)) { | | |
| 531 | curTitle = origTitle; | | |
| 532 | } | | |
| 533 | | | |
| 534 | return curTitle; | | |
| 535 | }, | | |
| 536 | | | |
| 537 | /** | | |
| 538 | * Prepare the HTML document for readability to scrape it. | | |
| 539 | * This includes things like stripping javascript, CSS, and handling terrible markup. | | |
| 540 | * | | |
| 541 | * @return void | | |
| 542 | **/ | | |
| 543 | _prepDocument: function() { | | |
| 544 | var doc = this._doc; | | |
| 545 | | | |
| 546 | // Remove all style tags in head | | |
| 547 | this._removeNodes(this._getAllNodesWithTag(doc, ["style"])); | | |
| 548 | | | |
| 549 | if (doc.body) { | | |
| 550 | this._replaceBrs(doc.body); | | |
| 551 | } | | |
| 552 | | | |
| 553 | this._replaceNodeTags(this._getAllNodesWithTag(doc, ["font"]), "SPAN"); | | |
| 554 | }, | | |
| 555 | | | |
| 556 | /** | | |
| 557 | * Finds the next node, starting from the given node, and ignoring | | |
| 558 | * whitespace in between. If the given node is an element, the same node is | | |
| 559 | * returned. | | |
| 560 | */ | | |
| 561 | _nextNode: function (node) { | | |
| 562 | var next = node; | | |
| 563 | while (next | | |
| 564 | && (next.nodeType != this.ELEMENT_NODE) | | |
| 565 | && this.REGEXPS.whitespace.test(next.textContent)) { | | |
| 566 | next = next.nextSibling; | | |
| 567 | } | | |
| 568 | return next; | | |
| 569 | }, | | |
| 570 | | | |
| 571 | /** | | |
| 572 | * Replaces 2 or more successive <br> elements with a single <p>. | | |
| 573 | * Whitespace between <br> elements are ignored. For example: | | |
| 574 | * <div>foo<br>bar<br> <br><br>abc</div> | | |
| 575 | * will become: | | |
| 576 | * <div>foo<br>bar<p>abc</p></div> | | |
| 577 | */ | | |
| 578 | _replaceBrs: function (elem) { | | |
| 579 | this._forEachNode(this._getAllNodesWithTag(elem, ["br"]), function(br) { | | |
| 580 | var next = br.nextSibling; | | |
| 581 | | | |
| 582 | // Whether 2 or more <br> elements have been found and replaced with a | | |
| 583 | // <p> block. | | |
| 584 | var replaced = false; | | |
| 585 | | | |
| 586 | // If we find a <br> chain, remove the <br>s until we hit another node | | |
| 587 | // or non-whitespace. This leaves behind the first <br> in the chain | | |
| 588 | // (which will be replaced with a <p> later). | | |
| 589 | while ((next = this._nextNode(next)) && (next.tagName == "BR")) { | | |
| 590 | replaced = true; | | |
| 591 | var brSibling = next.nextSibling; | | |
| 592 | next.parentNode.removeChild(next); | | |
| 593 | next = brSibling; | | |
| 594 | } | | |
| 595 | | | |
| 596 | // If we removed a <br> chain, replace the remaining <br> with a <p>. Add | | |
| 597 | // all sibling nodes as children of the <p> until we hit another <br> | | |
| 598 | // chain. | | |
| 599 | if (replaced) { | | |
| 600 | var p = this._doc.createElement("p"); | | |
| 601 | br.parentNode.replaceChild(p, br); | | |
| 602 | | | |
| 603 | next = p.nextSibling; | | |
| 604 | while (next) { | | |
| 605 | // If we've hit another <br><br>, we're done adding children to this <p>. | | |
| 606 | if (next.tagName == "BR") { | | |
| 607 | var nextElem = this._nextNode(next.nextSibling); | | |
| 608 | if (nextElem && nextElem.tagName == "BR") | | |
| 609 | break; | | |
| 610 | } | | |
| 611 | | | |
| 612 | if (!this._isPhrasingContent(next)) | | |
| 613 | break; | | |
| 614 | | | |
| 615 | // Otherwise, make this node a child of the new <p>. | | |
| 616 | var sibling = next.nextSibling; | | |
| 617 | p.appendChild(next); | | |
| 618 | next = sibling; | | |
| 619 | } | | |
| 620 | | | |
| 621 | while (p.lastChild && this._isWhitespace(p.lastChild)) { | | |
| 622 | p.removeChild(p.lastChild); | | |
| 623 | } | | |
| 624 | | | |
| 625 | if (p.parentNode.tagName === "P") | | |
| 626 | this._setNodeTag(p.parentNode, "DIV"); | | |
| 627 | } | | |
| 628 | }); | | |
| 629 | }, | | |
| 630 | | | |
| 631 | _setNodeTag: function (node, tag) { | | |
| 632 | this.log("_setNodeTag", node, tag); | | |
| 633 | if (this._docJSDOMParser) { | | |
| 634 | node.localName = tag.toLowerCase(); | | |
| 635 | node.tagName = tag.toUpperCase(); | | |
| 636 | return node; | | |
| 637 | } | | |
| 638 | | | |
| 639 | var replacement = node.ownerDocument.createElement(tag); | | |
| 640 | while (node.firstChild) { | | |
| 641 | replacement.appendChild(node.firstChild); | | |
| 642 | } | | |
| 643 | node.parentNode.replaceChild(replacement, node); | | |
| 644 | if (node.readability) | | |
| 645 | replacement.readability = node.readability; | | |
| 646 | | | |
| 647 | for (var i = 0; i < node.attributes.length; i++) { | | |
| 648 | try { | | |
| 649 | replacement.setAttribute(node.attributes[i].name, node.attributes[i].value); | | |
| 650 | } catch (ex) { | | |
| 651 | /* it's possible for setAttribute() to throw if the attribute name | | |
| 652 | * isn't a valid XML Name. Such attributes can however be parsed from | | |
| 653 | * source in HTML docs, see https://github.com/whatwg/html/issues/4275, | | |
| 654 | * so we can hit them here and then throw. We don't care about such | | |
| 655 | * attributes so we ignore them. | | |
| 656 | */ | | |
| 657 | } | | |
| 658 | } | | |
| 659 | return replacement; | | |
| 660 | }, | | |
| 661 | | | |
| 662 | /** | | |
| 663 | * Prepare the article node for display. Clean out any inline styles, | | |
| 664 | * iframes, forms, strip extraneous <p> tags, etc. | | |
| 665 | * | | |
| 666 | * @param Element | | |
| 667 | * @return void | | |
| 668 | **/ | | |
| 669 | _prepArticle: function(articleContent) { | | |
| 670 | this._cleanStyles(articleContent); | | |
| 671 | | | |
| 672 | // Check for data tables before we continue, to avoid removing items in | | |
| 673 | // those tables, which will often be isolated even though they're | | |
| 674 | // visually linked to other content-ful elements (text, images, etc.). | | |
| 675 | this._markDataTables(articleContent); | | |
| 676 | | | |
| 677 | this._fixLazyImages(articleContent); | | |
| 678 | | | |
| 679 | // Clean out junk from the article content | | |
| 680 | this._cleanConditionally(articleContent, "form"); | | |
| 681 | this._cleanConditionally(articleContent, "fieldset"); | | |
| 682 | this._clean(articleContent, "object"); | | |
| 683 | this._clean(articleContent, "embed"); | | |
| 684 | this._clean(articleContent, "footer"); | | |
| 685 | this._clean(articleContent, "link"); | | |
| 686 | this._clean(articleContent, "aside"); | | |
| 687 | | | |
| 688 | // Clean out elements with little content that have "share" in their id/class combinations from final top candidates, | | |
| 689 | // which means we don't remove the top candidates even they have "share". | | |
| 690 | | | |
| 691 | var shareElementThreshold = this.DEFAULT_CHAR_THRESHOLD; | | |
| 692 | | | |
| 693 | this._forEachNode(articleContent.children, function (topCandidate) { | | |
| 694 | this._cleanMatchedNodes(topCandidate, function (node, matchString) { | | |
| 695 | return this.REGEXPS.shareElements.test(matchString) && node.textContent.length < shareElementThreshold; | | |
| 696 | }); | | |
| 697 | }); | | |
| 698 | | | |
| 699 | this._clean(articleContent, "iframe"); | | |
| 700 | this._clean(articleContent, "input"); | | |
| 701 | this._clean(articleContent, "textarea"); | | |
| 702 | this._clean(articleContent, "select"); | | |
| 703 | this._clean(articleContent, "button"); | | |
| 704 | this._cleanHeaders(articleContent); | | |
| 705 | | | |
| 706 | // Do these last as the previous stuff may have removed junk | | |
| 707 | // that will affect these | | |
| 708 | this._cleanConditionally(articleContent, "table"); | | |
| 709 | this._cleanConditionally(articleContent, "ul"); | | |
| 710 | this._cleanConditionally(articleContent, "div"); | | |
| 711 | | | |
| 712 | // replace H1 with H2 as H1 should be only title that is displayed separately | | |
| 713 | this._replaceNodeTags(this._getAllNodesWithTag(articleContent, ["h1"]), "h2"); | | |
| 714 | | | |
| 715 | // Remove extra paragraphs | | |
| 716 | this._removeNodes(this._getAllNodesWithTag(articleContent, ["p"]), function (paragraph) { | | |
| 717 | var imgCount = paragraph.getElementsByTagName("img").length; | | |
| 718 | var embedCount = paragraph.getElementsByTagName("embed").length; | | |
| 719 | var objectCount = paragraph.getElementsByTagName("object").length; | | |
| 720 | // At this point, nasty iframes have been removed, only remain embedded video ones. | | |
| 721 | var iframeCount = paragraph.getElementsByTagName("iframe").length; | | |
| 722 | var totalCount = imgCount + embedCount + objectCount + iframeCount; | | |
| 723 | | | |
| 724 | return totalCount === 0 && !this._getInnerText(paragraph, false); | | |
| 725 | }); | | |
| 726 | | | |
| 727 | this._forEachNode(this._getAllNodesWithTag(articleContent, ["br"]), function(br) { | | |
| 728 | var next = this._nextNode(br.nextSibling); | | |
| 729 | if (next && next.tagName == "P") | | |
| 730 | br.parentNode.removeChild(br); | | |
| 731 | }); | | |
| 732 | | | |
| 733 | // Remove single-cell tables | | |
| 734 | this._forEachNode(this._getAllNodesWithTag(articleContent, ["table"]), function(table) { | | |
| 735 | var tbody = this._hasSingleTagInsideElement(table, "TBODY") ? table.firstElementChild : table; | | |
| 736 | if (this._hasSingleTagInsideElement(tbody, "TR")) { | | |
| 737 | var row = tbody.firstElementChild; | | |
| 738 | if (this._hasSingleTagInsideElement(row, "TD")) { | | |
| 739 | var cell = row.firstElementChild; | | |
| 740 | cell = this._setNodeTag(cell, this._everyNode(cell.childNodes, this._isPhrasingContent) ? "P" : "DIV"); | | |
| 741 | table.parentNode.replaceChild(cell, table); | | |
| 742 | } | | |
| 743 | } | | |
| 744 | }); | | |
| 745 | }, | | |
| 746 | | | |
| 747 | /** | | |
| 748 | * Initialize a node with the readability object. Also checks the | | |
| 749 | * className/id for special names to add to its score. | | |
| 750 | * | | |
| 751 | * @param Element | | |
| 752 | * @return void | | |
| 753 | **/ | | |
| 754 | _initializeNode: function(node) { | | |
| 755 | node.readability = {"contentScore": 0}; | | |
| 756 | | | |
| 757 | switch (node.tagName) { | | |
| 758 | case "DIV": | | |
| 759 | node.readability.contentScore += 5; | | |
| 760 | break; | | |
| 761 | | | |
| 762 | case "PRE": | | |
| 763 | case "TD": | | |
| 764 | case "BLOCKQUOTE": | | |
| 765 | node.readability.contentScore += 3; | | |
| 766 | break; | | |
| 767 | | | |
| 768 | case "ADDRESS": | | |
| 769 | case "OL": | | |
| 770 | case "UL": | | |
| 771 | case "DL": | | |
| 772 | case "DD": | | |
| 773 | case "DT": | | |
| 774 | case "LI": | | |
| 775 | case "FORM": | | |
| 776 | node.readability.contentScore -= 3; | | |
| 777 | break; | | |
| 778 | | | |
| 779 | case "H1": | | |
| 780 | case "H2": | | |
| 781 | case "H3": | | |
| 782 | case "H4": | | |
| 783 | case "H5": | | |
| 784 | case "H6": | | |
| 785 | case "TH": | | |
| 786 | node.readability.contentScore -= 5; | | |
| 787 | break; | | |
| 788 | } | | |
| 789 | | | |
| 790 | node.readability.contentScore += this._getClassWeight(node); | | |
| 791 | }, | | |
| 792 | | | |
| 793 | _removeAndGetNext: function(node) { | | |
| 794 | var nextNode = this._getNextNode(node, true); | | |
| 795 | node.parentNode.removeChild(node); | | |
| 796 | return nextNode; | | |
| 797 | }, | | |
| 798 | | | |
| 799 | /** | | |
| 800 | * Traverse the DOM from node to node, starting at the node passed in. | | |
| 801 | * Pass true for the second parameter to indicate this node itself | | |
| 802 | * (and its kids) are going away, and we want the next node over. | | |
| 803 | * | | |
| 804 | * Calling this in a loop will traverse the DOM depth-first. | | |
| 805 | */ | | |
| 806 | _getNextNode: function(node, ignoreSelfAndKids) { | | |
| 807 | // First check for kids if those aren't being ignored | | |
| 808 | if (!ignoreSelfAndKids && node.firstElementChild) { | | |
| 809 | return node.firstElementChild; | | |
| 810 | } | | |
| 811 | // Then for siblings... | | |
| 812 | if (node.nextElementSibling) { | | |
| 813 | return node.nextElementSibling; | | |
| 814 | } | | |
| 815 | // And finally, move up the parent chain *and* find a sibling | | |
| 816 | // (because this is depth-first traversal, we will have already | | |
| 817 | // seen the parent nodes themselves). | | |
| 818 | do { | | |
| 819 | node = node.parentNode; | | |
| 820 | } while (node && !node.nextElementSibling); | | |
| 821 | return node && node.nextElementSibling; | | |
| 822 | }, | | |
| 823 | | | |
| 824 | // compares second text to first one | | |
| 825 | // 1 = same text, 0 = completely different text | | |
| 826 | // works the way that it splits both texts into words and then finds words that are unique in second text | | |
| 827 | // the result is given by the lower length of unique parts | | |
| 828 | _textSimilarity: function(textA, textB) { | | |
| 829 | var tokensA = textA.toLowerCase().split(this.REGEXPS.tokenize).filter(Boolean); | | |
| 830 | var tokensB = textB.toLowerCase().split(this.REGEXPS.tokenize).filter(Boolean); | | |
| 831 | if (!tokensA.length || !tokensB.length) { | | |
| 832 | return 0; | | |
| 833 | } | | |
| 834 | var uniqTokensB = tokensB.filter(token => !tokensA.includes(token)); | | |
| 835 | var distanceB = uniqTokensB.join(" ").length / tokensB.join(" ").length; | | |
| 836 | return 1 - distanceB; | | |
| 837 | }, | | |
| 838 | | | |
| 839 | _checkByline: function(node, matchString) { | | |
| 840 | if (this._articleByline) { | | |
| 841 | return false; | | |
| 842 | } | | |
| 843 | | | |
| 844 | if (node.getAttribute !== undefined) { | | |
| 845 | var rel = node.getAttribute("rel"); | | |
| 846 | var itemprop = node.getAttribute("itemprop"); | | |
| 847 | } | | |
| 848 | | | |
| 849 | if ((rel === "author" || (itemprop && itemprop.indexOf("author") !== -1) || this.REGEXPS.byline.test(matchString)) && this._isValidByline(node.textContent)) { | | |
| 850 | this._articleByline = node.textContent.trim(); | | |
| 851 | return true; | | |
| 852 | } | | |
| 853 | | | |
| 854 | return false; | | |
| 855 | }, | | |
| 856 | | | |
| 857 | _getNodeAncestors: function(node, maxDepth) { | | |
| 858 | maxDepth = maxDepth || 0; | | |
| 859 | var i = 0, ancestors = []; | | |
| 860 | while (node.parentNode) { | | |
| 861 | ancestors.push(node.parentNode); | | |
| 862 | if (maxDepth && ++i === maxDepth) | | |
| 863 | break; | | |
| 864 | node = node.parentNode; | | |
| 865 | } | | |
| 866 | return ancestors; | | |
| 867 | }, | | |
| 868 | | | |
| 869 | /*** | | |
| 870 | * grabArticle - Using a variety of metrics (content score, classname, element types), find the content that is | | |
| 871 | * most likely to be the stuff a user wants to read. Then return it wrapped up in a div. | | |
| 872 | * | | |
| 873 | * @param page a document to run upon. Needs to be a full document, complete with body. | | |
| 874 | * @return Element | | |
| 875 | **/ | | |
| 876 | _grabArticle: function (page) { | | |
| 877 | this.log("**** grabArticle ****"); | | |
| 878 | var doc = this._doc; | | |
| 879 | var isPaging = page !== null; | | |
| 880 | page = page ? page : this._doc.body; | | |
| 881 | | | |
| 882 | // We can't grab an article if we don't have a page! | | |
| 883 | if (!page) { | | |
| 884 | this.log("No body found in document. Abort."); | | |
| 885 | return null; | | |
| 886 | } | | |
| 887 | | | |
| 888 | var pageCacheHtml = page.innerHTML; | | |
| 889 | | | |
| 890 | while (true) { | | |
| 891 | this.log("Starting grabArticle loop"); | | |
| 892 | var stripUnlikelyCandidates = this._flagIsActive(this.FLAG_STRIP_UNLIKELYS); | | |
| 893 | | | |
| 894 | // First, node prepping. Trash nodes that look cruddy (like ones with the | | |
| 895 | // class name "comment", etc), and turn divs into P tags where they have been | | |
| 896 | // used inappropriately (as in, where they contain no other block level elements.) | | |
| 897 | var elementsToScore = []; | | |
| 898 | var node = this._doc.documentElement; | | |
| 899 | | | |
| 900 | let shouldRemoveTitleHeader = true; | | |
| 901 | | | |
| 902 | while (node) { | | |
| 903 | | | |
| 904 | if (node.tagName === "HTML") { | | |
| 905 | this._articleLang = node.getAttribute("lang"); | | |
| 906 | } | | |
| 907 | | | |
| 908 | var matchString = node.className + " " + node.id; | | |
| 909 | | | |
| 910 | if (!this._isProbablyVisible(node)) { | | |
| 911 | this.log("Removing hidden node - " + matchString); | | |
| 912 | node = this._removeAndGetNext(node); | | |
| 913 | continue; | | |
| 914 | } | | |
| 915 | | | |
| 916 | // User is not able to see elements applied with both "aria-modal = true" and "role = dialog" | | |
| 917 | if (node.getAttribute("aria-modal") == "true" && node.getAttribute("role") == "dialog") { | | |
| 918 | node = this._removeAndGetNext(node); | | |
| 919 | continue; | | |
| 920 | } | | |
| 921 | | | |
| 922 | // Check to see if this node is a byline, and remove it if it is. | | |
| 923 | if (this._checkByline(node, matchString)) { | | |
| 924 | node = this._removeAndGetNext(node); | | |
| 925 | continue; | | |
| 926 | } | | |
| 927 | | | |
| 928 | if (shouldRemoveTitleHeader && this._headerDuplicatesTitle(node)) { | | |
| 929 | this.log("Removing header: ", node.textContent.trim(), this._articleTitle.trim()); | | |
| 930 | shouldRemoveTitleHeader = false; | | |
| 931 | node = this._removeAndGetNext(node); | | |
| 932 | continue; | | |
| 933 | } | | |
| 934 | | | |
| 935 | // Remove unlikely candidates | | |
| 936 | if (stripUnlikelyCandidates) { | | |
| 937 | if (this.REGEXPS.unlikelyCandidates.test(matchString) && | | |
| 938 | !this.REGEXPS.okMaybeItsACandidate.test(matchString) && | | |
| 939 | !this._hasAncestorTag(node, "table") && | | |
| 940 | !this._hasAncestorTag(node, "code") && | | |
| 941 | node.tagName !== "BODY" && | | |
| 942 | node.tagName !== "A") { | | |
| 943 | this.log("Removing unlikely candidate - " + matchString); | | |
| 944 | node = this._removeAndGetNext(node); | | |
| 945 | continue; | | |
| 946 | } | | |
| 947 | | | |
| 948 | if (this.UNLIKELY_ROLES.includes(node.getAttribute("role"))) { | | |
| 949 | this.log("Removing content with role " + node.getAttribute("role") + " - " + matchString); | | |
| 950 | node = this._removeAndGetNext(node); | | |
| 951 | continue; | | |
| 952 | } | | |
| 953 | } | | |
| 954 | | | |
| 955 | // Remove DIV, SECTION, and HEADER nodes without any content(e.g. text, image, video, or iframe). | | |
| 956 | if ((node.tagName === "DIV" || node.tagName === "SECTION" || node.tagName === "HEADER" || | | |
| 957 | node.tagName === "H1" || node.tagName === "H2" || node.tagName === "H3" || | | |
| 958 | node.tagName === "H4" || node.tagName === "H5" || node.tagName === "H6") && | | |
| 959 | this._isElementWithoutContent(node)) { | | |
| 960 | node = this._removeAndGetNext(node); | | |
| 961 | continue; | | |
| 962 | } | | |
| 963 | | | |
| 964 | if (this.DEFAULT_TAGS_TO_SCORE.indexOf(node.tagName) !== -1) { | | |
| 965 | elementsToScore.push(node); | | |
| 966 | } | | |
| 967 | | | |
| 968 | // Turn all divs that don't have children block level elements into p's | | |
| 969 | if (node.tagName === "DIV") { | | |
| 970 | // Put phrasing content into paragraphs. | | |
| 971 | var p = null; | | |
| 972 | var childNode = node.firstChild; | | |
| 973 | while (childNode) { | | |
| 974 | var nextSibling = childNode.nextSibling; | | |
| 975 | if (this._isPhrasingContent(childNode)) { | | |
| 976 | if (p !== null) { | | |
| 977 | p.appendChild(childNode); | | |
| 978 | } else if (!this._isWhitespace(childNode)) { | | |
| 979 | p = doc.createElement("p"); | | |
| 980 | node.replaceChild(p, childNode); | | |
| 981 | p.appendChild(childNode); | | |
| 982 | } | | |
| 983 | } else if (p !== null) { | | |
| 984 | while (p.lastChild && this._isWhitespace(p.lastChild)) { | | |
| 985 | p.removeChild(p.lastChild); | | |
| 986 | } | | |
| 987 | p = null; | | |
| 988 | } | | |
| 989 | childNode = nextSibling; | | |
| 990 | } | | |
| 991 | | | |
| 992 | // Sites like http://mobile.slate.com encloses each paragraph with a DIV | | |
| 993 | // element. DIVs with only a P element inside and no text content can be | | |
| 994 | // safely converted into plain P elements to avoid confusing the scoring | | |
| 995 | // algorithm with DIVs with are, in practice, paragraphs. | | |
| 996 | if (this._hasSingleTagInsideElement(node, "P") && this._getLinkDensity(node) < 0.25) { | | |
| 997 | var newNode = node.children[0]; | | |
| 998 | node.parentNode.replaceChild(newNode, node); | | |
| 999 | node = newNode; | | |
| 1000 | elementsToScore.push(node); | | |
| 1001 | } else if (!this._hasChildBlockElement(node)) { | | |
| 1002 | node = this._setNodeTag(node, "P"); | | |
| 1003 | elementsToScore.push(node); | | |
| 1004 | } | | |
| 1005 | } | | |
| 1006 | node = this._getNextNode(node); | | |
| 1007 | } | | |
| 1008 | | | |
| 1009 | /** | | |
| 1010 | * Loop through all paragraphs, and assign a score to them based on how content-y they look. | | |
| 1011 | * Then add their score to their parent node. | | |
| 1012 | * | | |
| 1013 | * A score is determined by things like number of commas, class names, etc. Maybe eventually link density. | | |
| 1014 | **/ | | |
| 1015 | var candidates = []; | | |
| 1016 | this._forEachNode(elementsToScore, function(elementToScore) { | | |
| 1017 | if (!elementToScore.parentNode || typeof(elementToScore.parentNode.tagName) === "undefined") | | |
| 1018 | return; | | |
| 1019 | | | |
| 1020 | // If this paragraph is less than 25 characters, don't even count it. | | |
| 1021 | var innerText = this._getInnerText(elementToScore); | | |
| 1022 | if (innerText.length < 25) | | |
| 1023 | return; | | |
| 1024 | | | |
| 1025 | // Exclude nodes with no ancestor. | | |
| 1026 | var ancestors = this._getNodeAncestors(elementToScore, 5); | | |
| 1027 | if (ancestors.length === 0) | | |
| 1028 | return; | | |
| 1029 | | | |
| 1030 | var contentScore = 0; | | |
| 1031 | | | |
| 1032 | // Add a point for the paragraph itself as a base. | | |
| 1033 | contentScore += 1; | | |
| 1034 | | | |
| 1035 | // Add points for any commas within this paragraph. | | |
| 1036 | contentScore += innerText.split(this.REGEXPS.commas).length; | | |
| 1037 | | | |
| 1038 | // For every 100 characters in this paragraph, add another point. Up to 3 points. | | |
| 1039 | contentScore += Math.min(Math.floor(innerText.length / 100), 3); | | |
| 1040 | | | |
| 1041 | // Initialize and score ancestors. | | |
| 1042 | this._forEachNode(ancestors, function(ancestor, level) { | | |
| 1043 | if (!ancestor.tagName || !ancestor.parentNode || typeof(ancestor.parentNode.tagName) === "undefined") | | |
| 1044 | return; | | |
| 1045 | | | |
| 1046 | if (typeof(ancestor.readability) === "undefined") { | | |
| 1047 | this._initializeNode(ancestor); | | |
| 1048 | candidates.push(ancestor); | | |
| 1049 | } | | |
| 1050 | | | |
| 1051 | // Node score divider: | | |
| 1052 | // - parent: 1 (no division) | | |
| 1053 | // - grandparent: 2 | | |
| 1054 | // - great grandparent+: ancestor level * 3 | | |
| 1055 | if (level === 0) | | |
| 1056 | var scoreDivider = 1; | | |
| 1057 | else if (level === 1) | | |
| 1058 | scoreDivider = 2; | | |
| 1059 | else | | |
| 1060 | scoreDivider = level * 3; | | |
| 1061 | ancestor.readability.contentScore += contentScore / scoreDivider; | | |
| 1062 | }); | | |
| 1063 | }); | | |
| 1064 | | | |
| 1065 | // After we've calculated scores, loop through all of the possible | | |
| 1066 | // candidate nodes we found and find the one with the highest score. | | |
| 1067 | var topCandidates = []; | | |
| 1068 | for (var c = 0, cl = candidates.length; c < cl; c += 1) { | | |
| 1069 | var candidate = candidates[c]; | | |
| 1070 | | | |
| 1071 | // Scale the final candidates score based on link density. Good content | | |
| 1072 | // should have a relatively small link density (5% or less) and be mostly | | |
| 1073 | // unaffected by this operation. | | |
| 1074 | var candidateScore = candidate.readability.contentScore * (1 - this._getLinkDensity(candidate)); | | |
| 1075 | candidate.readability.contentScore = candidateScore; | | |
| 1076 | | | |
| 1077 | this.log("Candidate:", candidate, "with score " + candidateScore); | | |
| 1078 | | | |
| 1079 | for (var t = 0; t < this._nbTopCandidates; t++) { | | |
| 1080 | var aTopCandidate = topCandidates[t]; | | |
| 1081 | | | |
| 1082 | if (!aTopCandidate || candidateScore > aTopCandidate.readability.contentScore) { | | |
| 1083 | topCandidates.splice(t, 0, candidate); | | |
| 1084 | if (topCandidates.length > this._nbTopCandidates) | | |
| 1085 | topCandidates.pop(); | | |
| 1086 | break; | | |
| 1087 | } | | |
| 1088 | } | | |
| 1089 | } | | |
| 1090 | | | |
| 1091 | var topCandidate = topCandidates[0] || null; | | |
| 1092 | var neededToCreateTopCandidate = false; | | |
| 1093 | var parentOfTopCandidate; | | |
| 1094 | | | |
| 1095 | // If we still have no top candidate, just use the body as a last resort. | | |
| 1096 | // We also have to copy the body node so it is something we can modify. | | |
| 1097 | if (topCandidate === null || topCandidate.tagName === "BODY") { | | |
| 1098 | // Move all of the page's children into topCandidate | | |
| 1099 | topCandidate = doc.createElement("DIV"); | | |
| 1100 | neededToCreateTopCandidate = true; | | |
| 1101 | // Move everything (not just elements, also text nodes etc.) into the container | | |
| 1102 | // so we even include text directly in the body: | | |
| 1103 | while (page.firstChild) { | | |
| 1104 | this.log("Moving child out:", page.firstChild); | | |
| 1105 | topCandidate.appendChild(page.firstChild); | | |
| 1106 | } | | |
| 1107 | | | |
| 1108 | page.appendChild(topCandidate); | | |
| 1109 | | | |
| 1110 | this._initializeNode(topCandidate); | | |
| 1111 | } else if (topCandidate) { | | |
| 1112 | // Find a better top candidate node if it contains (at least three) nodes which belong to `topCandidates` array | | |
| 1113 | // and whose scores are quite closed with current `topCandidate` node. | | |
| 1114 | var alternativeCandidateAncestors = []; | | |
| 1115 | for (var i = 1; i < topCandidates.length; i++) { | | |
| 1116 | if (topCandidates[i].readability.contentScore / topCandidate.readability.contentScore >= 0.75) { | | |
| 1117 | alternativeCandidateAncestors.push(this._getNodeAncestors(topCandidates[i])); | | |
| 1118 | } | | |
| 1119 | } | | |
| 1120 | var MINIMUM_TOPCANDIDATES = 3; | | |
| 1121 | if (alternativeCandidateAncestors.length >= MINIMUM_TOPCANDIDATES) { | | |
| 1122 | parentOfTopCandidate = topCandidate.parentNode; | | |
| 1123 | while (parentOfTopCandidate.tagName !== "BODY") { | | |
| 1124 | var listsContainingThisAncestor = 0; | | |
| 1125 | for (var ancestorIndex = 0; ancestorIndex < alternativeCandidateAncestors.length && listsContainingThisAncestor < MINIMUM_TOPCANDIDATES; ancestorIndex++) { | | |
| 1126 | listsContainingThisAncestor += Number(alternativeCandidateAncestors[ancestorIndex].includes(parentOfTopCandidate)); | | |
| 1127 | } | | |
| 1128 | if (listsContainingThisAncestor >= MINIMUM_TOPCANDIDATES) { | | |
| 1129 | topCandidate = parentOfTopCandidate; | | |
| 1130 | break; | | |
| 1131 | } | | |
| 1132 | parentOfTopCandidate = parentOfTopCandidate.parentNode; | | |
| 1133 | } | | |
| 1134 | } | | |
| 1135 | if (!topCandidate.readability) { | | |
| 1136 | this._initializeNode(topCandidate); | | |
| 1137 | } | | |
| 1138 | | | |
| 1139 | // Because of our bonus system, parents of candidates might have scores | | |
| 1140 | // themselves. They get half of the node. There won't be nodes with higher | | |
| 1141 | // scores than our topCandidate, but if we see the score going *up* in the first | | |
| 1142 | // few steps up the tree, that's a decent sign that there might be more content | | |
| 1143 | // lurking in other places that we want to unify in. The sibling stuff | | |
| 1144 | // below does some of that - but only if we've looked high enough up the DOM | | |
| 1145 | // tree. | | |
| 1146 | parentOfTopCandidate = topCandidate.parentNode; | | |
| 1147 | var lastScore = topCandidate.readability.contentScore; | | |
| 1148 | // The scores shouldn't get too low. | | |
| 1149 | var scoreThreshold = lastScore / 3; | | |
| 1150 | while (parentOfTopCandidate.tagName !== "BODY") { | | |
| 1151 | if (!parentOfTopCandidate.readability) { | | |
| 1152 | parentOfTopCandidate = parentOfTopCandidate.parentNode; | | |
| 1153 | continue; | | |
| 1154 | } | | |
| 1155 | var parentScore = parentOfTopCandidate.readability.contentScore; | | |
| 1156 | if (parentScore < scoreThreshold) | | |
| 1157 | break; | | |
| 1158 | if (parentScore > lastScore) { | | |
| 1159 | // Alright! We found a better parent to use. | | |
| 1160 | topCandidate = parentOfTopCandidate; | | |
| 1161 | break; | | |
| 1162 | } | | |
| 1163 | lastScore = parentOfTopCandidate.readability.contentScore; | | |
| 1164 | parentOfTopCandidate = parentOfTopCandidate.parentNode; | | |
| 1165 | } | | |
| 1166 | | | |
| 1167 | // If the top candidate is the only child, use parent instead. This will help sibling | | |
| 1168 | // joining logic when adjacent content is actually located in parent's sibling node. | | |
| 1169 | parentOfTopCandidate = topCandidate.parentNode; | | |
| 1170 | while (parentOfTopCandidate.tagName != "BODY" && parentOfTopCandidate.children.length == 1) { | | |
| 1171 | topCandidate = parentOfTopCandidate; | | |
| 1172 | parentOfTopCandidate = topCandidate.parentNode; | | |
| 1173 | } | | |
| 1174 | if (!topCandidate.readability) { | | |
| 1175 | this._initializeNode(topCandidate); | | |
| 1176 | } | | |
| 1177 | } | | |
| 1178 | | | |
| 1179 | // Now that we have the top candidate, look through its siblings for content | | |
| 1180 | // that might also be related. Things like preambles, content split by ads | | |
| 1181 | // that we removed, etc. | | |
| 1182 | var articleContent = doc.createElement("DIV"); | | |
| 1183 | if (isPaging) | | |
| 1184 | articleContent.id = "readability-content"; | | |
| 1185 | | | |
| 1186 | var siblingScoreThreshold = Math.max(10, topCandidate.readability.contentScore * 0.2); | | |
| 1187 | // Keep potential top candidate's parent node to try to get text direction of it later. | | |
| 1188 | parentOfTopCandidate = topCandidate.parentNode; | | |
| 1189 | var siblings = parentOfTopCandidate.children; | | |
| 1190 | | | |
| 1191 | for (var s = 0, sl = siblings.length; s < sl; s++) { | | |
| 1192 | var sibling = siblings[s]; | | |
| 1193 | var append = false; | | |
| 1194 | | | |
| 1195 | this.log("Looking at sibling node:", sibling, sibling.readability ? ("with score " + sibling.readability.contentScore) : ""); | | |
| 1196 | this.log("Sibling has score", sibling.readability ? sibling.readability.contentScore : "Unknown"); | | |
| 1197 | | | |
| 1198 | if (sibling === topCandidate) { | | |
| 1199 | append = true; | | |
| 1200 | } else { | | |
| 1201 | var contentBonus = 0; | | |
| 1202 | | | |
| 1203 | // Give a bonus if sibling nodes and top candidates have the example same classname | | |
| 1204 | if (sibling.className === topCandidate.className && topCandidate.className !== "") | | |
| 1205 | contentBonus += topCandidate.readability.contentScore * 0.2; | | |
| 1206 | | | |
| 1207 | if (sibling.readability && | | |
| 1208 | ((sibling.readability.contentScore + contentBonus) >= siblingScoreThreshold)) { | | |
| 1209 | append = true; | | |
| 1210 | } else if (sibling.nodeName === "P") { | | |
| 1211 | var linkDensity = this._getLinkDensity(sibling); | | |
| 1212 | var nodeContent = this._getInnerText(sibling); | | |
| 1213 | var nodeLength = nodeContent.length; | | |
| 1214 | | | |
| 1215 | if (nodeLength > 80 && linkDensity < 0.25) { | | |
| 1216 | append = true; | | |
| 1217 | } else if (nodeLength < 80 && nodeLength > 0 && linkDensity === 0 && | | |
| 1218 | nodeContent.search(/\.( |$)/) !== -1) { | | |
| 1219 | append = true; | | |
| 1220 | } | | |
| 1221 | } | | |
| 1222 | } | | |
| 1223 | | | |
| 1224 | if (append) { | | |
| 1225 | this.log("Appending node:", sibling); | | |
| 1226 | | | |
| 1227 | if (this.ALTER_TO_DIV_EXCEPTIONS.indexOf(sibling.nodeName) === -1) { | | |
| 1228 | // We have a node that isn't a common block level element, like a form or td tag. | | |
| 1229 | // Turn it into a div so it doesn't get filtered out later by accident. | | |
| 1230 | this.log("Altering sibling:", sibling, "to div."); | | |
| 1231 | | | |
| 1232 | sibling = this._setNodeTag(sibling, "DIV"); | | |
| 1233 | } | | |
| 1234 | | | |
| 1235 | articleContent.appendChild(sibling); | | |
| 1236 | // Fetch children again to make it compatible | | |
| 1237 | // with DOM parsers without live collection support. | | |
| 1238 | siblings = parentOfTopCandidate.children; | | |
| 1239 | // siblings is a reference to the children array, and | | |
| 1240 | // sibling is removed from the array when we call appendChild(). | | |
| 1241 | // As a result, we must revisit this index since the nodes | | |
| 1242 | // have been shifted. | | |
| 1243 | s -= 1; | | |
| 1244 | sl -= 1; | | |
| 1245 | } | | |
| 1246 | } | | |
| 1247 | | | |
| 1248 | if (this._debug) | | |
| 1249 | this.log("Article content pre-prep: " + articleContent.innerHTML); | | |
| 1250 | // So we have all of the content that we need. Now we clean it up for presentation. | | |
| 1251 | this._prepArticle(articleContent); | | |
| 1252 | if (this._debug) | | |
| 1253 | this.log("Article content post-prep: " + articleContent.innerHTML); | | |
| 1254 | | | |
| 1255 | if (neededToCreateTopCandidate) { | | |
| 1256 | // We already created a fake div thing, and there wouldn't have been any siblings left | | |
| 1257 | // for the previous loop, so there's no point trying to create a new div, and then | | |
| 1258 | // move all the children over. Just assign IDs and class names here. No need to append | | |
| 1259 | // because that already happened anyway. | | |
| 1260 | topCandidate.id = "readability-page-1"; | | |
| 1261 | topCandidate.className = "page"; | | |
| 1262 | } else { | | |
| 1263 | var div = doc.createElement("DIV"); | | |
| 1264 | div.id = "readability-page-1"; | | |
| 1265 | div.className = "page"; | | |
| 1266 | while (articleContent.firstChild) { | | |
| 1267 | div.appendChild(articleContent.firstChild); | | |
| 1268 | } | | |
| 1269 | articleContent.appendChild(div); | | |
| 1270 | } | | |
| 1271 | | | |
| 1272 | if (this._debug) | | |
| 1273 | this.log("Article content after paging: " + articleContent.innerHTML); | | |
| 1274 | | | |
| 1275 | var parseSuccessful = true; | | |
| 1276 | | | |
| 1277 | // Now that we've gone through the full algorithm, check to see if | | |
| 1278 | // we got any meaningful content. If we didn't, we may need to re-run | | |
| 1279 | // grabArticle with different flags set. This gives us a higher likelihood of | | |
| 1280 | // finding the content, and the sieve approach gives us a higher likelihood of | | |
| 1281 | // finding the -right- content. | | |
| 1282 | var textLength = this._getInnerText(articleContent, true).length; | | |
| 1283 | if (textLength < this._charThreshold) { | | |
| 1284 | parseSuccessful = false; | | |
| 1285 | page.innerHTML = pageCacheHtml; | | |
| 1286 | | | |
| 1287 | if (this._flagIsActive(this.FLAG_STRIP_UNLIKELYS)) { | | |
| 1288 | this._removeFlag(this.FLAG_STRIP_UNLIKELYS); | | |
| 1289 | this._attempts.push({articleContent: articleContent, textLength: textLength}); | | |
| 1290 | } else if (this._flagIsActive(this.FLAG_WEIGHT_CLASSES)) { | | |
| 1291 | this._removeFlag(this.FLAG_WEIGHT_CLASSES); | | |
| 1292 | this._attempts.push({articleContent: articleContent, textLength: textLength}); | | |
| 1293 | } else if (this._flagIsActive(this.FLAG_CLEAN_CONDITIONALLY)) { | | |
| 1294 | this._removeFlag(this.FLAG_CLEAN_CONDITIONALLY); | | |
| 1295 | this._attempts.push({articleContent: articleContent, textLength: textLength}); | | |
| 1296 | } else { | | |
| 1297 | this._attempts.push({articleContent: articleContent, textLength: textLength}); | | |
| 1298 | // No luck after removing flags, just return the longest text we found during the different loops | | |
| 1299 | this._attempts.sort(function (a, b) { | | |
| 1300 | return b.textLength - a.textLength; | | |
| 1301 | }); | | |
| 1302 | | | |
| 1303 | // But first check if we actually have something | | |
| 1304 | if (!this._attempts[0].textLength) { | | |
| 1305 | return null; | | |
| 1306 | } | | |
| 1307 | | | |
| 1308 | articleContent = this._attempts[0].articleContent; | | |
| 1309 | parseSuccessful = true; | | |
| 1310 | } | | |
| 1311 | } | | |
| 1312 | | | |
| 1313 | if (parseSuccessful) { | | |
| 1314 | // Find out text direction from ancestors of final top candidate. | | |
| 1315 | var ancestors = [parentOfTopCandidate, topCandidate].concat(this._getNodeAncestors(parentOfTopCandidate)); | | |
| 1316 | this._someNode(ancestors, function(ancestor) { | | |
| 1317 | if (!ancestor.tagName) | | |
| 1318 | return false; | | |
| 1319 | var articleDir = ancestor.getAttribute("dir"); | | |
| 1320 | if (articleDir) { | | |
| 1321 | this._articleDir = articleDir; | | |
| 1322 | return true; | | |
| 1323 | } | | |
| 1324 | return false; | | |
| 1325 | }); | | |
| 1326 | return articleContent; | | |
| 1327 | } | | |
| 1328 | } | | |
| 1329 | }, | | |
| 1330 | | | |
| 1331 | /** | | |
| 1332 | * Check whether the input string could be a byline. | | |
| 1333 | * This verifies that the input is a string, and that the length | | |
| 1334 | * is less than 100 chars. | | |
| 1335 | * | | |
| 1336 | * @param possibleByline {string} - a string to check whether its a byline. | | |
| 1337 | * @return Boolean - whether the input string is a byline. | | |
| 1338 | */ | | |
| 1339 | _isValidByline: function(byline) { | | |
| 1340 | if (typeof byline == "string" || byline instanceof String) { | | |
| 1341 | byline = byline.trim(); | | |
| 1342 | return (byline.length > 0) && (byline.length < 100); | | |
| 1343 | } | | |
| 1344 | return false; | | |
| 1345 | }, | | |
| 1346 | | | |
| 1347 | /** | | |
| 1348 | * Converts some of the common HTML entities in string to their corresponding characters. | | |
| 1349 | * | | |
| 1350 | * @param str {string} - a string to unescape. | | |
| 1351 | * @return string without HTML entity. | | |
| 1352 | */ | | |
| 1353 | _unescapeHtmlEntities: function(str) { | | |
| 1354 | if (!str) { | | |
| 1355 | return str; | | |
| 1356 | } | | |
| 1357 | | | |
| 1358 | var htmlEscapeMap = this.HTML_ESCAPE_MAP; | | |
| 1359 | return str.replace(/&(quot|amp|apos|lt|gt);/g, function(_, tag) { | | |
| 1360 | return htmlEscapeMap[tag]; | | |
| 1361 | }).replace(/&#(?:x([0-9a-z]{1,4})|([0-9]{1,4}));/gi, function(_, hex, numStr) { | | |
| 1362 | var num = parseInt(hex || numStr, hex ? 16 : 10); | | |
| 1363 | return String.fromCharCode(num); | | |
| 1364 | }); | | |
| 1365 | }, | | |
| 1366 | | | |
| 1367 | /** | | |
| 1368 | * Try to extract metadata from JSON-LD object. | | |
| 1369 | * For now, only Schema.org objects of type Article or its subtypes are supported. | | |
| 1370 | * @return Object with any metadata that could be extracted (possibly none) | | |
| 1371 | */ | | |
| 1372 | _getJSONLD: function (doc) { | | |
| 1373 | var scripts = this._getAllNodesWithTag(doc, ["script"]); | | |
| 1374 | | | |
| 1375 | var metadata; | | |
| 1376 | | | |
| 1377 | this._forEachNode(scripts, function(jsonLdElement) { | | |
| 1378 | if (!metadata && jsonLdElement.getAttribute("type") === "application/ld+json") { | | |
| 1379 | try { | | |
| 1380 | // Strip CDATA markers if present | | |
| 1381 | var content = jsonLdElement.textContent.replace(/^\s*<!\[CDATA\[|\]\]>\s*$/g, ""); | | |
| 1382 | var parsed = JSON.parse(content); | | |
| 1383 | if ( | | |
| 1384 | !parsed["@context"] || | | |
| 1385 | !parsed["@context"].match(/^https?\:\/\/schema\.org$/) | | |
| 1386 | ) { | | |
| 1387 | return; | | |
| 1388 | } | | |
| 1389 | | | |
| 1390 | if (!parsed["@type"] && Array.isArray(parsed["@graph"])) { | | |
| 1391 | parsed = parsed["@graph"].find(function(it) { | | |
| 1392 | return (it["@type"] || "").match( | | |
| 1393 | this.REGEXPS.jsonLdArticleTypes | | |
| 1394 | ); | | |
| 1395 | }); | | |
| 1396 | } | | |
| 1397 | | | |
| 1398 | if ( | | |
| 1399 | !parsed || | | |
| 1400 | !parsed["@type"] || | | |
| 1401 | !parsed["@type"].match(this.REGEXPS.jsonLdArticleTypes) | | |
| 1402 | ) { | | |
| 1403 | return; | | |
| 1404 | } | | |
| 1405 | | | |
| 1406 | metadata = {}; | | |
| 1407 | | | |
| 1408 | if (typeof parsed.name === "string" && typeof parsed.headline === "string" && parsed.name !== parsed.headline) { | | |
| 1409 | // we have both name and headline element in the JSON-LD. They should both be the same but some websites like aktualne.cz | | |
| 1410 | // put their own name into "name" and the article title to "headline" which confuses Readability. So we try to check if either | | |
| 1411 | // "name" or "headline" closely matches the html title, and if so, use that one. If not, then we use "name" by default. | | |
| 1412 | | | |
| 1413 | var title = this._getArticleTitle(); | | |
| 1414 | var nameMatches = this._textSimilarity(parsed.name, title) > 0.75; | | |
| 1415 | var headlineMatches = this._textSimilarity(parsed.headline, title) > 0.75; | | |
| 1416 | | | |
| 1417 | if (headlineMatches && !nameMatches) { | | |
| 1418 | metadata.title = parsed.headline; | | |
| 1419 | } else { | | |
| 1420 | metadata.title = parsed.name; | | |
| 1421 | } | | |
| 1422 | } else if (typeof parsed.name === "string") { | | |
| 1423 | metadata.title = parsed.name.trim(); | | |
| 1424 | } else if (typeof parsed.headline === "string") { | | |
| 1425 | metadata.title = parsed.headline.trim(); | | |
| 1426 | } | | |
| 1427 | if (parsed.author) { | | |
| 1428 | if (typeof parsed.author.name === "string") { | | |
| 1429 | metadata.byline = parsed.author.name.trim(); | | |
| 1430 | } else if (Array.isArray(parsed.author) && parsed.author[0] && typeof parsed.author[0].name === "string") { | | |
| 1431 | metadata.byline = parsed.author | | |
| 1432 | .filter(function(author) { | | |
| 1433 | return author && typeof author.name === "string"; | | |
| 1434 | }) | | |
| 1435 | .map(function(author) { | | |
| 1436 | return author.name.trim(); | | |
| 1437 | }) | | |
| 1438 | .join(", "); | | |
| 1439 | } | | |
| 1440 | } | | |
| 1441 | if (typeof parsed.description === "string") { | | |
| 1442 | metadata.excerpt = parsed.description.trim(); | | |
| 1443 | } | | |
| 1444 | if ( | | |
| 1445 | parsed.publisher && | | |
| 1446 | typeof parsed.publisher.name === "string" | | |
| 1447 | ) { | | |
| 1448 | metadata.siteName = parsed.publisher.name.trim(); | | |
| 1449 | } | | |
| 1450 | if (typeof parsed.datePublished === "string") { | | |
| 1451 | metadata.datePublished = parsed.datePublished.trim(); | | |
| 1452 | } | | |
| 1453 | return; | | |
| 1454 | } catch (err) { | | |
| 1455 | this.log(err.message); | | |
| 1456 | } | | |
| 1457 | } | | |
| 1458 | }); | | |
| 1459 | return metadata ? metadata : {}; | | |
| 1460 | }, | | |
| 1461 | | | |
| 1462 | /** | | |
| 1463 | * Attempts to get excerpt and byline metadata for the article. | | |
| 1464 | * | | |
| 1465 | * @param {Object} jsonld — object containing any metadata that | | |
| 1466 | * could be extracted from JSON-LD object. | | |
| 1467 | * | | |
| 1468 | * @return Object with optional "excerpt" and "byline" properties | | |
| 1469 | */ | | |
| 1470 | _getArticleMetadata: function(jsonld) { | | |
| 1471 | var metadata = {}; | | |
| 1472 | var values = {}; | | |
| 1473 | var metaElements = this._doc.getElementsByTagName("meta"); | | |
| 1474 | | | |
| 1475 | // property is a space-separated list of values | | |
| 1476 | var propertyPattern = /\s*(article|dc|dcterm|og|twitter)\s*:\s*(author|creator|description|published_time|title|site_name)\s*/gi; | | |
| 1477 | | | |
| 1478 | // name is a single value | | |
| 1479 | var namePattern = /^\s*(?:(dc|dcterm|og|twitter|weibo:(article|webpage))\s*[\.:]\s*)?(author|creator|description|title|site_name)\s*$/i; | | |
| 1480 | | | |
| 1481 | // Find description tags. | | |
| 1482 | this._forEachNode(metaElements, function(element) { | | |
| 1483 | var elementName = element.getAttribute("name"); | | |
| 1484 | var elementProperty = element.getAttribute("property"); | | |
| 1485 | var content = element.getAttribute("content"); | | |
| 1486 | if (!content) { | | |
| 1487 | return; | | |
| 1488 | } | | |
| 1489 | var matches = null; | | |
| 1490 | var name = null; | | |
| 1491 | | | |
| 1492 | if (elementProperty) { | | |
| 1493 | matches = elementProperty.match(propertyPattern); | | |
| 1494 | if (matches) { | | |
| 1495 | // Convert to lowercase, and remove any whitespace | | |
| 1496 | // so we can match below. | | |
| 1497 | name = matches[0].toLowerCase().replace(/\s/g, ""); | | |
| 1498 | // multiple authors | | |
| 1499 | values[name] = content.trim(); | | |
| 1500 | } | | |
| 1501 | } | | |
| 1502 | if (!matches && elementName && namePattern.test(elementName)) { | | |
| 1503 | name = elementName; | | |
| 1504 | if (content) { | | |
| 1505 | // Convert to lowercase, remove any whitespace, and convert dots | | |
| 1506 | // to colons so we can match below. | | |
| 1507 | name = name.toLowerCase().replace(/\s/g, "").replace(/\./g, ":"); | | |
| 1508 | values[name] = content.trim(); | | |
| 1509 | } | | |
| 1510 | } | | |
| 1511 | }); | | |
| 1512 | | | |
| 1513 | // get title | | |
| 1514 | metadata.title = jsonld.title || | | |
| 1515 | values["dc:title"] || | | |
| 1516 | values["dcterm:title"] || | | |
| 1517 | values["og:title"] || | | |
| 1518 | values["weibo:article:title"] || | | |
| 1519 | values["weibo:webpage:title"] || | | |
| 1520 | values["title"] || | | |
| 1521 | values["twitter:title"]; | | |
| 1522 | | | |
| 1523 | if (!metadata.title) { | | |
| 1524 | metadata.title = this._getArticleTitle(); | | |
| 1525 | } | | |
| 1526 | | | |
| 1527 | // get author | | |
| 1528 | metadata.byline = jsonld.byline || | | |
| 1529 | values["dc:creator"] || | | |
| 1530 | values["dcterm:creator"] || | | |
| 1531 | values["author"]; | | |
| 1532 | | | |
| 1533 | // get description | | |
| 1534 | metadata.excerpt = jsonld.excerpt || | | |
| 1535 | values["dc:description"] || | | |
| 1536 | values["dcterm:description"] || | | |
| 1537 | values["og:description"] || | | |
| 1538 | values["weibo:article:description"] || | | |
| 1539 | values["weibo:webpage:description"] || | | |
| 1540 | values["description"] || | | |
| 1541 | values["twitter:description"]; | | |
| 1542 | | | |
| 1543 | // get site name | | |
| 1544 | metadata.siteName = jsonld.siteName || | | |
| 1545 | values["og:site_name"]; | | |
| 1546 | | | |
| 1547 | // get article published time | | |
| 1548 | metadata.publishedTime = jsonld.datePublished || | | |
| 1549 | values["article:published_time"] || null; | | |
| 1550 | | | |
| 1551 | // in many sites the meta value is escaped with HTML entities, | | |
| 1552 | // so here we need to unescape it | | |
| 1553 | metadata.title = this._unescapeHtmlEntities(metadata.title); | | |
| 1554 | metadata.byline = this._unescapeHtmlEntities(metadata.byline); | | |
| 1555 | metadata.excerpt = this._unescapeHtmlEntities(metadata.excerpt); | | |
| 1556 | metadata.siteName = this._unescapeHtmlEntities(metadata.siteName); | | |
| 1557 | metadata.publishedTime = this._unescapeHtmlEntities(metadata.publishedTime); | | |
| 1558 | | | |
| 1559 | return metadata; | | |
| 1560 | }, | | |
| 1561 | | | |
| 1562 | /** | | |
| 1563 | * Check if node is image, or if node contains exactly only one image | | |
| 1564 | * whether as a direct child or as its descendants. | | |
| 1565 | * | | |
| 1566 | * @param Element | | |
| 1567 | **/ | | |
| 1568 | _isSingleImage: function(node) { | | |
| 1569 | if (node.tagName === "IMG") { | | |
| 1570 | return true; | | |
| 1571 | } | | |
| 1572 | | | |
| 1573 | if (node.children.length !== 1 || node.textContent.trim() !== "") { | | |
| 1574 | return false; | | |
| 1575 | } | | |
| 1576 | | | |
| 1577 | return this._isSingleImage(node.children[0]); | | |
| 1578 | }, | | |
| 1579 | | | |
| 1580 | /** | | |
| 1581 | * Find all <noscript> that are located after <img> nodes, and which contain only one | | |
| 1582 | * <img> element. Replace the first image with the image from inside the <noscript> tag, | | |
| 1583 | * and remove the <noscript> tag. This improves the quality of the images we use on | | |
| 1584 | * some sites (e.g. Medium). | | |
| 1585 | * | | |
| 1586 | * @param Element | | |
| 1587 | **/ | | |
| 1588 | _unwrapNoscriptImages: function(doc) { | | |
| 1589 | // Find img without source or attributes that might contains image, and remove it. | | |
| 1590 | // This is done to prevent a placeholder img is replaced by img from noscript in next step. | | |
| 1591 | var imgs = Array.from(doc.getElementsByTagName("img")); | | |
| 1592 | this._forEachNode(imgs, function(img) { | | |
| 1593 | for (var i = 0; i < img.attributes.length; i++) { | | |
| 1594 | var attr = img.attributes[i]; | | |
| 1595 | switch (attr.name) { | | |
| 1596 | case "src": | | |
| 1597 | case "srcset": | | |
| 1598 | case "data-src": | | |
| 1599 | case "data-srcset": | | |
| 1600 | return; | | |
| 1601 | } | | |
| 1602 | | | |
| 1603 | if (/\.(jpg|jpeg|png|webp)/i.test(attr.value)) { | | |
| 1604 | return; | | |
| 1605 | } | | |
| 1606 | } | | |
| 1607 | | | |
| 1608 | img.parentNode.removeChild(img); | | |
| 1609 | }); | | |
| 1610 | | | |
| 1611 | // Next find noscript and try to extract its image | | |
| 1612 | var noscripts = Array.from(doc.getElementsByTagName("noscript")); | | |
| 1613 | this._forEachNode(noscripts, function(noscript) { | | |
| 1614 | // Parse content of noscript and make sure it only contains image | | |
| 1615 | var tmp = doc.createElement("div"); | | |
| 1616 | tmp.innerHTML = noscript.innerHTML; | | |
| 1617 | if (!this._isSingleImage(tmp)) { | | |
| 1618 | return; | | |
| 1619 | } | | |
| 1620 | | | |
| 1621 | // If noscript has previous sibling and it only contains image, | | |
| 1622 | // replace it with noscript content. However we also keep old | | |
| 1623 | // attributes that might contains image. | | |
| 1624 | var prevElement = noscript.previousElementSibling; | | |
| 1625 | if (prevElement && this._isSingleImage(prevElement)) { | | |
| 1626 | var prevImg = prevElement; | | |
| 1627 | if (prevImg.tagName !== "IMG") { | | |
| 1628 | prevImg = prevElement.getElementsByTagName("img")[0]; | | |
| 1629 | } | | |
| 1630 | | | |
| 1631 | var newImg = tmp.getElementsByTagName("img")[0]; | | |
| 1632 | for (var i = 0; i < prevImg.attributes.length; i++) { | | |
| 1633 | var attr = prevImg.attributes[i]; | | |
| 1634 | if (attr.value === "") { | | |
| 1635 | continue; | | |
| 1636 | } | | |
| 1637 | | | |
| 1638 | if (attr.name === "src" || attr.name === "srcset" || /\.(jpg|jpeg|png|webp)/i.test(attr.value)) { | | |
| 1639 | if (newImg.getAttribute(attr.name) === attr.value) { | | |
| 1640 | continue; | | |
| 1641 | } | | |
| 1642 | | | |
| 1643 | var attrName = attr.name; | | |
| 1644 | if (newImg.hasAttribute(attrName)) { | | |
| 1645 | attrName = "data-old-" + attrName; | | |
| 1646 | } | | |
| 1647 | | | |
| 1648 | newImg.setAttribute(attrName, attr.value); | | |
| 1649 | } | | |
| 1650 | } | | |
| 1651 | | | |
| 1652 | noscript.parentNode.replaceChild(tmp.firstElementChild, prevElement); | | |
| 1653 | } | | |
| 1654 | }); | | |
| 1655 | }, | | |
| 1656 | | | |
| 1657 | /** | | |
| 1658 | * Removes script tags from the document. | | |
| 1659 | * | | |
| 1660 | * @param Element | | |
| 1661 | **/ | | |
| 1662 | _removeScripts: function(doc) { | | |
| 1663 | this._removeNodes(this._getAllNodesWithTag(doc, ["script", "noscript"])); | | |
| 1664 | }, | | |
| 1665 | | | |
| 1666 | /** | | |
| 1667 | * Check if this node has only whitespace and a single element with given tag | | |
| 1668 | * Returns false if the DIV node contains non-empty text nodes | | |
| 1669 | * or if it contains no element with given tag or more than 1 element. | | |
| 1670 | * | | |
| 1671 | * @param Element | | |
| 1672 | * @param string tag of child element | | |
| 1673 | **/ | | |
| 1674 | _hasSingleTagInsideElement: function(element, tag) { | | |
| 1675 | // There should be exactly 1 element child with given tag | | |
| 1676 | if (element.children.length != 1 || element.children[0].tagName !== tag) { | | |
| 1677 | return false; | | |
| 1678 | } | | |
| 1679 | | | |
| 1680 | // And there should be no text nodes with real content | | |
| 1681 | return !this._someNode(element.childNodes, function(node) { | | |
| 1682 | return node.nodeType === this.TEXT_NODE && | | |
| 1683 | this.REGEXPS.hasContent.test(node.textContent); | | |
| 1684 | }); | | |
| 1685 | }, | | |
| 1686 | | | |
| 1687 | _isElementWithoutContent: function(node) { | | |
| 1688 | return node.nodeType === this.ELEMENT_NODE && | | |
| 1689 | node.textContent.trim().length == 0 && | | |
| 1690 | (node.children.length == 0 || | | |
| 1691 | node.children.length == node.getElementsByTagName("br").length + node.getElementsByTagName("hr").length); | | |
| 1692 | }, | | |
| 1693 | | | |
| 1694 | /** | | |
| 1695 | * Determine whether element has any children block level elements. | | |
| 1696 | * | | |
| 1697 | * @param Element | | |
| 1698 | */ | | |
| 1699 | _hasChildBlockElement: function (element) { | | |
| 1700 | return this._someNode(element.childNodes, function(node) { | | |
| 1701 | return this.DIV_TO_P_ELEMS.has(node.tagName) || | | |
| 1702 | this._hasChildBlockElement(node); | | |
| 1703 | }); | | |
| 1704 | }, | | |
| 1705 | | | |
| 1706 | /*** | | |
| 1707 | * Determine if a node qualifies as phrasing content. | | |
| 1708 | * https://developer.mozilla.org/en-US/docs/Web/Guide/HTML/Content_categories#Phrasing_content | | |
| 1709 | **/ | | |
| 1710 | _isPhrasingContent: function(node) { | | |
| 1711 | return node.nodeType === this.TEXT_NODE || this.PHRASING_ELEMS.indexOf(node.tagName) !== -1 || | | |
| 1712 | ((node.tagName === "A" || node.tagName === "DEL" || node.tagName === "INS") && | | |
| 1713 | this._everyNode(node.childNodes, this._isPhrasingContent)); | | |
| 1714 | }, | | |
| 1715 | | | |
| 1716 | _isWhitespace: function(node) { | | |
| 1717 | return (node.nodeType === this.TEXT_NODE && node.textContent.trim().length === 0) || | | |
| 1718 | (node.nodeType === this.ELEMENT_NODE && node.tagName === "BR"); | | |
| 1719 | }, | | |
| 1720 | | | |
| 1721 | /** | | |
| 1722 | * Get the inner text of a node - cross browser compatibly. | | |
| 1723 | * This also strips out any excess whitespace to be found. | | |
| 1724 | * | | |
| 1725 | * @param Element | | |
| 1726 | * @param Boolean normalizeSpaces (default: true) | | |
| 1727 | * @return string | | |
| 1728 | **/ | | |
| 1729 | _getInnerText: function(e, normalizeSpaces) { | | |
| 1730 | normalizeSpaces = (typeof normalizeSpaces === "undefined") ? true : normalizeSpaces; | | |
| 1731 | var textContent = e.textContent.trim(); | | |
| 1732 | | | |
| 1733 | if (normalizeSpaces) { | | |
| 1734 | return textContent.replace(this.REGEXPS.normalize, " "); | | |
| 1735 | } | | |
| 1736 | return textContent; | | |
| 1737 | }, | | |
| 1738 | | | |
| 1739 | /** | | |
| 1740 | * Get the number of times a string s appears in the node e. | | |
| 1741 | * | | |
| 1742 | * @param Element | | |
| 1743 | * @param string - what to split on. Default is "," | | |
| 1744 | * @return number (integer) | | |
| 1745 | **/ | | |
| 1746 | _getCharCount: function(e, s) { | | |
| 1747 | s = s || ","; | | |
| 1748 | return this._getInnerText(e).split(s).length - 1; | | |
| 1749 | }, | | |
| 1750 | | | |
| 1751 | /** | | |
| 1752 | * Remove the style attribute on every e and under. | | |
| 1753 | * TODO: Test if getElementsByTagName(*) is faster. | | |
| 1754 | * | | |
| 1755 | * @param Element | | |
| 1756 | * @return void | | |
| 1757 | **/ | | |
| 1758 | _cleanStyles: function(e) { | | |
| 1759 | if (!e || e.tagName.toLowerCase() === "svg") | | |
| 1760 | return; | | |
| 1761 | | | |
| 1762 | // Remove `style` and deprecated presentational attributes | | |
| 1763 | for (var i = 0; i < this.PRESENTATIONAL_ATTRIBUTES.length; i++) { | | |
| 1764 | e.removeAttribute(this.PRESENTATIONAL_ATTRIBUTES[i]); | | |
| 1765 | } | | |
| 1766 | | | |
| 1767 | if (this.DEPRECATED_SIZE_ATTRIBUTE_ELEMS.indexOf(e.tagName) !== -1) { | | |
| 1768 | e.removeAttribute("width"); | | |
| 1769 | e.removeAttribute("height"); | | |
| 1770 | } | | |
| 1771 | | | |
| 1772 | var cur = e.firstElementChild; | | |
| 1773 | while (cur !== null) { | | |
| 1774 | this._cleanStyles(cur); | | |
| 1775 | cur = cur.nextElementSibling; | | |
| 1776 | } | | |
| 1777 | }, | | |
| 1778 | | | |
| 1779 | /** | | |
| 1780 | * Get the density of links as a percentage of the content | | |
| 1781 | * This is the amount of text that is inside a link divided by the total text in the node. | | |
| 1782 | * | | |
| 1783 | * @param Element | | |
| 1784 | * @return number (float) | | |
| 1785 | **/ | | |
| 1786 | _getLinkDensity: function(element) { | | |
| 1787 | var textLength = this._getInnerText(element).length; | | |
| 1788 | if (textLength === 0) | | |
| 1789 | return 0; | | |
| 1790 | | | |
| 1791 | var linkLength = 0; | | |
| 1792 | | | |
| 1793 | // XXX implement _reduceNodeList? | | |
| 1794 | this._forEachNode(element.getElementsByTagName("a"), function(linkNode) { | | |
| 1795 | var href = linkNode.getAttribute("href"); | | |
| 1796 | var coefficient = href && this.REGEXPS.hashUrl.test(href) ? 0.3 : 1; | | |
| 1797 | linkLength += this._getInnerText(linkNode).length * coefficient; | | |
| 1798 | }); | | |
| 1799 | | | |
| 1800 | return linkLength / textLength; | | |
| 1801 | }, | | |
| 1802 | | | |
| 1803 | /** | | |
| 1804 | * Get an elements class/id weight. Uses regular expressions to tell if this | | |
| 1805 | * element looks good or bad. | | |
| 1806 | * | | |
| 1807 | * @param Element | | |
| 1808 | * @return number (Integer) | | |
| 1809 | **/ | | |
| 1810 | _getClassWeight: function(e) { | | |
| 1811 | if (!this._flagIsActive(this.FLAG_WEIGHT_CLASSES)) | | |
| 1812 | return 0; | | |
| 1813 | | | |
| 1814 | var weight = 0; | | |
| 1815 | | | |
| 1816 | // Look for a special classname | | |
| 1817 | if (typeof(e.className) === "string" && e.className !== "") { | | |
| 1818 | if (this.REGEXPS.negative.test(e.className)) | | |
| 1819 | weight -= 25; | | |
| 1820 | | | |
| 1821 | if (this.REGEXPS.positive.test(e.className)) | | |
| 1822 | weight += 25; | | |
| 1823 | } | | |
| 1824 | | | |
| 1825 | // Look for a special ID | | |
| 1826 | if (typeof(e.id) === "string" && e.id !== "") { | | |
| 1827 | if (this.REGEXPS.negative.test(e.id)) | | |
| 1828 | weight -= 25; | | |
| 1829 | | | |
| 1830 | if (this.REGEXPS.positive.test(e.id)) | | |
| 1831 | weight += 25; | | |
| 1832 | } | | |
| 1833 | | | |
| 1834 | return weight; | | |
| 1835 | }, | | |
| 1836 | | | |
| 1837 | /** | | |
| 1838 | * Clean a node of all elements of type "tag". | | |
| 1839 | * (Unless it's a youtube/vimeo video. People love movies.) | | |
| 1840 | * | | |
| 1841 | * @param Element | | |
| 1842 | * @param string tag to clean | | |
| 1843 | * @return void | | |
| 1844 | **/ | | |
| 1845 | _clean: function(e, tag) { | | |
| 1846 | var isEmbed = ["object", "embed", "iframe"].indexOf(tag) !== -1; | | |
| 1847 | | | |
| 1848 | this._removeNodes(this._getAllNodesWithTag(e, [tag]), function(element) { | | |
| 1849 | // Allow youtube and vimeo videos through as people usually want to see those. | | |
| 1850 | if (isEmbed) { | | |
| 1851 | // First, check the elements attributes to see if any of them contain youtube or vimeo | | |
| 1852 | for (var i = 0; i < element.attributes.length; i++) { | | |
| 1853 | if (this._allowedVideoRegex.test(element.attributes[i].value)) { | | |
| 1854 | return false; | | |
| 1855 | } | | |
| 1856 | } | | |
| 1857 | | | |
| 1858 | // For embed with <object> tag, check inner HTML as well. | | |
| 1859 | if (element.tagName === "object" && this._allowedVideoRegex.test(element.innerHTML)) { | | |
| 1860 | return false; | | |
| 1861 | } | | |
| 1862 | } | | |
| 1863 | | | |
| 1864 | return true; | | |
| 1865 | }); | | |
| 1866 | }, | | |
| 1867 | | | |
| 1868 | /** | | |
| 1869 | * Check if a given node has one of its ancestor tag name matching the | | |
| 1870 | * provided one. | | |
| 1871 | * @param HTMLElement node | | |
| 1872 | * @param String tagName | | |
| 1873 | * @param Number maxDepth | | |
| 1874 | * @param Function filterFn a filter to invoke to determine whether this node 'counts' | | |
| 1875 | * @return Boolean | | |
| 1876 | */ | | |
| 1877 | _hasAncestorTag: function(node, tagName, maxDepth, filterFn) { | | |
| 1878 | maxDepth = maxDepth || 3; | | |
| 1879 | tagName = tagName.toUpperCase(); | | |
| 1880 | var depth = 0; | | |
| 1881 | while (node.parentNode) { | | |
| 1882 | if (maxDepth > 0 && depth > maxDepth) | | |
| 1883 | return false; | | |
| 1884 | if (node.parentNode.tagName === tagName && (!filterFn || filterFn(node.parentNode))) | | |
| 1885 | return true; | | |
| 1886 | node = node.parentNode; | | |
| 1887 | depth++; | | |
| 1888 | } | | |
| 1889 | return false; | | |
| 1890 | }, | | |
| 1891 | | | |
| 1892 | /** | | |
| 1893 | * Return an object indicating how many rows and columns this table has. | | |
| 1894 | */ | | |
| 1895 | _getRowAndColumnCount: function(table) { | | |
| 1896 | var rows = 0; | | |
| 1897 | var columns = 0; | | |
| 1898 | var trs = table.getElementsByTagName("tr"); | | |
| 1899 | for (var i = 0; i < trs.length; i++) { | | |
| 1900 | var rowspan = trs[i].getAttribute("rowspan") || 0; | | |
| 1901 | if (rowspan) { | | |
| 1902 | rowspan = parseInt(rowspan, 10); | | |
| 1903 | } | | |
| 1904 | rows += (rowspan || 1); | | |
| 1905 | | | |
| 1906 | // Now look for column-related info | | |
| 1907 | var columnsInThisRow = 0; | | |
| 1908 | var cells = trs[i].getElementsByTagName("td"); | | |
| 1909 | for (var j = 0; j < cells.length; j++) { | | |
| 1910 | var colspan = cells[j].getAttribute("colspan") || 0; | | |
| 1911 | if (colspan) { | | |
| 1912 | colspan = parseInt(colspan, 10); | | |
| 1913 | } | | |
| 1914 | columnsInThisRow += (colspan || 1); | | |
| 1915 | } | | |
| 1916 | columns = Math.max(columns, columnsInThisRow); | | |
| 1917 | } | | |
| 1918 | return {rows: rows, columns: columns}; | | |
| 1919 | }, | | |
| 1920 | | | |
| 1921 | /** | | |
| 1922 | * Look for 'data' (as opposed to 'layout') tables, for which we use | | |
| 1923 | * similar checks as | | |
| 1924 | * https://searchfox.org/mozilla-central/rev/f82d5c549f046cb64ce5602bfd894b7ae807c8f8/accessible/generic/TableAccessible.cpp#19 | | |
| 1925 | */ | | |
| 1926 | _markDataTables: function(root) { | | |
| 1927 | var tables = root.getElementsByTagName("table"); | | |
| 1928 | for (var i = 0; i < tables.length; i++) { | | |
| 1929 | var table = tables[i]; | | |
| 1930 | var role = table.getAttribute("role"); | | |
| 1931 | if (role == "presentation") { | | |
| 1932 | table._readabilityDataTable = false; | | |
| 1933 | continue; | | |
| 1934 | } | | |
| 1935 | var datatable = table.getAttribute("datatable"); | | |
| 1936 | if (datatable == "0") { | | |
| 1937 | table._readabilityDataTable = false; | | |
| 1938 | continue; | | |
| 1939 | } | | |
| 1940 | var summary = table.getAttribute("summary"); | | |
| 1941 | if (summary) { | | |
| 1942 | table._readabilityDataTable = true; | | |
| 1943 | continue; | | |
| 1944 | } | | |
| 1945 | | | |
| 1946 | var caption = table.getElementsByTagName("caption")[0]; | | |
| 1947 | if (caption && caption.childNodes.length > 0) { | | |
| 1948 | table._readabilityDataTable = true; | | |
| 1949 | continue; | | |
| 1950 | } | | |
| 1951 | | | |
| 1952 | // If the table has a descendant with any of these tags, consider a data table: | | |
| 1953 | var dataTableDescendants = ["col", "colgroup", "tfoot", "thead", "th"]; | | |
| 1954 | var descendantExists = function(tag) { | | |
| 1955 | return !!table.getElementsByTagName(tag)[0]; | | |
| 1956 | }; | | |
| 1957 | if (dataTableDescendants.some(descendantExists)) { | | |
| 1958 | this.log("Data table because found data-y descendant"); | | |
| 1959 | table._readabilityDataTable = true; | | |
| 1960 | continue; | | |
| 1961 | } | | |
| 1962 | | | |
| 1963 | // Nested tables indicate a layout table: | | |
| 1964 | if (table.getElementsByTagName("table")[0]) { | | |
| 1965 | table._readabilityDataTable = false; | | |
| 1966 | continue; | | |
| 1967 | } | | |
| 1968 | | | |
| 1969 | var sizeInfo = this._getRowAndColumnCount(table); | | |
| 1970 | if (sizeInfo.rows >= 10 || sizeInfo.columns > 4) { | | |
| 1971 | table._readabilityDataTable = true; | | |
| 1972 | continue; | | |
| 1973 | } | | |
| 1974 | // Now just go by size entirely: | | |
| 1975 | table._readabilityDataTable = sizeInfo.rows * sizeInfo.columns > 10; | | |
| 1976 | } | | |
| 1977 | }, | | |
| 1978 | | | |
| 1979 | /* convert images and figures that have properties like data-src into images that can be loaded without JS */ | | |
| 1980 | _fixLazyImages: function (root) { | | |
| 1981 | this._forEachNode(this._getAllNodesWithTag(root, ["img", "picture", "figure"]), function (elem) { | | |
| 1982 | // In some sites (e.g. Kotaku), they put 1px square image as base64 data uri in the src attribute. | | |
| 1983 | // So, here we check if the data uri is too short, just might as well remove it. | | |
| 1984 | if (elem.src && this.REGEXPS.b64DataUrl.test(elem.src)) { | | |
| 1985 | // Make sure it's not SVG, because SVG can have a meaningful image in under 133 bytes. | | |
| 1986 | var parts = this.REGEXPS.b64DataUrl.exec(elem.src); | | |
| 1987 | if (parts[1] === "image/svg+xml") { | | |
| 1988 | return; | | |
| 1989 | } | | |
| 1990 | | | |
| 1991 | // Make sure this element has other attributes which contains image. | | |
| 1992 | // If it doesn't, then this src is important and shouldn't be removed. | | |
| 1993 | var srcCouldBeRemoved = false; | | |
| 1994 | for (var i = 0; i < elem.attributes.length; i++) { | | |
| 1995 | var attr = elem.attributes[i]; | | |
| 1996 | if (attr.name === "src") { | | |
| 1997 | continue; | | |
| 1998 | } | | |
| 1999 | | | |
| 2000 | if (/\.(jpg|jpeg|png|webp)/i.test(attr.value)) { | | |
| 2001 | srcCouldBeRemoved = true; | | |
| 2002 | break; | | |
| 2003 | } | | |
| 2004 | } | | |
| 2005 | | | |
| 2006 | // Here we assume if image is less than 100 bytes (or 133B after encoded to base64) | | |
| 2007 | // it will be too small, therefore it might be placeholder image. | | |
| 2008 | if (srcCouldBeRemoved) { | | |
| 2009 | var b64starts = elem.src.search(/base64\s*/i) + 7; | | |
| 2010 | var b64length = elem.src.length - b64starts; | | |
| 2011 | if (b64length < 133) { | | |
| 2012 | elem.removeAttribute("src"); | | |
| 2013 | } | | |
| 2014 | } | | |
| 2015 | } | | |
| 2016 | | | |
| 2017 | // also check for "null" to work around https://github.com/jsdom/jsdom/issues/2580 | | |
| 2018 | if ((elem.src || (elem.srcset && elem.srcset != "null")) && elem.className.toLowerCase().indexOf("lazy") === -1) { | | |
| 2019 | return; | | |
| 2020 | } | | |
| 2021 | | | |
| 2022 | for (var j = 0; j < elem.attributes.length; j++) { | | |
| 2023 | attr = elem.attributes[j]; | | |
| 2024 | if (attr.name === "src" || attr.name === "srcset" || attr.name === "alt") { | | |
| 2025 | continue; | | |
| 2026 | } | | |
| 2027 | var copyTo = null; | | |
| 2028 | if (/\.(jpg|jpeg|png|webp)\s+\d/.test(attr.value)) { | | |
| 2029 | copyTo = "srcset"; | | |
| 2030 | } else if (/^\s*\S+\.(jpg|jpeg|png|webp)\S*\s*$/.test(attr.value)) { | | |
| 2031 | copyTo = "src"; | | |
| 2032 | } | | |
| 2033 | if (copyTo) { | | |
| 2034 | //if this is an img or picture, set the attribute directly | | |
| 2035 | if (elem.tagName === "IMG" || elem.tagName === "PICTURE") { | | |
| 2036 | elem.setAttribute(copyTo, attr.value); | | |
| 2037 | } else if (elem.tagName === "FIGURE" && !this._getAllNodesWithTag(elem, ["img", "picture"]).length) { | | |
| 2038 | //if the item is a <figure> that does not contain an image or picture, create one and place it inside the figure | | |
| 2039 | //see the nytimes-3 testcase for an example | | |
| 2040 | var img = this._doc.createElement("img"); | | |
| 2041 | img.setAttribute(copyTo, attr.value); | | |
| 2042 | elem.appendChild(img); | | |
| 2043 | } | | |
| 2044 | } | | |
| 2045 | } | | |
| 2046 | }); | | |
| 2047 | }, | | |
| 2048 | | | |
| 2049 | _getTextDensity: function(e, tags) { | | |
| 2050 | var textLength = this._getInnerText(e, true).length; | | |
| 2051 | if (textLength === 0) { | | |
| 2052 | return 0; | | |
| 2053 | } | | |
| 2054 | var childrenLength = 0; | | |
| 2055 | var children = this._getAllNodesWithTag(e, tags); | | |
| 2056 | this._forEachNode(children, (child) => childrenLength += this._getInnerText(child, true).length); | | |
| 2057 | return childrenLength / textLength; | | |
| 2058 | }, | | |
| 2059 | | | |
| 2060 | /** | | |
| 2061 | * Clean an element of all tags of type "tag" if they look fishy. | | |
| 2062 | * "Fishy" is an algorithm based on content length, classnames, link density, number of images & embeds, etc. | | |
| 2063 | * | | |
| 2064 | * @return void | | |
| 2065 | **/ | | |
| 2066 | _cleanConditionally: function(e, tag) { | | |
| 2067 | if (!this._flagIsActive(this.FLAG_CLEAN_CONDITIONALLY)) | | |
| 2068 | return; | | |
| 2069 | | | |
| 2070 | // Gather counts for other typical elements embedded within. | | |
| 2071 | // Traverse backwards so we can remove nodes at the same time | | |
| 2072 | // without effecting the traversal. | | |
| 2073 | // | | |
| 2074 | // TODO: Consider taking into account original contentScore here. | | |
| 2075 | this._removeNodes(this._getAllNodesWithTag(e, [tag]), function(node) { | | |
| 2076 | // First check if this node IS data table, in which case don't remove it. | | |
| 2077 | var isDataTable = function(t) { | | |
| 2078 | return t._readabilityDataTable; | | |
| 2079 | }; | | |
| 2080 | | | |
| 2081 | var isList = tag === "ul" || tag === "ol"; | | |
| 2082 | if (!isList) { | | |
| 2083 | var listLength = 0; | | |
| 2084 | var listNodes = this._getAllNodesWithTag(node, ["ul", "ol"]); | | |
| 2085 | this._forEachNode(listNodes, (list) => listLength += this._getInnerText(list).length); | | |
| 2086 | isList = listLength / this._getInnerText(node).length > 0.9; | | |
| 2087 | } | | |
| 2088 | | | |
| 2089 | if (tag === "table" && isDataTable(node)) { | | |
| 2090 | return false; | | |
| 2091 | } | | |
| 2092 | | | |
| 2093 | // Next check if we're inside a data table, in which case don't remove it as well. | | |
| 2094 | if (this._hasAncestorTag(node, "table", -1, isDataTable)) { | | |
| 2095 | return false; | | |
| 2096 | } | | |
| 2097 | | | |
| 2098 | if (this._hasAncestorTag(node, "code")) { | | |
| 2099 | return false; | | |
| 2100 | } | | |
| 2101 | | | |
| 2102 | var weight = this._getClassWeight(node); | | |
| 2103 | | | |
| 2104 | this.log("Cleaning Conditionally", node); | | |
| 2105 | | | |
| 2106 | var contentScore = 0; | | |
| 2107 | | | |
| 2108 | if (weight + contentScore < 0) { | | |
| 2109 | return true; | | |
| 2110 | } | | |
| 2111 | | | |
| 2112 | if (this._getCharCount(node, ",") < 10) { | | |
| 2113 | // If there are not very many commas, and the number of | | |
| 2114 | // non-paragraph elements is more than paragraphs or other | | |
| 2115 | // ominous signs, remove the element. | | |
| 2116 | var p = node.getElementsByTagName("p").length; | | |
| 2117 | var img = node.getElementsByTagName("img").length; | | |
| 2118 | var li = node.getElementsByTagName("li").length - 100; | | |
| 2119 | var input = node.getElementsByTagName("input").length; | | |
| 2120 | var headingDensity = this._getTextDensity(node, ["h1", "h2", "h3", "h4", "h5", "h6"]); | | |
| 2121 | | | |
| 2122 | var embedCount = 0; | | |
| 2123 | var embeds = this._getAllNodesWithTag(node, ["object", "embed", "iframe"]); | | |
| 2124 | | | |
| 2125 | for (var i = 0; i < embeds.length; i++) { | | |
| 2126 | // If this embed has attribute that matches video regex, don't delete it. | | |
| 2127 | for (var j = 0; j < embeds[i].attributes.length; j++) { | | |
| 2128 | if (this._allowedVideoRegex.test(embeds[i].attributes[j].value)) { | | |
| 2129 | return false; | | |
| 2130 | } | | |
| 2131 | } | | |
| 2132 | | | |
| 2133 | // For embed with <object> tag, check inner HTML as well. | | |
| 2134 | if (embeds[i].tagName === "object" && this._allowedVideoRegex.test(embeds[i].innerHTML)) { | | |
| 2135 | return false; | | |
| 2136 | } | | |
| 2137 | | | |
| 2138 | embedCount++; | | |
| 2139 | } | | |
| 2140 | | | |
| 2141 | var linkDensity = this._getLinkDensity(node); | | |
| 2142 | var contentLength = this._getInnerText(node).length; | | |
| 2143 | | | |
| 2144 | var haveToRemove = | | |
| 2145 | (img > 1 && p / img < 0.5 && !this._hasAncestorTag(node, "figure")) || | | |
| 2146 | (!isList && li > p) || | | |
| 2147 | (input > Math.floor(p/3)) || | | |
| 2148 | (!isList && headingDensity < 0.9 && contentLength < 25 && (img === 0 || img > 2) && !this._hasAncestorTag(node, "figure")) || | | |
| 2149 | (!isList && weight < 25 && linkDensity > 0.2) || | | |
| 2150 | (weight >= 25 && linkDensity > 0.5) || | | |
| 2151 | ((embedCount === 1 && contentLength < 75) || embedCount > 1); | | |
| 2152 | // Allow simple lists of images to remain in pages | | |
| 2153 | if (isList && haveToRemove) { | | |
| 2154 | for (var x = 0; x < node.children.length; x++) { | | |
| 2155 | let child = node.children[x]; | | |
| 2156 | // Don't filter in lists with li's that contain more than one child | | |
| 2157 | if (child.children.length > 1) { | | |
| 2158 | return haveToRemove; | | |
| 2159 | } | | |
| 2160 | } | | |
| 2161 | let li_count = node.getElementsByTagName("li").length; | | |
| 2162 | // Only allow the list to remain if every li contains an image | | |
| 2163 | if (img == li_count) { | | |
| 2164 | return false; | | |
| 2165 | } | | |
| 2166 | } | | |
| 2167 | return haveToRemove; | | |
| 2168 | } | | |
| 2169 | return false; | | |
| 2170 | }); | | |
| 2171 | }, | | |
| 2172 | | | |
| 2173 | /** | | |
| 2174 | * Clean out elements that match the specified conditions | | |
| 2175 | * | | |
| 2176 | * @param Element | | |
| 2177 | * @param Function determines whether a node should be removed | | |
| 2178 | * @return void | | |
| 2179 | **/ | | |
| 2180 | _cleanMatchedNodes: function(e, filter) { | | |
| 2181 | var endOfSearchMarkerNode = this._getNextNode(e, true); | | |
| 2182 | var next = this._getNextNode(e); | | |
| 2183 | while (next && next != endOfSearchMarkerNode) { | | |
| 2184 | if (filter.call(this, next, next.className + " " + next.id)) { | | |
| 2185 | next = this._removeAndGetNext(next); | | |
| 2186 | } else { | | |
| 2187 | next = this._getNextNode(next); | | |
| 2188 | } | | |
| 2189 | } | | |
| 2190 | }, | | |
| 2191 | | | |
| 2192 | /** | | |
| 2193 | * Clean out spurious headers from an Element. | | |
| 2194 | * | | |
| 2195 | * @param Element | | |
| 2196 | * @return void | | |
| 2197 | **/ | | |
| 2198 | _cleanHeaders: function(e) { | | |
| 2199 | let headingNodes = this._getAllNodesWithTag(e, ["h1", "h2"]); | | |
| 2200 | this._removeNodes(headingNodes, function(node) { | | |
| 2201 | let shouldRemove = this._getClassWeight(node) < 0; | | |
| 2202 | if (shouldRemove) { | | |
| 2203 | this.log("Removing header with low class weight:", node); | | |
| 2204 | } | | |
| 2205 | return shouldRemove; | | |
| 2206 | }); | | |
| 2207 | }, | | |
| 2208 | | | |
| 2209 | /** | | |
| 2210 | * Check if this node is an H1 or H2 element whose content is mostly | | |
| 2211 | * the same as the article title. | | |
| 2212 | * | | |
| 2213 | * @param Element the node to check. | | |
| 2214 | * @return boolean indicating whether this is a title-like header. | | |
| 2215 | */ | | |
| 2216 | _headerDuplicatesTitle: function(node) { | | |
| 2217 | if (node.tagName != "H1" && node.tagName != "H2") { | | |
| 2218 | return false; | | |
| 2219 | } | | |
| 2220 | var heading = this._getInnerText(node, false); | | |
| 2221 | this.log("Evaluating similarity of header:", heading, this._articleTitle); | | |
| 2222 | return this._textSimilarity(this._articleTitle, heading) > 0.75; | | |
| 2223 | }, | | |
| 2224 | | | |
| 2225 | _flagIsActive: function(flag) { | | |
| 2226 | return (this._flags & flag) > 0; | | |
| 2227 | }, | | |
| 2228 | | | |
| 2229 | _removeFlag: function(flag) { | | |
| 2230 | this._flags = this._flags & ~flag; | | |
| 2231 | }, | | |
| 2232 | | | |
| 2233 | _isProbablyVisible: function(node) { | | |
| 2234 | // Have to null-check node.style and node.className.indexOf to deal with SVG and MathML nodes. | | |
| 2235 | return (!node.style || node.style.display != "none") | | |
| 2236 | && (!node.style || node.style.visibility != "hidden") | | |
| 2237 | && !node.hasAttribute("hidden") | | |
| 2238 | //check for "fallback-image" so that wikimedia math images are displayed | | |
| 2239 | && (!node.hasAttribute("aria-hidden") || node.getAttribute("aria-hidden") != "true" || (node.className && node.className.indexOf && node.className.indexOf("fallback-image") !== -1)); | | |
| 2240 | }, | | |
| 2241 | | | |
| 2242 | /** | | |
| 2243 | * Runs readability. | | |
| 2244 | * | | |
| 2245 | * Workflow: | | |
| 2246 | * 1. Prep the document by removing script tags, css, etc. | | |
| 2247 | * 2. Build readability's DOM tree. | | |
| 2248 | * 3. Grab the article content from the current dom tree. | | |
| 2249 | * 4. Replace the current DOM tree with the new one. | | |
| 2250 | * 5. Read peacefully. | | |
| 2251 | * | | |
| 2252 | * @return void | | |
| 2253 | **/ | | |
| 2254 | parse: function () { | | |
| 2255 | // Avoid parsing too large documents, as per configuration option | | |
| 2256 | if (this._maxElemsToParse > 0) { | | |
| 2257 | var numTags = this._doc.getElementsByTagName("*").length; | | |
| 2258 | if (numTags > this._maxElemsToParse) { | | |
| 2259 | throw new Error("Aborting parsing document; " + numTags + " elements found"); | | |
| 2260 | } | | |
| 2261 | } | | |
| 2262 | | | |
| 2263 | // Unwrap image from noscript | | |
| 2264 | this._unwrapNoscriptImages(this._doc); | | |
| 2265 | | | |
| 2266 | // Extract JSON-LD metadata before removing scripts | | |
| 2267 | var jsonLd = this._disableJSONLD ? {} : this._getJSONLD(this._doc); | | |
| 2268 | | | |
| 2269 | // Remove script tags from the document. | | |
| 2270 | this._removeScripts(this._doc); | | |
| 2271 | | | |
| 2272 | this._prepDocument(); | | |
| 2273 | | | |
| 2274 | var metadata = this._getArticleMetadata(jsonLd); | | |
| 2275 | this._articleTitle = metadata.title; | | |
| 2276 | | | |
| 2277 | var articleContent = this._grabArticle(); | | |
| 2278 | if (!articleContent) | | |
| 2279 | return null; | | |
| 2280 | | | |
| 2281 | this.log("Grabbed: " + articleContent.innerHTML); | | |
| 2282 | | | |
| 2283 | this._postProcessContent(articleContent); | | |
| 2284 | | | |
| 2285 | // If we haven't found an excerpt in the article's metadata, use the article's | | |
| 2286 | // first paragraph as the excerpt. This is used for displaying a preview of | | |
| 2287 | // the article's content. | | |
| 2288 | if (!metadata.excerpt) { | | |
| 2289 | var paragraphs = articleContent.getElementsByTagName("p"); | | |
| 2290 | if (paragraphs.length > 0) { | | |
| 2291 | metadata.excerpt = paragraphs[0].textContent.trim(); | | |
| 2292 | } | | |
| 2293 | } | | |
| 2294 | | | |
| 2295 | var textContent = articleContent.textContent; | | |
| 2296 | return { | | |
| 2297 | title: this._articleTitle, | | |
| 2298 | byline: metadata.byline || this._articleByline, | | |
| 2299 | dir: this._articleDir, | | |
| 2300 | lang: this._articleLang, | | |
| 2301 | content: this._serializer(articleContent), | | |
| 2302 | textContent: textContent, | | |
| 2303 | length: textContent.length, | | |
| 2304 | excerpt: metadata.excerpt, | | |
| 2305 | siteName: metadata.siteName || this._articleSiteName, | | |
| 2306 | publishedTime: metadata.publishedTime | | |
| 2307 | }; | | |
| 2308 | } | | |
| 2309 | }; | | |
| 2310 | | | |
| 2311 | if (typeof module === "object") { | | |
| 2312 | /* global module */ | | |
| 2313 | module.exports = Readability; | | |
| 2314 | } | | |