From 72b8b5b63465ea649b4762e06ce530b2a0ba2d4b Mon Sep 17 00:00:00 2001 From: rleed <101502594+rleed@users.noreply.github.com> Date: Fri, 20 Oct 2023 21:23:50 -0300 Subject: [PATCH] Attempt to get publication year when auto-titling links (#520) * port date scraper from python * bug fixes and cleanup * bug fixes and cleanup * refactor * address comments * make it intuitive * Update timedate-scraper.js - lint * address review comments * cleanup * simplfy and don't use side effects --------- Co-authored-by: rleed --- api/resolvers/item.js | 8 +++- lib/timedate-scraper.js | 88 +++++++++++++++++++++++++++++++++++++++++ 2 files changed, 95 insertions(+), 1 deletion(-) create mode 100644 lib/timedate-scraper.js diff --git a/api/resolvers/item.js b/api/resolvers/item.js index 249f4931..6648dbfd 100644 --- a/api/resolvers/item.js +++ b/api/resolvers/item.js @@ -3,6 +3,7 @@ import { ensureProtocol, removeTracking } from '../../lib/url' import { serializeInvoicable } from './serial' import { decodeCursor, LIMIT, nextCursorEncoded } from '../../lib/cursor' import { getMetadata, metadataRuleSets } from 'page-metadata-parser' +import { ruleSet as publicationDateRuleSet } from '../../lib/timedate-scraper' import domino from 'domino' import { ITEM_SPAM_INTERVAL, ITEM_FILTER_THRESHOLD, @@ -17,6 +18,7 @@ import { advSchema, amountSchema, bountySchema, commentSchema, discussionSchema, import { sendUserNotification } from '../webPush' import { defaultCommentSort, isJob } from '../../lib/item' import { notifyItemParents, notifyUserSubscribers, notifyZapped } from '../../lib/push-notifications' +import { datePivot } from '../../lib/time' export async function commentFilterClause (me, models) { let clause = ` AND ("Item"."weightedVotes" - "Item"."weightedDownVotes" > -${ITEM_FILTER_THRESHOLD}` @@ -540,8 +542,12 @@ export default { const response = await fetch(ensureProtocol(url), { redirect: 'follow' }) const html = await response.text() const doc = domino.createWindow(html).document - const metadata = getMetadata(doc, url, { title: metadataRuleSets.title }) + const metadata = getMetadata(doc, url, { title: metadataRuleSets.title, publicationDate: publicationDateRuleSet }) + const dateHint = ` (${metadata.publicationDate?.getFullYear()})` + const moreThanOneYearAgo = metadata.publicationDate && metadata.publicationDate < datePivot(new Date(), { years: -1 }) + res.title = metadata?.title + if (moreThanOneYearAgo) res.title += dateHint } catch { } try { diff --git a/lib/timedate-scraper.js b/lib/timedate-scraper.js new file mode 100644 index 00000000..0774b05f --- /dev/null +++ b/lib/timedate-scraper.js @@ -0,0 +1,88 @@ +// Date rule for use with page-metadata-parser. +// Based on https://github.com/Webhose/article-date-extractor/blob/master/articleDateExtractor/__init__.py +// Usage: import ruleSet and use in a call similar to: getMetadata(doc, url, { publicationDate: ruleSet.publicationDate }) +// Some example URLs for testing purposes: + +// ld+json example from 2018: +// https://mhagemann.medium.com/how-to-add-structured-json-ld-data-to-nuxt-js-8bb5f7c8a2d + +// meta tag example from 2011: +// https://www.newyorker.com/magazine/2011/10/10/the-crypto-currency + +// YouTube example from 2018: +// https://www.youtube.com/watch?v=YSUVRj8iznU + +// A news article from 2023 (< 1 year, should not display a date): +// https://edition.cnn.com/politics/live-news/matt-gaetz-kevin-mccarthy-house-speakership-10-03-23/index.html + +function cleanDateStr (dateString) { + try { + return new Date(dateString.match(String.raw`[0-9].*[0-9]`)[0]) + } catch {} +} + +export function extractFromURL (url) { + // Regex by Newspaper3k - https://github.com/codelucas/newspaper/blob/master/newspaper/urls.py + const m = url.match(String.raw`([\./\-_]{0,1}(19|20)\d{2})[\./\-_]{0,1}(([0-3]{0,1}[0-9][\./\-_])|(\w{3,5}[\./\-_]))([0-3]{0,1}[0-9][\./\-]{0,1})?`) + if (m) { + return cleanDateStr(m[0]) + } +} + +function asDate (str) { + if (str) { + try { + const d = new Date(str) + if (!isNaN(d)) return d + } catch { } + } +} + +export const ruleSet = { + // note meta names are case sensitive, and scorers must not favor rules when they will not return good results. + rules: [ + ['script[type="application/ld+json"]', node => asDate(JSON.parse(node.innerHTML)?.datePublished)], + ['script[type="application/ld+json"]', node => asDate(JSON.parse(node.innerHTML)?.dateCreated)], + ['script[type="application/ld+json"]', node => asDate(JSON.parse(node.innerHTML)?.dateModified)], + + ['meta[property="article:published_time"]', node => asDate(node.getAttribute('content'))], + ['meta[name="pubdate"]', node => asDate(node.getAttribute('content'))], + ['meta[name="publishdate"]', node => asDate(node.getAttribute('content'))], + ['meta[name="timestamp"]', node => asDate(node.getAttribute('content'))], + ['meta[name="dc.date.issued"]', node => asDate(node.getAttribute('content'))], + ['meta[name="date"]', node => asDate(node.getAttribute('content'))], + ['meta[property="bt:pubdate"]', node => asDate(node.getAttribute('content'))], + ['meta[name="parsely-pub-date"]', node => asDate(node.getAttribute('content'))], + ['meta[name="sailthru.date"]', node => asDate(node.getAttribute('content'))], + ['meta[name="article.published"]', node => asDate(node.getAttribute('content'))], + ['meta[name="published-date"]', node => asDate(node.getAttribute('content'))], + ['meta[name="article.created"]', node => asDate(node.getAttribute('content'))], + ['meta[name="article_date_original"]', node => asDate(node.getAttribute('content'))], + ['meta[name="cxenseparse:recs:publishtime"]', node => asDate(node.getAttribute('content'))], + ['meta[name="date_published"]', node => asDate(node.getAttribute('content'))], + ['meta[itemprop="datePublished"]', node => asDate(node.getAttribute('content'))], + ['meta[itemprop="datepublished"]', node => asDate(node.getAttribute('content'))], + ['meta[itemprop="datecreated"]', node => asDate(node.getAttribute('content'))], + ['meta[http-equiv="date"]', node => asDate(node.getAttribute('content'))], + ['meta[property="og:image"]', node => asDate(extractFromURL(node.getAttribute('content')))], + ['meta[itemprop="image"]', node => asDate(extractFromURL(node.getAttribute('content')))], + + ['time', node => asDate(node.getAttribute('datetime') || (node.getAttribute('class') === 'timestamp' && node.innerHTML))], + ['span[itemprop="datePublished"]', node => asDate(node.getAttribute('content') || cleanDateStr(node.innerHTML))], + ...['span', 'p', 'div'].map(tag => { + return ['pubdate', 'timestamp', 'article_date', 'articledate', 'date'].map(className => { + return [`${tag}[class="${className}"]`, node => asDate(cleanDateStr(node.innerHTML))] + }) + }).flat() + ], + scorers: [ + (el, score) => { + if (el.localName === 'script' && el.getAttribute('type') === 'application/ld+json' && el.innerHTML) { + const data = JSON.parse(el.innerHTML) + return data?.datePublished || data?.dateCreated || data?.dateModified ? 1000000 + score : 0 + } + }, + (el, score) => el.localName === 'meta' && el.getAttribute('content') && cleanDateStr(el.getAttribute('content')) ? 1000 + score : 0, + (el, score) => !['script', 'meta'].includes(el.localName) ? score : 0 + ] +}