The Tor Browser / file comparison
comparison: mobile/android/chrome/content/Readability.js
mobile/android/chrome/content/Readability.js
- changeset 0
- 6474c204b198
equal
deleted
inserted
replaced
| -1:000000000000 | 0:dbc61788cda5 |
|---|---|
| 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 var Readability = function(uri, doc) { | |
| 23 const ENABLE_LOGGING = false; | |
| 24 | |
| 25 this._uri = uri; | |
| 26 this._doc = doc; | |
| 27 this._biggestFrame = false; | |
| 28 this._articleByline = null; | |
| 29 this._articleDir = null; | |
| 30 | |
| 31 // Start with all flags set | |
| 32 this._flags = this.FLAG_STRIP_UNLIKELYS | | |
| 33 this.FLAG_WEIGHT_CLASSES | | |
| 34 this.FLAG_CLEAN_CONDITIONALLY; | |
| 35 | |
| 36 // The list of pages we've parsed in this call of readability, | |
| 37 // for autopaging. As a key store for easier searching. | |
| 38 this._parsedPages = {}; | |
| 39 | |
| 40 // A list of the ETag headers of pages we've parsed, in case they happen to match, | |
| 41 // we'll know it's a duplicate. | |
| 42 this._pageETags = {}; | |
| 43 | |
| 44 // Make an AJAX request for each page and append it to the document. | |
| 45 this._curPageNum = 1; | |
| 46 | |
| 47 // Control whether log messages are sent to the console | |
| 48 if (ENABLE_LOGGING) { | |
| 49 this.log = function (msg) { | |
| 50 dump("Reader: (Readability) " + msg); | |
| 51 }; | |
| 52 } else { | |
| 53 this.log = function () {}; | |
| 54 } | |
| 55 } | |
| 56 | |
| 57 Readability.prototype = { | |
| 58 FLAG_STRIP_UNLIKELYS: 0x1, | |
| 59 FLAG_WEIGHT_CLASSES: 0x2, | |
| 60 FLAG_CLEAN_CONDITIONALLY: 0x4, | |
| 61 | |
| 62 // The number of top candidates to consider when analysing how | |
| 63 // tight the competition is among candidates. | |
| 64 N_TOP_CANDIDATES: 5, | |
| 65 | |
| 66 // The maximum number of pages to loop through before we call | |
| 67 // it quits and just show a link. | |
| 68 MAX_PAGES: 5, | |
| 69 | |
| 70 // All of the regular expressions in use within readability. | |
| 71 // Defined up here so we don't instantiate them repeatedly in loops. | |
| 72 REGEXPS: { | |
| 73 unlikelyCandidates: /combx|comment|community|disqus|extra|foot|header|menu|remark|rss|shoutbox|sidebar|sponsor|ad-break|agegate|pagination|pager|popup|tweet|twitter/i, | |
| 74 okMaybeItsACandidate: /and|article|body|column|main|shadow/i, | |
| 75 positive: /article|body|content|entry|hentry|main|page|pagination|post|text|blog|story/i, | |
| 76 negative: /hidden|combx|comment|com-|contact|foot|footer|footnote|masthead|media|meta|outbrain|promo|related|scroll|shoutbox|sidebar|sponsor|shopping|tags|tool|widget/i, | |
| 77 extraneous: /print|archive|comment|discuss|e[\-]?mail|share|reply|all|login|sign|single|utility/i, | |
| 78 byline: /byline|author|dateline|writtenby/i, | |
| 79 replaceFonts: /<(\/?)font[^>]*>/gi, | |
| 80 trim: /^\s+|\s+$/g, | |
| 81 normalize: /\s{2,}/g, | |
| 82 videos: /http:\/\/(www\.)?(youtube|vimeo)\.com/i, | |
| 83 nextLink: /(next|weiter|continue|>([^\|]|$)|»([^\|]|$))/i, | |
| 84 prevLink: /(prev|earl|old|new|<|«)/i, | |
| 85 whitespace: /^\s*$/ | |
| 86 }, | |
| 87 | |
| 88 DIV_TO_P_ELEMS: [ "A", "BLOCKQUOTE", "DL", "DIV", "IMG", "OL", "P", "PRE", "TABLE", "UL", "SELECT" ], | |
| 89 | |
| 90 /** | |
| 91 * Run any post-process modifications to article content as necessary. | |
| 92 * | |
| 93 * @param Element | |
| 94 * @return void | |
| 95 **/ | |
| 96 _postProcessContent: function(articleContent) { | |
| 97 // Readability cannot open relative uris so we convert them to absolute uris. | |
| 98 this._fixRelativeUris(articleContent); | |
| 99 }, | |
| 100 | |
| 101 /** | |
| 102 * Converts each <a> and <img> uri in the given element to an absolute URI. | |
| 103 * | |
| 104 * @param Element | |
| 105 * @return void | |
| 106 */ | |
| 107 _fixRelativeUris: function(articleContent) { | |
| 108 let scheme = this._uri.scheme; | |
| 109 let prePath = this._uri.prePath; | |
| 110 let pathBase = this._uri.pathBase; | |
| 111 | |
| 112 function toAbsoluteURI(uri) { | |
| 113 // If this is already an absolute URI, return it. | |
| 114 if (/^[a-zA-Z][a-zA-Z0-9\+\-\.]*:/.test(uri)) | |
| 115 return uri; | |
| 116 | |
| 117 // Scheme-rooted relative URI. | |
| 118 if (uri.substr(0, 2) == "//") | |
| 119 return scheme + "://" + uri.substr(2); | |
| 120 | |
| 121 // Prepath-rooted relative URI. | |
| 122 if (uri[0] == "/") | |
| 123 return prePath + uri; | |
| 124 | |
| 125 // Standard relative URI; add entire path. pathBase already includes a | |
| 126 // trailing "/". | |
| 127 return pathBase + uri; | |
| 128 } | |
| 129 | |
| 130 function convertRelativeURIs(tagName, propName) { | |
| 131 let elems = articleContent.getElementsByTagName(tagName); | |
| 132 for (let i = elems.length; --i >= 0;) { | |
| 133 let elem = elems[i]; | |
| 134 let relativeURI = elem.getAttribute(propName); | |
| 135 if (relativeURI != null) | |
| 136 elems[i].setAttribute(propName, toAbsoluteURI(relativeURI)); | |
| 137 } | |
| 138 } | |
| 139 | |
| 140 // Fix links. | |
| 141 convertRelativeURIs("a", "href"); | |
| 142 | |
| 143 // Fix images. | |
| 144 convertRelativeURIs("img", "src"); | |
| 145 }, | |
| 146 | |
| 147 /** | |
| 148 * Get the article title as an H1. | |
| 149 * | |
| 150 * @return void | |
| 151 **/ | |
| 152 _getArticleTitle: function() { | |
| 153 let doc = this._doc; | |
| 154 let curTitle = ""; | |
| 155 let origTitle = ""; | |
| 156 | |
| 157 try { | |
| 158 curTitle = origTitle = doc.title; | |
| 159 | |
| 160 // If they had an element with id "title" in their HTML | |
| 161 if (typeof curTitle !== "string") | |
| 162 curTitle = origTitle = this._getInnerText(doc.getElementsByTagName('title')[0]); | |
| 163 } catch(e) {} | |
| 164 | |
| 165 if (curTitle.match(/ [\|\-] /)) { | |
| 166 curTitle = origTitle.replace(/(.*)[\|\-] .*/gi,'$1'); | |
| 167 | |
| 168 if (curTitle.split(' ').length < 3) | |
| 169 curTitle = origTitle.replace(/[^\|\-]*[\|\-](.*)/gi,'$1'); | |
| 170 } else if (curTitle.indexOf(': ') !== -1) { | |
| 171 curTitle = origTitle.replace(/.*:(.*)/gi, '$1'); | |
| 172 | |
| 173 if (curTitle.split(' ').length < 3) | |
| 174 curTitle = origTitle.replace(/[^:]*[:](.*)/gi,'$1'); | |
| 175 } else if (curTitle.length > 150 || curTitle.length < 15) { | |
| 176 let hOnes = doc.getElementsByTagName('h1'); | |
| 177 | |
| 178 if (hOnes.length === 1) | |
| 179 curTitle = this._getInnerText(hOnes[0]); | |
| 180 } | |
| 181 | |
| 182 curTitle = curTitle.replace(this.REGEXPS.trim, ""); | |
| 183 | |
| 184 if (curTitle.split(' ').length <= 4) | |
| 185 curTitle = origTitle; | |
| 186 | |
| 187 return curTitle; | |
| 188 }, | |
| 189 | |
| 190 /** | |
| 191 * Prepare the HTML document for readability to scrape it. | |
| 192 * This includes things like stripping javascript, CSS, and handling terrible markup. | |
| 193 * | |
| 194 * @return void | |
| 195 **/ | |
| 196 _prepDocument: function() { | |
| 197 let doc = this._doc; | |
| 198 | |
| 199 // In some cases a body element can't be found (if the HTML is | |
| 200 // totally hosed for example) so we create a new body node and | |
| 201 // append it to the document. | |
| 202 if (doc.body === null) { | |
| 203 let body = doc.createElement("body"); | |
| 204 | |
| 205 try { | |
| 206 doc.body = body; | |
| 207 } catch(e) { | |
| 208 doc.documentElement.appendChild(body); | |
| 209 this.log(e); | |
| 210 } | |
| 211 } | |
| 212 | |
| 213 // Remove all style tags in head | |
| 214 let styleTags = doc.getElementsByTagName("style"); | |
| 215 for (let st = 0; st < styleTags.length; st += 1) { | |
| 216 styleTags[st].textContent = ""; | |
| 217 } | |
| 218 | |
| 219 this._replaceBrs(doc.body); | |
| 220 | |
| 221 let fonts = doc.getElementsByTagName("FONT"); | |
| 222 for (let i = fonts.length; --i >=0;) { | |
| 223 this._setNodeTag(fonts[i], "SPAN"); | |
| 224 } | |
| 225 }, | |
| 226 | |
| 227 /** | |
| 228 * Finds the next element, starting from the given node, and ignoring | |
| 229 * whitespace in between. If the given node is an element, the same node is | |
| 230 * returned. | |
| 231 */ | |
| 232 _nextElement: function (node) { | |
| 233 let next = node; | |
| 234 while (next | |
| 235 && (next.nodeType != Node.ELEMENT_NODE) | |
| 236 && this.REGEXPS.whitespace.test(next.textContent)) { | |
| 237 next = next.nextSibling; | |
| 238 } | |
| 239 return next; | |
| 240 }, | |
| 241 | |
| 242 /** | |
| 243 * Replaces 2 or more successive <br> elements with a single <p>. | |
| 244 * Whitespace between <br> elements are ignored. For example: | |
| 245 * <div>foo<br>bar<br> <br><br>abc</div> | |
| 246 * will become: | |
| 247 * <div>foo<br>bar<p>abc</p></div> | |
| 248 */ | |
| 249 _replaceBrs: function (elem) { | |
| 250 let brs = elem.getElementsByTagName("br"); | |
| 251 for (let i = 0; i < brs.length; i++) { | |
| 252 let br = brs[i]; | |
| 253 let next = br.nextSibling; | |
| 254 | |
| 255 // Whether 2 or more <br> elements have been found and replaced with a | |
| 256 // <p> block. | |
| 257 let replaced = false; | |
| 258 | |
| 259 // If we find a <br> chain, remove the <br>s until we hit another element | |
| 260 // or non-whitespace. This leaves behind the first <br> in the chain | |
| 261 // (which will be replaced with a <p> later). | |
| 262 while ((next = this._nextElement(next)) && (next.tagName == "BR")) { | |
| 263 replaced = true; | |
| 264 let sibling = next.nextSibling; | |
| 265 next.parentNode.removeChild(next); | |
| 266 next = sibling; | |
| 267 } | |
| 268 | |
| 269 // If we removed a <br> chain, replace the remaining <br> with a <p>. Add | |
| 270 // all sibling nodes as children of the <p> until we hit another <br> | |
| 271 // chain. | |
| 272 if (replaced) { | |
| 273 let p = this._doc.createElement("p"); | |
| 274 br.parentNode.replaceChild(p, br); | |
| 275 | |
| 276 next = p.nextSibling; | |
| 277 while (next) { | |
| 278 // If we've hit another <br><br>, we're done adding children to this <p>. | |
| 279 if (next.tagName == "BR") { | |
| 280 let nextElem = this._nextElement(next); | |
| 281 if (nextElem && nextElem.tagName == "BR") | |
| 282 break; | |
| 283 } | |
| 284 | |
| 285 // Otherwise, make this node a child of the new <p>. | |
| 286 let sibling = next.nextSibling; | |
| 287 p.appendChild(next); | |
| 288 next = sibling; | |
| 289 } | |
| 290 } | |
| 291 } | |
| 292 }, | |
| 293 | |
| 294 _setNodeTag: function (node, tag) { | |
| 295 node.localName = tag.toLowerCase(); | |
| 296 node.tagName = tag.toUpperCase(); | |
| 297 }, | |
| 298 | |
| 299 /** | |
| 300 * Prepare the article node for display. Clean out any inline styles, | |
| 301 * iframes, forms, strip extraneous <p> tags, etc. | |
| 302 * | |
| 303 * @param Element | |
| 304 * @return void | |
| 305 **/ | |
| 306 _prepArticle: function(articleContent) { | |
| 307 this._cleanStyles(articleContent); | |
| 308 | |
| 309 // Clean out junk from the article content | |
| 310 this._cleanConditionally(articleContent, "form"); | |
| 311 this._clean(articleContent, "object"); | |
| 312 this._clean(articleContent, "h1"); | |
| 313 | |
| 314 // If there is only one h2, they are probably using it as a header | |
| 315 // and not a subheader, so remove it since we already have a header. | |
| 316 if (articleContent.getElementsByTagName('h2').length === 1) | |
| 317 this._clean(articleContent, "h2"); | |
| 318 | |
| 319 this._clean(articleContent, "iframe"); | |
| 320 this._cleanHeaders(articleContent); | |
| 321 | |
| 322 // Do these last as the previous stuff may have removed junk | |
| 323 // that will affect these | |
| 324 this._cleanConditionally(articleContent, "table"); | |
| 325 this._cleanConditionally(articleContent, "ul"); | |
| 326 this._cleanConditionally(articleContent, "div"); | |
| 327 | |
| 328 // Remove extra paragraphs | |
| 329 let articleParagraphs = articleContent.getElementsByTagName('p'); | |
| 330 for (let i = articleParagraphs.length - 1; i >= 0; i -= 1) { | |
| 331 let imgCount = articleParagraphs[i].getElementsByTagName('img').length; | |
| 332 let embedCount = articleParagraphs[i].getElementsByTagName('embed').length; | |
| 333 let objectCount = articleParagraphs[i].getElementsByTagName('object').length; | |
| 334 | |
| 335 if (imgCount === 0 && | |
| 336 embedCount === 0 && | |
| 337 objectCount === 0 && | |
| 338 this._getInnerText(articleParagraphs[i], false) === '') | |
| 339 articleParagraphs[i].parentNode.removeChild(articleParagraphs[i]); | |
| 340 } | |
| 341 | |
| 342 let brs = articleContent.getElementsByTagName("BR"); | |
| 343 for (let i = brs.length; --i >= 0;) { | |
| 344 let br = brs[i]; | |
| 345 let next = this._nextElement(br.nextSibling); | |
| 346 if (next && next.tagName == "P") | |
| 347 br.parentNode.removeChild(br); | |
| 348 } | |
| 349 }, | |
| 350 | |
| 351 /** | |
| 352 * Initialize a node with the readability object. Also checks the | |
| 353 * className/id for special names to add to its score. | |
| 354 * | |
| 355 * @param Element | |
| 356 * @return void | |
| 357 **/ | |
| 358 _initializeNode: function(node) { | |
| 359 node.readability = {"contentScore": 0}; | |
| 360 | |
| 361 switch(node.tagName) { | |
| 362 case 'DIV': | |
| 363 node.readability.contentScore += 5; | |
| 364 break; | |
| 365 | |
| 366 case 'PRE': | |
| 367 case 'TD': | |
| 368 case 'BLOCKQUOTE': | |
| 369 node.readability.contentScore += 3; | |
| 370 break; | |
| 371 | |
| 372 case 'ADDRESS': | |
| 373 case 'OL': | |
| 374 case 'UL': | |
| 375 case 'DL': | |
| 376 case 'DD': | |
| 377 case 'DT': | |
| 378 case 'LI': | |
| 379 case 'FORM': | |
| 380 node.readability.contentScore -= 3; | |
| 381 break; | |
| 382 | |
| 383 case 'H1': | |
| 384 case 'H2': | |
| 385 case 'H3': | |
| 386 case 'H4': | |
| 387 case 'H5': | |
| 388 case 'H6': | |
| 389 case 'TH': | |
| 390 node.readability.contentScore -= 5; | |
| 391 break; | |
| 392 } | |
| 393 | |
| 394 node.readability.contentScore += this._getClassWeight(node); | |
| 395 }, | |
| 396 | |
| 397 /*** | |
| 398 * grabArticle - Using a variety of metrics (content score, classname, element types), find the content that is | |
| 399 * most likely to be the stuff a user wants to read. Then return it wrapped up in a div. | |
| 400 * | |
| 401 * @param page a document to run upon. Needs to be a full document, complete with body. | |
| 402 * @return Element | |
| 403 **/ | |
| 404 _grabArticle: function (page) { | |
| 405 let doc = this._doc; | |
| 406 let isPaging = (page !== null ? true: false); | |
| 407 page = page ? page : this._doc.body; | |
| 408 let pageCacheHtml = page.innerHTML; | |
| 409 | |
| 410 // Check if any "dir" is set on the toplevel document element | |
| 411 this._articleDir = doc.documentElement.getAttribute("dir"); | |
| 412 | |
| 413 while (true) { | |
| 414 let stripUnlikelyCandidates = this._flagIsActive(this.FLAG_STRIP_UNLIKELYS); | |
| 415 let allElements = page.getElementsByTagName('*'); | |
| 416 | |
| 417 // First, node prepping. Trash nodes that look cruddy (like ones with the | |
| 418 // class name "comment", etc), and turn divs into P tags where they have been | |
| 419 // used inappropriately (as in, where they contain no other block level elements.) | |
| 420 // | |
| 421 // Note: Assignment from index for performance. See http://www.peachpit.com/articles/article.aspx?p=31567&seqNum=5 | |
| 422 // TODO: Shouldn't this be a reverse traversal? | |
| 423 let node = null; | |
| 424 let nodesToScore = []; | |
| 425 | |
| 426 // Let each node know its index in the allElements array. | |
| 427 for (let i = allElements.length; --i >= 0;) { | |
| 428 allElements[i]._index = i; | |
| 429 } | |
| 430 | |
| 431 /** | |
| 432 * JSDOMParser returns static node lists, not live ones. When we remove | |
| 433 * an element from the document, we need to manually remove it - and all | |
| 434 * of its children - from the allElements array. | |
| 435 */ | |
| 436 function purgeNode(node) { | |
| 437 for (let i = node.childNodes.length; --i >= 0;) { | |
| 438 purgeNode(node.childNodes[i]); | |
| 439 } | |
| 440 if (node._index !== undefined && allElements[node._index] == node) | |
| 441 delete allElements[node._index]; | |
| 442 } | |
| 443 | |
| 444 for (let nodeIndex = 0; nodeIndex < allElements.length; nodeIndex++) { | |
| 445 if (!(node = allElements[nodeIndex])) | |
| 446 continue; | |
| 447 | |
| 448 let matchString = node.className + node.id; | |
| 449 if (matchString.search(this.REGEXPS.byline) !== -1 && !this._articleByline) { | |
| 450 this._articleByline = node.textContent; | |
| 451 node.parentNode.removeChild(node); | |
| 452 purgeNode(node); | |
| 453 continue; | |
| 454 } | |
| 455 | |
| 456 // Remove unlikely candidates | |
| 457 if (stripUnlikelyCandidates) { | |
| 458 if (matchString.search(this.REGEXPS.unlikelyCandidates) !== -1 && | |
| 459 matchString.search(this.REGEXPS.okMaybeItsACandidate) === -1 && | |
| 460 node.tagName !== "BODY") { | |
| 461 this.log("Removing unlikely candidate - " + matchString); | |
| 462 node.parentNode.removeChild(node); | |
| 463 purgeNode(node); | |
| 464 continue; | |
| 465 } | |
| 466 } | |
| 467 | |
| 468 if (node.tagName === "P" || node.tagName === "TD" || node.tagName === "PRE") | |
| 469 nodesToScore[nodesToScore.length] = node; | |
| 470 | |
| 471 // Turn all divs that don't have children block level elements into p's | |
| 472 if (node.tagName === "DIV") { | |
| 473 // Sites like http://mobile.slate.com encloses each paragraph with a DIV | |
| 474 // element. DIVs with only a P element inside and no text content can be | |
| 475 // safely converted into plain P elements to avoid confusing the scoring | |
| 476 // algorithm with DIVs with are, in practice, paragraphs. | |
| 477 let pIndex = this._getSinglePIndexInsideDiv(node); | |
| 478 | |
| 479 if (pIndex >= 0 || !this._hasChildBlockElement(node)) { | |
| 480 if (pIndex >= 0) { | |
| 481 let newNode = node.childNodes[pIndex]; | |
| 482 node.parentNode.replaceChild(newNode, node); | |
| 483 purgeNode(node); | |
| 484 } else { | |
| 485 this._setNodeTag(node, "P"); | |
| 486 nodesToScore[nodesToScore.length] = node; | |
| 487 } | |
| 488 } else { | |
| 489 // EXPERIMENTAL | |
| 490 for (let i = 0, il = node.childNodes.length; i < il; i += 1) { | |
| 491 let childNode = node.childNodes[i]; | |
| 492 if (!childNode) | |
| 493 continue; | |
| 494 | |
| 495 if (childNode.nodeType === 3) { // Node.TEXT_NODE | |
| 496 let p = doc.createElement('p'); | |
| 497 p.textContent = childNode.textContent; | |
| 498 p.style.display = 'inline'; | |
| 499 p.className = 'readability-styled'; | |
| 500 childNode.parentNode.replaceChild(p, childNode); | |
| 501 } | |
| 502 } | |
| 503 } | |
| 504 } | |
| 505 } | |
| 506 | |
| 507 /** | |
| 508 * Loop through all paragraphs, and assign a score to them based on how content-y they look. | |
| 509 * Then add their score to their parent node. | |
| 510 * | |
| 511 * A score is determined by things like number of commas, class names, etc. Maybe eventually link density. | |
| 512 **/ | |
| 513 let candidates = []; | |
| 514 for (let pt = 0; pt < nodesToScore.length; pt += 1) { | |
| 515 let parentNode = nodesToScore[pt].parentNode; | |
| 516 let grandParentNode = parentNode ? parentNode.parentNode : null; | |
| 517 let innerText = this._getInnerText(nodesToScore[pt]); | |
| 518 | |
| 519 if (!parentNode || typeof(parentNode.tagName) === 'undefined') | |
| 520 continue; | |
| 521 | |
| 522 // If this paragraph is less than 25 characters, don't even count it. | |
| 523 if (innerText.length < 25) | |
| 524 continue; | |
| 525 | |
| 526 // Initialize readability data for the parent. | |
| 527 if (typeof parentNode.readability === 'undefined') { | |
| 528 this._initializeNode(parentNode); | |
| 529 candidates.push(parentNode); | |
| 530 } | |
| 531 | |
| 532 // Initialize readability data for the grandparent. | |
| 533 if (grandParentNode && | |
| 534 typeof(grandParentNode.readability) === 'undefined' && | |
| 535 typeof(grandParentNode.tagName) !== 'undefined') { | |
| 536 this._initializeNode(grandParentNode); | |
| 537 candidates.push(grandParentNode); | |
| 538 } | |
| 539 | |
| 540 let contentScore = 0; | |
| 541 | |
| 542 // Add a point for the paragraph itself as a base. | |
| 543 contentScore += 1; | |
| 544 | |
| 545 // Add points for any commas within this paragraph. | |
| 546 contentScore += innerText.split(',').length; | |
| 547 | |
| 548 // For every 100 characters in this paragraph, add another point. Up to 3 points. | |
| 549 contentScore += Math.min(Math.floor(innerText.length / 100), 3); | |
| 550 | |
| 551 // Add the score to the parent. The grandparent gets half. | |
| 552 parentNode.readability.contentScore += contentScore; | |
| 553 | |
| 554 if (grandParentNode) | |
| 555 grandParentNode.readability.contentScore += contentScore / 2; | |
| 556 } | |
| 557 | |
| 558 // After we've calculated scores, loop through all of the possible | |
| 559 // candidate nodes we found and find the one with the highest score. | |
| 560 let topCandidates = []; | |
| 561 for (let c = 0, cl = candidates.length; c < cl; c += 1) { | |
| 562 let candidate = candidates[c]; | |
| 563 | |
| 564 // Scale the final candidates score based on link density. Good content | |
| 565 // should have a relatively small link density (5% or less) and be mostly | |
| 566 // unaffected by this operation. | |
| 567 let candidateScore = candidate.readability.contentScore * (1 - this._getLinkDensity(candidate)); | |
| 568 candidate.readability.contentScore = candidateScore; | |
| 569 | |
| 570 this.log('Candidate: ' + candidate + " (" + candidate.className + ":" + | |
| 571 candidate.id + ") with score " + candidateScore); | |
| 572 | |
| 573 for (let t = 0; t < this.N_TOP_CANDIDATES; t++) { | |
| 574 let aTopCandidate = topCandidates[t]; | |
| 575 | |
| 576 if (!aTopCandidate || candidateScore > aTopCandidate.readability.contentScore) { | |
| 577 topCandidates.splice(t, 0, candidate); | |
| 578 if (topCandidates.length > this.N_TOP_CANDIDATES) | |
| 579 topCandidates.pop(); | |
| 580 break; | |
| 581 } | |
| 582 } | |
| 583 } | |
| 584 | |
| 585 let topCandidate = topCandidates[0] || null; | |
| 586 let lastTopCandidate = (topCandidates.length > 3 ? topCandidates[topCandidates.length - 1] : null); | |
| 587 | |
| 588 // If we still have no top candidate, just use the body as a last resort. | |
| 589 // We also have to copy the body node so it is something we can modify. | |
| 590 if (topCandidate === null || topCandidate.tagName === "BODY") { | |
| 591 // Move all of the page's children into topCandidate | |
| 592 topCandidate = doc.createElement("DIV"); | |
| 593 let children = page.childNodes; | |
| 594 for (let i = 0; i < children.length; ++i) { | |
| 595 topCandidate.appendChild(children[i]); | |
| 596 } | |
| 597 | |
| 598 page.appendChild(topCandidate); | |
| 599 | |
| 600 this._initializeNode(topCandidate); | |
| 601 } | |
| 602 | |
| 603 // Now that we have the top candidate, look through its siblings for content | |
| 604 // that might also be related. Things like preambles, content split by ads | |
| 605 // that we removed, etc. | |
| 606 let articleContent = doc.createElement("DIV"); | |
| 607 if (isPaging) | |
| 608 articleContent.id = "readability-content"; | |
| 609 | |
| 610 let siblingScoreThreshold = Math.max(10, topCandidate.readability.contentScore * 0.2); | |
| 611 let siblingNodes = topCandidate.parentNode.childNodes; | |
| 612 | |
| 613 for (let s = 0, sl = siblingNodes.length; s < sl; s += 1) { | |
| 614 let siblingNode = siblingNodes[s]; | |
| 615 let append = false; | |
| 616 | |
| 617 this.log("Looking at sibling node: " + siblingNode + " (" + siblingNode.className + ":" + siblingNode.id + ")" + ((typeof siblingNode.readability !== 'undefined') ? (" with score " + siblingNode.readability.contentScore) : '')); | |
| 618 this.log("Sibling has score " + (siblingNode.readability ? siblingNode.readability.contentScore : 'Unknown')); | |
| 619 | |
| 620 if (siblingNode === topCandidate) | |
| 621 append = true; | |
| 622 | |
| 623 let contentBonus = 0; | |
| 624 | |
| 625 // Give a bonus if sibling nodes and top candidates have the example same classname | |
| 626 if (siblingNode.className === topCandidate.className && topCandidate.className !== "") | |
| 627 contentBonus += topCandidate.readability.contentScore * 0.2; | |
| 628 | |
| 629 if (typeof siblingNode.readability !== 'undefined' && | |
| 630 (siblingNode.readability.contentScore+contentBonus) >= siblingScoreThreshold) | |
| 631 append = true; | |
| 632 | |
| 633 if (siblingNode.nodeName === "P") { | |
| 634 let linkDensity = this._getLinkDensity(siblingNode); | |
| 635 let nodeContent = this._getInnerText(siblingNode); | |
| 636 let nodeLength = nodeContent.length; | |
| 637 | |
| 638 if (nodeLength > 80 && linkDensity < 0.25) { | |
| 639 append = true; | |
| 640 } else if (nodeLength < 80 && linkDensity === 0 && nodeContent.search(/\.( |$)/) !== -1) { | |
| 641 append = true; | |
| 642 } | |
| 643 } | |
| 644 | |
| 645 if (append) { | |
| 646 this.log("Appending node: " + siblingNode); | |
| 647 | |
| 648 // siblingNodes is a reference to the childNodes array, and | |
| 649 // siblingNode is removed from the array when we call appendChild() | |
| 650 // below. As a result, we must revisit this index since the nodes | |
| 651 // have been shifted. | |
| 652 s -= 1; | |
| 653 sl -= 1; | |
| 654 | |
| 655 if (siblingNode.nodeName !== "DIV" && siblingNode.nodeName !== "P") { | |
| 656 // We have a node that isn't a common block level element, like a form or td tag. | |
| 657 // Turn it into a div so it doesn't get filtered out later by accident. */ | |
| 658 this.log("Altering siblingNode of " + siblingNode.nodeName + ' to div.'); | |
| 659 | |
| 660 this._setNodeTag(siblingNode, "DIV"); | |
| 661 } | |
| 662 | |
| 663 // To ensure a node does not interfere with readability styles, | |
| 664 // remove its classnames. | |
| 665 siblingNode.className = ""; | |
| 666 | |
| 667 // Append sibling and subtract from our list because it removes | |
| 668 // the node when you append to another node. | |
| 669 articleContent.appendChild(siblingNode); | |
| 670 } | |
| 671 } | |
| 672 | |
| 673 // So we have all of the content that we need. Now we clean it up for presentation. | |
| 674 this._prepArticle(articleContent); | |
| 675 | |
| 676 if (this._curPageNum === 1) { | |
| 677 let div = doc.createElement("DIV"); | |
| 678 div.id = "readability-page-1"; | |
| 679 div.className = "page"; | |
| 680 let children = articleContent.childNodes; | |
| 681 for (let i = 0; i < children.length; ++i) { | |
| 682 div.appendChild(children[i]); | |
| 683 } | |
| 684 articleContent.appendChild(div); | |
| 685 } | |
| 686 | |
| 687 // Now that we've gone through the full algorithm, check to see if | |
| 688 // we got any meaningful content. If we didn't, we may need to re-run | |
| 689 // grabArticle with different flags set. This gives us a higher likelihood of | |
| 690 // finding the content, and the sieve approach gives us a higher likelihood of | |
| 691 // finding the -right- content. | |
| 692 if (this._getInnerText(articleContent, true).length < 500) { | |
| 693 page.innerHTML = pageCacheHtml; | |
| 694 | |
| 695 if (this._flagIsActive(this.FLAG_STRIP_UNLIKELYS)) { | |
| 696 this._removeFlag(this.FLAG_STRIP_UNLIKELYS); | |
| 697 } else if (this._flagIsActive(this.FLAG_WEIGHT_CLASSES)) { | |
| 698 this._removeFlag(this.FLAG_WEIGHT_CLASSES); | |
| 699 } else if (this._flagIsActive(this.FLAG_CLEAN_CONDITIONALLY)) { | |
| 700 this._removeFlag(this.FLAG_CLEAN_CONDITIONALLY); | |
| 701 } else { | |
| 702 return null; | |
| 703 } | |
| 704 } else { | |
| 705 if (lastTopCandidate !== null) { | |
| 706 // EXPERIMENTAL: Contrast ratio is how we measure the level of competition between candidates in the | |
| 707 // readability algorithm. This is to avoid offering reader mode on pages that are more like | |
| 708 // a list or directory of links with summaries. It takes the score of the last top candidate | |
| 709 // (see N_TOP_CANDIDATES) and checks how it compares to the top candidate's. On pages that are not | |
| 710 // actual articles, there will likely be many candidates with similar score (i.e. higher contrast ratio). | |
| 711 let contrastRatio = lastTopCandidate.readability.contentScore / topCandidate.readability.contentScore; | |
| 712 if (contrastRatio > 0.45) | |
| 713 return null; | |
| 714 } | |
| 715 | |
| 716 return articleContent; | |
| 717 } | |
| 718 } | |
| 719 }, | |
| 720 | |
| 721 /** | |
| 722 * Attempts to get the excerpt from these | |
| 723 * sources in the following order: | |
| 724 * - meta description tag | |
| 725 * - open-graph description | |
| 726 * - twitter cards description | |
| 727 * - article's first paragraph | |
| 728 * If no excerpt is found, an empty string will be | |
| 729 * returned. | |
| 730 * | |
| 731 * @param Element - root element of the processed version page | |
| 732 * @return String - excerpt of the article | |
| 733 **/ | |
| 734 _getExcerpt: function(articleContent) { | |
| 735 let values = {}; | |
| 736 let metaElements = this._doc.getElementsByTagName("meta"); | |
| 737 | |
| 738 // Match "description", or Twitter's "twitter:description" (Cards) | |
| 739 // in name attribute. | |
| 740 let namePattern = /^\s*((twitter)\s*:\s*)?description\s*$/gi; | |
| 741 | |
| 742 // Match Facebook's og:description (Open Graph) in property attribute. | |
| 743 let propertyPattern = /^\s*og\s*:\s*description\s*$/gi; | |
| 744 | |
| 745 // Find description tags. | |
| 746 for (let i = 0; i < metaElements.length; i++) { | |
| 747 let element = metaElements[i]; | |
| 748 let elementName = element.getAttribute("name"); | |
| 749 let elementProperty = element.getAttribute("property"); | |
| 750 | |
| 751 let name; | |
| 752 if (namePattern.test(elementName)) { | |
| 753 name = elementName; | |
| 754 } else if (propertyPattern.test(elementProperty)) { | |
| 755 name = elementProperty; | |
| 756 } | |
| 757 | |
| 758 if (name) { | |
| 759 let content = element.getAttribute("content"); | |
| 760 if (content) { | |
| 761 // Convert to lowercase and remove any whitespace | |
| 762 // so we can match below. | |
| 763 name = name.toLowerCase().replace(/\s/g, ''); | |
| 764 values[name] = content.trim(); | |
| 765 } | |
| 766 } | |
| 767 } | |
| 768 | |
| 769 if ("description" in values) { | |
| 770 return values["description"]; | |
| 771 } | |
| 772 | |
| 773 if ("og:description" in values) { | |
| 774 // Use facebook open graph description. | |
| 775 return values["og:description"]; | |
| 776 } | |
| 777 | |
| 778 if ("twitter:description" in values) { | |
| 779 // Use twitter cards description. | |
| 780 return values["twitter:description"]; | |
| 781 } | |
| 782 | |
| 783 // No description meta tags, use the article's first paragraph. | |
| 784 let paragraphs = articleContent.getElementsByTagName("p"); | |
| 785 if (paragraphs.length > 0) { | |
| 786 return paragraphs[0].textContent; | |
| 787 } | |
| 788 | |
| 789 return ""; | |
| 790 }, | |
| 791 | |
| 792 /** | |
| 793 * Removes script tags from the document. | |
| 794 * | |
| 795 * @param Element | |
| 796 **/ | |
| 797 _removeScripts: function(doc) { | |
| 798 let scripts = doc.getElementsByTagName('script'); | |
| 799 for (let i = scripts.length - 1; i >= 0; i -= 1) { | |
| 800 scripts[i].nodeValue=""; | |
| 801 scripts[i].removeAttribute('src'); | |
| 802 | |
| 803 if (scripts[i].parentNode) | |
| 804 scripts[i].parentNode.removeChild(scripts[i]); | |
| 805 } | |
| 806 }, | |
| 807 | |
| 808 /** | |
| 809 * Get child index of the only P element inside a DIV with no | |
| 810 * text content. Returns -1 if the DIV node contains non-empty | |
| 811 * text nodes or if it contains other element nodes. | |
| 812 * | |
| 813 * @param Element | |
| 814 **/ | |
| 815 _getSinglePIndexInsideDiv: function(e) { | |
| 816 let childNodes = e.childNodes; | |
| 817 let pIndex = -1; | |
| 818 | |
| 819 for (let i = childNodes.length; --i >= 0;) { | |
| 820 let node = childNodes[i]; | |
| 821 | |
| 822 if (node.nodeType === Node.ELEMENT_NODE) { | |
| 823 if (node.tagName !== "P") | |
| 824 return -1; | |
| 825 | |
| 826 if (pIndex >= 0) | |
| 827 return -1; | |
| 828 | |
| 829 pIndex = i; | |
| 830 } else if (node.nodeType == Node.TEXT_NODE && this._getInnerText(node, false)) { | |
| 831 return -1; | |
| 832 } | |
| 833 } | |
| 834 | |
| 835 return pIndex; | |
| 836 }, | |
| 837 | |
| 838 /** | |
| 839 * Determine whether element has any children block level elements. | |
| 840 * | |
| 841 * @param Element | |
| 842 */ | |
| 843 _hasChildBlockElement: function (e) { | |
| 844 let length = e.childNodes.length; | |
| 845 for (let i = 0; i < length; i++) { | |
| 846 let child = e.childNodes[i]; | |
| 847 if (child.nodeType != 1) | |
| 848 continue; | |
| 849 | |
| 850 if (this.DIV_TO_P_ELEMS.indexOf(child.tagName) !== -1 || this._hasChildBlockElement(child)) | |
| 851 return true; | |
| 852 } | |
| 853 return false; | |
| 854 }, | |
| 855 | |
| 856 /** | |
| 857 * Get the inner text of a node - cross browser compatibly. | |
| 858 * This also strips out any excess whitespace to be found. | |
| 859 * | |
| 860 * @param Element | |
| 861 * @return string | |
| 862 **/ | |
| 863 _getInnerText: function(e, normalizeSpaces) { | |
| 864 let textContent = e.textContent.replace(this.REGEXPS.trim, ""); | |
| 865 normalizeSpaces = (typeof normalizeSpaces === 'undefined') ? true : normalizeSpaces; | |
| 866 | |
| 867 if (normalizeSpaces) { | |
| 868 return textContent.replace(this.REGEXPS.normalize, " "); | |
| 869 } else { | |
| 870 return textContent; | |
| 871 } | |
| 872 }, | |
| 873 | |
| 874 /** | |
| 875 * Get the number of times a string s appears in the node e. | |
| 876 * | |
| 877 * @param Element | |
| 878 * @param string - what to split on. Default is "," | |
| 879 * @return number (integer) | |
| 880 **/ | |
| 881 _getCharCount: function(e,s) { | |
| 882 s = s || ","; | |
| 883 return this._getInnerText(e).split(s).length - 1; | |
| 884 }, | |
| 885 | |
| 886 /** | |
| 887 * Remove the style attribute on every e and under. | |
| 888 * TODO: Test if getElementsByTagName(*) is faster. | |
| 889 * | |
| 890 * @param Element | |
| 891 * @return void | |
| 892 **/ | |
| 893 _cleanStyles: function(e) { | |
| 894 e = e || this._doc; | |
| 895 let cur = e.firstChild; | |
| 896 | |
| 897 if (!e) | |
| 898 return; | |
| 899 | |
| 900 // Remove any root styles, if we're able. | |
| 901 if (typeof e.removeAttribute === 'function' && e.className !== 'readability-styled') | |
| 902 e.removeAttribute('style'); | |
| 903 | |
| 904 // Go until there are no more child nodes | |
| 905 while (cur !== null) { | |
| 906 if (cur.nodeType === 1) { | |
| 907 // Remove style attribute(s) : | |
| 908 if (cur.className !== "readability-styled") | |
| 909 cur.removeAttribute("style"); | |
| 910 | |
| 911 this._cleanStyles(cur); | |
| 912 } | |
| 913 | |
| 914 cur = cur.nextSibling; | |
| 915 } | |
| 916 }, | |
| 917 | |
| 918 /** | |
| 919 * Get the density of links as a percentage of the content | |
| 920 * This is the amount of text that is inside a link divided by the total text in the node. | |
| 921 * | |
| 922 * @param Element | |
| 923 * @return number (float) | |
| 924 **/ | |
| 925 _getLinkDensity: function(e) { | |
| 926 let links = e.getElementsByTagName("a"); | |
| 927 let textLength = this._getInnerText(e).length; | |
| 928 let linkLength = 0; | |
| 929 | |
| 930 for (let i = 0, il = links.length; i < il; i += 1) { | |
| 931 linkLength += this._getInnerText(links[i]).length; | |
| 932 } | |
| 933 | |
| 934 return linkLength / textLength; | |
| 935 }, | |
| 936 | |
| 937 /** | |
| 938 * Find a cleaned up version of the current URL, to use for comparing links for possible next-pageyness. | |
| 939 * | |
| 940 * @author Dan Lacy | |
| 941 * @return string the base url | |
| 942 **/ | |
| 943 _findBaseUrl: function() { | |
| 944 let uri = this._uri; | |
| 945 let noUrlParams = uri.path.split("?")[0]; | |
| 946 let urlSlashes = noUrlParams.split("/").reverse(); | |
| 947 let cleanedSegments = []; | |
| 948 let possibleType = ""; | |
| 949 | |
| 950 for (let i = 0, slashLen = urlSlashes.length; i < slashLen; i += 1) { | |
| 951 let segment = urlSlashes[i]; | |
| 952 | |
| 953 // Split off and save anything that looks like a file type. | |
| 954 if (segment.indexOf(".") !== -1) { | |
| 955 possibleType = segment.split(".")[1]; | |
| 956 | |
| 957 // If the type isn't alpha-only, it's probably not actually a file extension. | |
| 958 if (!possibleType.match(/[^a-zA-Z]/)) | |
| 959 segment = segment.split(".")[0]; | |
| 960 } | |
| 961 | |
| 962 // EW-CMS specific segment replacement. Ugly. | |
| 963 // Example: http://www.ew.com/ew/article/0,,20313460_20369436,00.html | |
| 964 if (segment.indexOf(',00') !== -1) | |
| 965 segment = segment.replace(',00', ''); | |
| 966 | |
| 967 // If our first or second segment has anything looking like a page number, remove it. | |
| 968 if (segment.match(/((_|-)?p[a-z]*|(_|-))[0-9]{1,2}$/i) && ((i === 1) || (i === 0))) | |
| 969 segment = segment.replace(/((_|-)?p[a-z]*|(_|-))[0-9]{1,2}$/i, ""); | |
| 970 | |
| 971 let del = false; | |
| 972 | |
| 973 // If this is purely a number, and it's the first or second segment, | |
| 974 // it's probably a page number. Remove it. | |
| 975 if (i < 2 && segment.match(/^\d{1,2}$/)) | |
| 976 del = true; | |
| 977 | |
| 978 // If this is the first segment and it's just "index", remove it. | |
| 979 if (i === 0 && segment.toLowerCase() === "index") | |
| 980 del = true; | |
| 981 | |
| 982 // If our first or second segment is smaller than 3 characters, | |
| 983 // and the first segment was purely alphas, remove it. | |
| 984 if (i < 2 && segment.length < 3 && !urlSlashes[0].match(/[a-z]/i)) | |
| 985 del = true; | |
| 986 | |
| 987 // If it's not marked for deletion, push it to cleanedSegments. | |
| 988 if (!del) | |
| 989 cleanedSegments.push(segment); | |
| 990 } | |
| 991 | |
| 992 // This is our final, cleaned, base article URL. | |
| 993 return uri.scheme + "://" + uri.host + cleanedSegments.reverse().join("/"); | |
| 994 }, | |
| 995 | |
| 996 /** | |
| 997 * Look for any paging links that may occur within the document. | |
| 998 * | |
| 999 * @param body | |
| 1000 * @return object (array) | |
| 1001 **/ | |
| 1002 _findNextPageLink: function(elem) { | |
| 1003 let uri = this._uri; | |
| 1004 let possiblePages = {}; | |
| 1005 let allLinks = elem.getElementsByTagName('a'); | |
| 1006 let articleBaseUrl = this._findBaseUrl(); | |
| 1007 | |
| 1008 // Loop through all links, looking for hints that they may be next-page links. | |
| 1009 // Things like having "page" in their textContent, className or id, or being a child | |
| 1010 // of a node with a page-y className or id. | |
| 1011 // | |
| 1012 // Also possible: levenshtein distance? longest common subsequence? | |
| 1013 // | |
| 1014 // After we do that, assign each page a score, and | |
| 1015 for (let i = 0, il = allLinks.length; i < il; i += 1) { | |
| 1016 let link = allLinks[i]; | |
| 1017 let linkHref = allLinks[i].href.replace(/#.*$/, '').replace(/\/$/, ''); | |
| 1018 | |
| 1019 // If we've already seen this page, ignore it. | |
| 1020 if (linkHref === "" || | |
| 1021 linkHref === articleBaseUrl || | |
| 1022 linkHref === uri.spec || | |
| 1023 linkHref in this._parsedPages) { | |
| 1024 continue; | |
| 1025 } | |
| 1026 | |
| 1027 // If it's on a different domain, skip it. | |
| 1028 if (uri.host !== linkHref.split(/\/+/g)[1]) | |
| 1029 continue; | |
| 1030 | |
| 1031 let linkText = this._getInnerText(link); | |
| 1032 | |
| 1033 // If the linkText looks like it's not the next page, skip it. | |
| 1034 if (linkText.match(this.REGEXPS.extraneous) || linkText.length > 25) | |
| 1035 continue; | |
| 1036 | |
| 1037 // If the leftovers of the URL after removing the base URL don't contain | |
| 1038 // any digits, it's certainly not a next page link. | |
| 1039 let linkHrefLeftover = linkHref.replace(articleBaseUrl, ''); | |
| 1040 if (!linkHrefLeftover.match(/\d/)) | |
| 1041 continue; | |
| 1042 | |
| 1043 if (!(linkHref in possiblePages)) { | |
| 1044 possiblePages[linkHref] = {"score": 0, "linkText": linkText, "href": linkHref}; | |
| 1045 } else { | |
| 1046 possiblePages[linkHref].linkText += ' | ' + linkText; | |
| 1047 } | |
| 1048 | |
| 1049 let linkObj = possiblePages[linkHref]; | |
| 1050 | |
| 1051 // If the articleBaseUrl isn't part of this URL, penalize this link. It could | |
| 1052 // still be the link, but the odds are lower. | |
| 1053 // Example: http://www.actionscript.org/resources/articles/745/1/JavaScript-and-VBScript-Injection-in-ActionScript-3/Page1.html | |
| 1054 if (linkHref.indexOf(articleBaseUrl) !== 0) | |
| 1055 linkObj.score -= 25; | |
| 1056 | |
| 1057 let linkData = linkText + ' ' + link.className + ' ' + link.id; | |
| 1058 if (linkData.match(this.REGEXPS.nextLink)) | |
| 1059 linkObj.score += 50; | |
| 1060 | |
| 1061 if (linkData.match(/pag(e|ing|inat)/i)) | |
| 1062 linkObj.score += 25; | |
| 1063 | |
| 1064 if (linkData.match(/(first|last)/i)) { | |
| 1065 // -65 is enough to negate any bonuses gotten from a > or » in the text, | |
| 1066 // If we already matched on "next", last is probably fine. | |
| 1067 // If we didn't, then it's bad. Penalize. | |
| 1068 if (!linkObj.linkText.match(this.REGEXPS.nextLink)) | |
| 1069 linkObj.score -= 65; | |
| 1070 } | |
| 1071 | |
| 1072 if (linkData.match(this.REGEXPS.negative) || linkData.match(this.REGEXPS.extraneous)) | |
| 1073 linkObj.score -= 50; | |
| 1074 | |
| 1075 if (linkData.match(this.REGEXPS.prevLink)) | |
| 1076 linkObj.score -= 200; | |
| 1077 | |
| 1078 // If a parentNode contains page or paging or paginat | |
| 1079 let parentNode = link.parentNode; | |
| 1080 let positiveNodeMatch = false; | |
| 1081 let negativeNodeMatch = false; | |
| 1082 | |
| 1083 while (parentNode) { | |
| 1084 let parentNodeClassAndId = parentNode.className + ' ' + parentNode.id; | |
| 1085 | |
| 1086 if (!positiveNodeMatch && parentNodeClassAndId && parentNodeClassAndId.match(/pag(e|ing|inat)/i)) { | |
| 1087 positiveNodeMatch = true; | |
| 1088 linkObj.score += 25; | |
| 1089 } | |
| 1090 | |
| 1091 if (!negativeNodeMatch && parentNodeClassAndId && parentNodeClassAndId.match(this.REGEXPS.negative)) { | |
| 1092 // If this is just something like "footer", give it a negative. | |
| 1093 // If it's something like "body-and-footer", leave it be. | |
| 1094 if (!parentNodeClassAndId.match(this.REGEXPS.positive)) { | |
| 1095 linkObj.score -= 25; | |
| 1096 negativeNodeMatch = true; | |
| 1097 } | |
| 1098 } | |
| 1099 | |
| 1100 parentNode = parentNode.parentNode; | |
| 1101 } | |
| 1102 | |
| 1103 // If the URL looks like it has paging in it, add to the score. | |
| 1104 // Things like /page/2/, /pagenum/2, ?p=3, ?page=11, ?pagination=34 | |
| 1105 if (linkHref.match(/p(a|g|ag)?(e|ing|ination)?(=|\/)[0-9]{1,2}/i) || linkHref.match(/(page|paging)/i)) | |
| 1106 linkObj.score += 25; | |
| 1107 | |
| 1108 // If the URL contains negative values, give a slight decrease. | |
| 1109 if (linkHref.match(this.REGEXPS.extraneous)) | |
| 1110 linkObj.score -= 15; | |
| 1111 | |
| 1112 /** | |
| 1113 * Minor punishment to anything that doesn't match our current URL. | |
| 1114 * NOTE: I'm finding this to cause more harm than good where something is exactly 50 points. | |
| 1115 * Dan, can you show me a counterexample where this is necessary? | |
| 1116 * if (linkHref.indexOf(window.location.href) !== 0) { | |
| 1117 * linkObj.score -= 1; | |
| 1118 * } | |
| 1119 **/ | |
| 1120 | |
| 1121 // If the link text can be parsed as a number, give it a minor bonus, with a slight | |
| 1122 // bias towards lower numbered pages. This is so that pages that might not have 'next' | |
| 1123 // in their text can still get scored, and sorted properly by score. | |
| 1124 let linkTextAsNumber = parseInt(linkText, 10); | |
| 1125 if (linkTextAsNumber) { | |
| 1126 // Punish 1 since we're either already there, or it's probably | |
| 1127 // before what we want anyways. | |
| 1128 if (linkTextAsNumber === 1) { | |
| 1129 linkObj.score -= 10; | |
| 1130 } else { | |
| 1131 linkObj.score += Math.max(0, 10 - linkTextAsNumber); | |
| 1132 } | |
| 1133 } | |
| 1134 } | |
| 1135 | |
| 1136 // Loop thrugh all of our possible pages from above and find our top | |
| 1137 // candidate for the next page URL. Require at least a score of 50, which | |
| 1138 // is a relatively high confidence that this page is the next link. | |
| 1139 let topPage = null; | |
| 1140 for (let page in possiblePages) { | |
| 1141 if (possiblePages.hasOwnProperty(page)) { | |
| 1142 if (possiblePages[page].score >= 50 && | |
| 1143 (!topPage || topPage.score < possiblePages[page].score)) | |
| 1144 topPage = possiblePages[page]; | |
| 1145 } | |
| 1146 } | |
| 1147 | |
| 1148 if (topPage) { | |
| 1149 let nextHref = topPage.href.replace(/\/$/,''); | |
| 1150 | |
| 1151 this.log('NEXT PAGE IS ' + nextHref); | |
| 1152 this._parsedPages[nextHref] = true; | |
| 1153 return nextHref; | |
| 1154 } else { | |
| 1155 return null; | |
| 1156 } | |
| 1157 }, | |
| 1158 | |
| 1159 _successfulRequest: function(request) { | |
| 1160 return (request.status >= 200 && request.status < 300) || | |
| 1161 request.status === 304 || | |
| 1162 (request.status === 0 && request.responseText); | |
| 1163 }, | |
| 1164 | |
| 1165 _ajax: function(url, options) { | |
| 1166 let request = new XMLHttpRequest(); | |
| 1167 | |
| 1168 function respondToReadyState(readyState) { | |
| 1169 if (request.readyState === 4) { | |
| 1170 if (this._successfulRequest(request)) { | |
| 1171 if (options.success) | |
| 1172 options.success(request); | |
| 1173 } else { | |
| 1174 if (options.error) | |
| 1175 options.error(request); | |
| 1176 } | |
| 1177 } | |
| 1178 } | |
| 1179 | |
| 1180 if (typeof options === 'undefined') | |
| 1181 options = {}; | |
| 1182 | |
| 1183 request.onreadystatechange = respondToReadyState; | |
| 1184 | |
| 1185 request.open('get', url, true); | |
| 1186 request.setRequestHeader('Accept', 'text/html'); | |
| 1187 | |
| 1188 try { | |
| 1189 request.send(options.postBody); | |
| 1190 } catch (e) { | |
| 1191 if (options.error) | |
| 1192 options.error(); | |
| 1193 } | |
| 1194 | |
| 1195 return request; | |
| 1196 }, | |
| 1197 | |
| 1198 _appendNextPage: function(nextPageLink) { | |
| 1199 let doc = this._doc; | |
| 1200 this._curPageNum += 1; | |
| 1201 | |
| 1202 let articlePage = doc.createElement("DIV"); | |
| 1203 articlePage.id = 'readability-page-' + this._curPageNum; | |
| 1204 articlePage.className = 'page'; | |
| 1205 articlePage.innerHTML = '<p class="page-separator" title="Page ' + this._curPageNum + '">§</p>'; | |
| 1206 | |
| 1207 doc.getElementById("readability-content").appendChild(articlePage); | |
| 1208 | |
| 1209 if (this._curPageNum > this.MAX_PAGES) { | |
| 1210 let nextPageMarkup = "<div style='text-align: center'><a href='" + nextPageLink + "'>View Next Page</a></div>"; | |
| 1211 articlePage.innerHTML = articlePage.innerHTML + nextPageMarkup; | |
| 1212 return; | |
| 1213 } | |
| 1214 | |
| 1215 // Now that we've built the article page DOM element, get the page content | |
| 1216 // asynchronously and load the cleaned content into the div we created for it. | |
| 1217 (function(pageUrl, thisPage) { | |
| 1218 this._ajax(pageUrl, { | |
| 1219 success: function(r) { | |
| 1220 | |
| 1221 // First, check to see if we have a matching ETag in headers - if we do, this is a duplicate page. | |
| 1222 let eTag = r.getResponseHeader('ETag'); | |
| 1223 if (eTag) { | |
| 1224 if (eTag in this._pageETags) { | |
| 1225 this.log("Exact duplicate page found via ETag. Aborting."); | |
| 1226 articlePage.style.display = 'none'; | |
| 1227 return; | |
| 1228 } else { | |
| 1229 this._pageETags[eTag] = 1; | |
| 1230 } | |
| 1231 } | |
| 1232 | |
| 1233 // TODO: this ends up doubling up page numbers on NYTimes articles. Need to generically parse those away. | |
| 1234 let page = doc.createElement("DIV"); | |
| 1235 | |
| 1236 // Do some preprocessing to our HTML to make it ready for appending. | |
| 1237 // - Remove any script tags. Swap and reswap newlines with a unicode | |
| 1238 // character because multiline regex doesn't work in javascript. | |
| 1239 // - Turn any noscript tags into divs so that we can parse them. This | |
| 1240 // allows us to find any next page links hidden via javascript. | |
| 1241 // - Turn all double br's into p's - was handled by prepDocument in the original view. | |
| 1242 // Maybe in the future abstract out prepDocument to work for both the original document | |
| 1243 // and AJAX-added pages. | |
| 1244 let responseHtml = r.responseText.replace(/\n/g,'\uffff').replace(/<script.*?>.*?<\/script>/gi, ''); | |
| 1245 responseHtml = responseHtml.replace(/\n/g,'\uffff').replace(/<script.*?>.*?<\/script>/gi, ''); | |
| 1246 responseHtml = responseHtml.replace(/\uffff/g,'\n').replace(/<(\/?)noscript/gi, '<$1div'); | |
| 1247 responseHtml = responseHtml.replace(this.REGEXPS.replaceFonts, '<$1span>'); | |
| 1248 | |
| 1249 page.innerHTML = responseHtml; | |
| 1250 this._replaceBrs(page); | |
| 1251 | |
| 1252 // Reset all flags for the next page, as they will search through it and | |
| 1253 // disable as necessary at the end of grabArticle. | |
| 1254 this._flags = 0x1 | 0x2 | 0x4; | |
| 1255 | |
| 1256 let nextPageLink = this._findNextPageLink(page); | |
| 1257 | |
| 1258 // NOTE: if we end up supporting _appendNextPage(), we'll need to | |
| 1259 // change this call to be async | |
| 1260 let content = this._grabArticle(page); | |
| 1261 | |
| 1262 if (!content) { | |
| 1263 this.log("No content found in page to append. Aborting."); | |
| 1264 return; | |
| 1265 } | |
| 1266 | |
| 1267 // Anti-duplicate mechanism. Essentially, get the first paragraph of our new page. | |
| 1268 // Compare it against all of the the previous document's we've gotten. If the previous | |
| 1269 // document contains exactly the innerHTML of this first paragraph, it's probably a duplicate. | |
| 1270 let firstP = content.getElementsByTagName("P").length ? content.getElementsByTagName("P")[0] : null; | |
| 1271 if (firstP && firstP.innerHTML.length > 100) { | |
| 1272 for (let i = 1; i <= this._curPageNum; i += 1) { | |
| 1273 let rPage = doc.getElementById('readability-page-' + i); | |
| 1274 if (rPage && rPage.innerHTML.indexOf(firstP.innerHTML) !== -1) { | |
| 1275 this.log('Duplicate of page ' + i + ' - skipping.'); | |
| 1276 articlePage.style.display = 'none'; | |
| 1277 this._parsedPages[pageUrl] = true; | |
| 1278 return; | |
| 1279 } | |
| 1280 } | |
| 1281 } | |
| 1282 | |
| 1283 this._removeScripts(content); | |
| 1284 | |
| 1285 thisPage.innerHTML = thisPage.innerHTML + content.innerHTML; | |
| 1286 | |
| 1287 // After the page has rendered, post process the content. This delay is necessary because, | |
| 1288 // in webkit at least, offsetWidth is not set in time to determine image width. We have to | |
| 1289 // wait a little bit for reflow to finish before we can fix floating images. | |
| 1290 setTimeout((function() { | |
| 1291 this._postProcessContent(thisPage); | |
| 1292 }).bind(this), 500); | |
| 1293 | |
| 1294 | |
| 1295 if (nextPageLink) | |
| 1296 this._appendNextPage(nextPageLink); | |
| 1297 } | |
| 1298 }); | |
| 1299 }).bind(this)(nextPageLink, articlePage); | |
| 1300 }, | |
| 1301 | |
| 1302 /** | |
| 1303 * Get an elements class/id weight. Uses regular expressions to tell if this | |
| 1304 * element looks good or bad. | |
| 1305 * | |
| 1306 * @param Element | |
| 1307 * @return number (Integer) | |
| 1308 **/ | |
| 1309 _getClassWeight: function(e) { | |
| 1310 if (!this._flagIsActive(this.FLAG_WEIGHT_CLASSES)) | |
| 1311 return 0; | |
| 1312 | |
| 1313 let weight = 0; | |
| 1314 | |
| 1315 // Look for a special classname | |
| 1316 if (typeof(e.className) === 'string' && e.className !== '') { | |
| 1317 if (e.className.search(this.REGEXPS.negative) !== -1) | |
| 1318 weight -= 25; | |
| 1319 | |
| 1320 if (e.className.search(this.REGEXPS.positive) !== -1) | |
| 1321 weight += 25; | |
| 1322 } | |
| 1323 | |
| 1324 // Look for a special ID | |
| 1325 if (typeof(e.id) === 'string' && e.id !== '') { | |
| 1326 if (e.id.search(this.REGEXPS.negative) !== -1) | |
| 1327 weight -= 25; | |
| 1328 | |
| 1329 if (e.id.search(this.REGEXPS.positive) !== -1) | |
| 1330 weight += 25; | |
| 1331 } | |
| 1332 | |
| 1333 return weight; | |
| 1334 }, | |
| 1335 | |
| 1336 /** | |
| 1337 * Clean a node of all elements of type "tag". | |
| 1338 * (Unless it's a youtube/vimeo video. People love movies.) | |
| 1339 * | |
| 1340 * @param Element | |
| 1341 * @param string tag to clean | |
| 1342 * @return void | |
| 1343 **/ | |
| 1344 _clean: function(e, tag) { | |
| 1345 let targetList = e.getElementsByTagName(tag); | |
| 1346 let isEmbed = (tag === 'object' || tag === 'embed'); | |
| 1347 | |
| 1348 for (let y = targetList.length - 1; y >= 0; y -= 1) { | |
| 1349 // Allow youtube and vimeo videos through as people usually want to see those. | |
| 1350 if (isEmbed) { | |
| 1351 let attributeValues = ""; | |
| 1352 for (let i = 0, il = targetList[y].attributes.length; i < il; i += 1) { | |
| 1353 attributeValues += targetList[y].attributes[i].value + '|'; | |
| 1354 } | |
| 1355 | |
| 1356 // First, check the elements attributes to see if any of them contain youtube or vimeo | |
| 1357 if (attributeValues.search(this.REGEXPS.videos) !== -1) | |
| 1358 continue; | |
| 1359 | |
| 1360 // Then check the elements inside this element for the same. | |
| 1361 if (targetList[y].innerHTML.search(this.REGEXPS.videos) !== -1) | |
| 1362 continue; | |
| 1363 } | |
| 1364 | |
| 1365 targetList[y].parentNode.removeChild(targetList[y]); | |
| 1366 } | |
| 1367 }, | |
| 1368 | |
| 1369 /** | |
| 1370 * Clean an element of all tags of type "tag" if they look fishy. | |
| 1371 * "Fishy" is an algorithm based on content length, classnames, link density, number of images & embeds, etc. | |
| 1372 * | |
| 1373 * @return void | |
| 1374 **/ | |
| 1375 _cleanConditionally: function(e, tag) { | |
| 1376 if (!this._flagIsActive(this.FLAG_CLEAN_CONDITIONALLY)) | |
| 1377 return; | |
| 1378 | |
| 1379 let tagsList = e.getElementsByTagName(tag); | |
| 1380 let curTagsLength = tagsList.length; | |
| 1381 | |
| 1382 // Gather counts for other typical elements embedded within. | |
| 1383 // Traverse backwards so we can remove nodes at the same time | |
| 1384 // without effecting the traversal. | |
| 1385 // | |
| 1386 // TODO: Consider taking into account original contentScore here. | |
| 1387 for (let i = curTagsLength-1; i >= 0; i -= 1) { | |
| 1388 let weight = this._getClassWeight(tagsList[i]); | |
| 1389 let contentScore = 0; | |
| 1390 | |
| 1391 this.log("Cleaning Conditionally " + tagsList[i] + " (" + tagsList[i].className + ":" + tagsList[i].id + ")"); | |
| 1392 | |
| 1393 if (weight + contentScore < 0) { | |
| 1394 tagsList[i].parentNode.removeChild(tagsList[i]); | |
| 1395 } else if (this._getCharCount(tagsList[i],',') < 10) { | |
| 1396 // If there are not very many commas, and the number of | |
| 1397 // non-paragraph elements is more than paragraphs or other | |
| 1398 // ominous signs, remove the element. | |
| 1399 let p = tagsList[i].getElementsByTagName("p").length; | |
| 1400 let img = tagsList[i].getElementsByTagName("img").length; | |
| 1401 let li = tagsList[i].getElementsByTagName("li").length-100; | |
| 1402 let input = tagsList[i].getElementsByTagName("input").length; | |
| 1403 | |
| 1404 let embedCount = 0; | |
| 1405 let embeds = tagsList[i].getElementsByTagName("embed"); | |
| 1406 for (let ei = 0, il = embeds.length; ei < il; ei += 1) { | |
| 1407 if (embeds[ei].src.search(this.REGEXPS.videos) === -1) | |
| 1408 embedCount += 1; | |
| 1409 } | |
| 1410 | |
| 1411 let linkDensity = this._getLinkDensity(tagsList[i]); | |
| 1412 let contentLength = this._getInnerText(tagsList[i]).length; | |
| 1413 let toRemove = false; | |
| 1414 | |
| 1415 if (img > p) { | |
| 1416 toRemove = true; | |
| 1417 } else if (li > p && tag !== "ul" && tag !== "ol") { | |
| 1418 toRemove = true; | |
| 1419 } else if ( input > Math.floor(p/3) ) { | |
| 1420 toRemove = true; | |
| 1421 } else if (contentLength < 25 && (img === 0 || img > 2) ) { | |
| 1422 toRemove = true; | |
| 1423 } else if (weight < 25 && linkDensity > 0.2) { | |
| 1424 toRemove = true; | |
| 1425 } else if (weight >= 25 && linkDensity > 0.5) { | |
| 1426 toRemove = true; | |
| 1427 } else if ((embedCount === 1 && contentLength < 75) || embedCount > 1) { | |
| 1428 toRemove = true; | |
| 1429 } | |
| 1430 | |
| 1431 if (toRemove) | |
| 1432 tagsList[i].parentNode.removeChild(tagsList[i]); | |
| 1433 } | |
| 1434 } | |
| 1435 }, | |
| 1436 | |
| 1437 /** | |
| 1438 * Clean out spurious headers from an Element. Checks things like classnames and link density. | |
| 1439 * | |
| 1440 * @param Element | |
| 1441 * @return void | |
| 1442 **/ | |
| 1443 _cleanHeaders: function(e) { | |
| 1444 for (let headerIndex = 1; headerIndex < 3; headerIndex += 1) { | |
| 1445 let headers = e.getElementsByTagName('h' + headerIndex); | |
| 1446 for (let i = headers.length - 1; i >= 0; i -= 1) { | |
| 1447 if (this._getClassWeight(headers[i]) < 0 || this._getLinkDensity(headers[i]) > 0.33) | |
| 1448 headers[i].parentNode.removeChild(headers[i]); | |
| 1449 } | |
| 1450 } | |
| 1451 }, | |
| 1452 | |
| 1453 _flagIsActive: function(flag) { | |
| 1454 return (this._flags & flag) > 0; | |
| 1455 }, | |
| 1456 | |
| 1457 _addFlag: function(flag) { | |
| 1458 this._flags = this._flags | flag; | |
| 1459 }, | |
| 1460 | |
| 1461 _removeFlag: function(flag) { | |
| 1462 this._flags = this._flags & ~flag; | |
| 1463 }, | |
| 1464 | |
| 1465 /** | |
| 1466 * Runs readability. | |
| 1467 * | |
| 1468 * Workflow: | |
| 1469 * 1. Prep the document by removing script tags, css, etc. | |
| 1470 * 2. Build readability's DOM tree. | |
| 1471 * 3. Grab the article content from the current dom tree. | |
| 1472 * 4. Replace the current DOM tree with the new one. | |
| 1473 * 5. Read peacefully. | |
| 1474 * | |
| 1475 * @return void | |
| 1476 **/ | |
| 1477 parse: function () { | |
| 1478 // Remove script tags from the document. | |
| 1479 this._removeScripts(this._doc); | |
| 1480 | |
| 1481 // FIXME: Disabled multi-page article support for now as it | |
| 1482 // needs more work on infrastructure. | |
| 1483 | |
| 1484 // Make sure this document is added to the list of parsed pages first, | |
| 1485 // so we don't double up on the first page. | |
| 1486 // this._parsedPages[uri.spec.replace(/\/$/, '')] = true; | |
| 1487 | |
| 1488 // Pull out any possible next page link first. | |
| 1489 // let nextPageLink = this._findNextPageLink(doc.body); | |
| 1490 | |
| 1491 this._prepDocument(); | |
| 1492 | |
| 1493 let articleTitle = this._getArticleTitle(); | |
| 1494 let articleContent = this._grabArticle(); | |
| 1495 if (!articleContent) | |
| 1496 return null; | |
| 1497 | |
| 1498 this._postProcessContent(articleContent); | |
| 1499 | |
| 1500 // if (nextPageLink) { | |
| 1501 // // Append any additional pages after a small timeout so that people | |
| 1502 // // can start reading without having to wait for this to finish processing. | |
| 1503 // setTimeout((function() { | |
| 1504 // this._appendNextPage(nextPageLink); | |
| 1505 // }).bind(this), 500); | |
| 1506 // } | |
| 1507 | |
| 1508 let excerpt = this._getExcerpt(articleContent); | |
| 1509 | |
| 1510 return { title: articleTitle, | |
| 1511 byline: this._articleByline, | |
| 1512 dir: this._articleDir, | |
| 1513 content: articleContent.innerHTML, | |
| 1514 length: articleContent.textContent.length, | |
| 1515 excerpt: excerpt }; | |
| 1516 } | |
| 1517 }; |
