refactor: extract lib/wiki-scraper.ts, make scraper composable, sync from Wikipedia
Move all scraping logic (fetchWikiHtml, scrapeYear, scrapeSquads and all helpers) into lib/wiki-scraper.ts as exported functions shared by both scripts. scrape-wikipedia.ts becomes a composable CLI: pnpm scrape [year] — matches + squads (default) pnpm scrape [year] --matches — matches/meta/stadiums only pnpm scrape [year] --squads — squads only sync.ts drops the openfootball GitHub dependency entirely and scrapes Wikipedia directly. Incremental: completed groups (all matches have FT scores) are detected via DB query and their sub-pages are skipped each run. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
@@ -0,0 +1,467 @@
|
||||
import { load } from 'cheerio'
|
||||
import type { CheerioAPI, Cheerio } from 'cheerio'
|
||||
import type { Element } from 'domhandler'
|
||||
|
||||
// ── Types ──────────────────────────────────────────────────────────────────
|
||||
|
||||
export type Goal = {
|
||||
name: string
|
||||
minute?: number
|
||||
offset?: number
|
||||
penalty?: boolean
|
||||
owngoal?: boolean
|
||||
}
|
||||
|
||||
export type ScoreObj = {
|
||||
ft?: [number, number]
|
||||
et?: [number, number]
|
||||
p?: [number, number]
|
||||
}
|
||||
|
||||
export type Match = {
|
||||
round: string
|
||||
group?: string
|
||||
date?: string
|
||||
time?: string
|
||||
team1: string
|
||||
team2: string
|
||||
score?: ScoreObj
|
||||
goals1?: Goal[]
|
||||
goals2?: Goal[]
|
||||
ground?: string
|
||||
}
|
||||
|
||||
export type Stadium = { name: string; city: string }
|
||||
export type Group = { name: string; teams: string[] }
|
||||
export type Meta = {
|
||||
host: string
|
||||
teams_count: number | null
|
||||
winner: string | null
|
||||
runner_up: string | null
|
||||
third_place: string | null
|
||||
fourth_place: string | null
|
||||
}
|
||||
|
||||
export type Player = { name: string; number?: number; pos?: string; date_of_birth?: string }
|
||||
export type Squad = { name: string; players: Player[] }
|
||||
|
||||
export type YearResult = {
|
||||
matches: Match[]
|
||||
stadiums: Map<string, Stadium>
|
||||
groups: Map<string, Set<string>>
|
||||
meta: Meta
|
||||
}
|
||||
|
||||
type State = { active: boolean; round: string; group: string | null }
|
||||
|
||||
// ── Fetch ──────────────────────────────────────────────────────────────────
|
||||
|
||||
const delay = (ms: number) => new Promise(r => setTimeout(r, ms))
|
||||
|
||||
export async function fetchWikiHtml(page: string, retries = 5): Promise<string | null> {
|
||||
const url = `https://en.wikipedia.org/w/api.php?action=parse&page=${encodeURIComponent(page)}&format=json&prop=text&disabletoc=1`
|
||||
for (let attempt = 0; attempt < retries; attempt++) {
|
||||
try {
|
||||
if (attempt > 0) await delay(3000 * attempt)
|
||||
const res = await fetch(url, { headers: { 'User-Agent': 'WorldCupScraper/1.0' } })
|
||||
if (!res.ok) continue
|
||||
const data = await res.json() as { parse?: { text?: { '*': string } } }
|
||||
const html = data?.parse?.text?.['*']
|
||||
if (html) return html
|
||||
} catch {
|
||||
// retry
|
||||
}
|
||||
}
|
||||
return null
|
||||
}
|
||||
|
||||
// ── Parsing helpers ────────────────────────────────────────────────────────
|
||||
|
||||
function parseScoreText(text: string): [number, number] | null {
|
||||
const m = text.match(/(\d+)\s*[–\-]\s*(\d+)/)
|
||||
if (!m) return null
|
||||
return [parseInt(m[1]), parseInt(m[2])]
|
||||
}
|
||||
|
||||
function extractTeam($: CheerioAPI, $cell: Cheerio<Element>): string {
|
||||
let name = ''
|
||||
$cell.find('a').each((_, a) => {
|
||||
const $a = $(a)
|
||||
if (!$a.find('img').length && $a.text().trim()) {
|
||||
name = $a.text().trim()
|
||||
return false
|
||||
}
|
||||
})
|
||||
return name
|
||||
}
|
||||
|
||||
function parseGoals($: CheerioAPI, $td: Cheerio<Element>): Goal[] {
|
||||
const goals: Goal[] = []
|
||||
$td.find('li').each((_, li) => {
|
||||
const $li = $(li)
|
||||
let playerName = ''
|
||||
$li.find('a').each((_, a) => {
|
||||
if (!$(a).closest('.fb-goal').length) {
|
||||
const t = $(a).text().trim()
|
||||
if (t) { playerName = t; return false }
|
||||
}
|
||||
})
|
||||
if (!playerName) return
|
||||
const $fbGoal = $li.find('.fb-goal')
|
||||
if (!$fbGoal.length) return
|
||||
$fbGoal.children('span').each((_, span) => {
|
||||
const $span = $(span)
|
||||
if ($span.attr('typeof')) return
|
||||
const text = $span.text()
|
||||
const minMatch = text.match(/(\d+)(?:\+(\d+))?['′]/)
|
||||
if (!minMatch) return
|
||||
const minute = parseInt(minMatch[1])
|
||||
const offset = minMatch[2] ? parseInt(minMatch[2]) : 0
|
||||
const goal: Goal = { name: playerName }
|
||||
if (!isNaN(minute)) goal.minute = minute
|
||||
if (offset) goal.offset = offset
|
||||
if (text.includes('pen.')) goal.penalty = true
|
||||
if (text.includes('o.g.')) goal.owngoal = true
|
||||
goals.push(goal)
|
||||
})
|
||||
})
|
||||
return goals
|
||||
}
|
||||
|
||||
function extractGround($: CheerioAPI, $box: Cheerio<Element>): string {
|
||||
const $loc = $box.find('[itemprop="name address"]').first()
|
||||
if ($loc.length) return $loc.text().trim()
|
||||
return $box.find('.fright').first().text().split('\n')[0].trim()
|
||||
}
|
||||
|
||||
function parseGroundParts(ground: string): { name: string; city: string } {
|
||||
const commaIdx = ground.indexOf(',')
|
||||
if (commaIdx !== -1) return { name: ground.slice(0, commaIdx).trim(), city: ground.slice(commaIdx + 1).trim() }
|
||||
return { name: ground, city: '' }
|
||||
}
|
||||
|
||||
function parseBox($: CheerioAPI, $box: Cheerio<Element>, round: string, group: string | null): Match | null {
|
||||
const team1 = extractTeam($, $box.find('.fhome'))
|
||||
const team2 = extractTeam($, $box.find('.faway'))
|
||||
if (!team1 || !team2) return null
|
||||
|
||||
const dateStr = $box.find('.bday, .dtstart').first().text().trim() || undefined
|
||||
const timeText = $box.find('.ftime').first().text().trim()
|
||||
const timeStr = timeText.match(/(\d{2}:\d{2})/)?.[1]
|
||||
const scoreText = $box.find('.fscore').first().text().trim()
|
||||
const hasAET = scoreText.toLowerCase().includes('a.e.t.')
|
||||
const scoreArr = parseScoreText(scoreText)
|
||||
|
||||
const $regularRow = $box.find('tr.fgoals').first()
|
||||
const goals1 = parseGoals($, $regularRow.find('.fhgoal'))
|
||||
const goals2 = parseGoals($, $regularRow.find('.fagoal'))
|
||||
|
||||
let penScore: [number, number] | undefined
|
||||
$box.find('tr').each((_, tr) => {
|
||||
const $tr = $(tr)
|
||||
if ($tr.find('th[colspan]').text().toLowerCase().includes('penalt')) {
|
||||
const ps = parseScoreText($tr.next('tr').find('th').not('.fhome,.faway').first().text().trim())
|
||||
if (ps) penScore = ps
|
||||
return false
|
||||
}
|
||||
})
|
||||
|
||||
let score: ScoreObj | undefined
|
||||
if (scoreArr) {
|
||||
if (hasAET) {
|
||||
const ftGoals = (gs: Goal[], includeOG = false) =>
|
||||
gs.filter(g => {
|
||||
const w90 = g.minute === undefined || g.minute <= 90
|
||||
return includeOG ? g.owngoal === true && w90 : !g.owngoal && w90
|
||||
}).length
|
||||
score = { ft: [ftGoals(goals1) + ftGoals(goals2, true), ftGoals(goals2) + ftGoals(goals1, true)], et: scoreArr }
|
||||
} else {
|
||||
score = { ft: scoreArr }
|
||||
}
|
||||
if (penScore) score.p = penScore
|
||||
}
|
||||
|
||||
const ground = extractGround($, $box) || undefined
|
||||
return {
|
||||
round,
|
||||
...(group ? { group } : {}),
|
||||
...(dateStr ? { date: dateStr } : {}),
|
||||
...(timeStr ? { time: timeStr } : {}),
|
||||
team1, team2,
|
||||
...(score ? { score } : {}),
|
||||
...(goals1.length ? { goals1 } : {}),
|
||||
...(goals2.length ? { goals2 } : {}),
|
||||
...(ground ? { ground } : {}),
|
||||
}
|
||||
}
|
||||
|
||||
function collectBoxes($: CheerioAPI, round: string, group: string | null): Match[] {
|
||||
const matches: Match[] = []
|
||||
$('.footballbox').each((_, el) => {
|
||||
const m = parseBox($, $(el), round, group)
|
||||
if (m) matches.push(m)
|
||||
})
|
||||
return matches
|
||||
}
|
||||
|
||||
function processHeading(text: string, level: number, state: State): void {
|
||||
const t = text.toLowerCase().trim()
|
||||
if (level === 2) {
|
||||
if (/group stage/i.test(t) && !/second/i.test(t)) {
|
||||
state.active = true; state.round = 'Group stage'; state.group = null
|
||||
} else if (/first group stage/i.test(t)) {
|
||||
state.active = true; state.round = 'Group stage'; state.group = null
|
||||
} else if (/second group stage/i.test(t)) {
|
||||
state.active = true; state.round = 'Second group stage'; state.group = null
|
||||
} else if (t === 'final round') {
|
||||
state.active = true; state.round = 'Final round'; state.group = null
|
||||
} else if (/final tournament/i.test(t)) {
|
||||
state.active = true; state.round = ''; state.group = null
|
||||
} else if (/knock.?out stage/i.test(t)) {
|
||||
state.active = true; state.round = ''; state.group = null
|
||||
} else if (/round of 16/i.test(t)) {
|
||||
state.active = true; state.round = 'Round of 16'; state.group = null
|
||||
} else if (/quarter.final/i.test(t)) {
|
||||
state.active = true; state.round = 'Quarter-finals'; state.group = null
|
||||
} else if (/semi.final/i.test(t)) {
|
||||
state.active = true; state.round = 'Semi-finals'; state.group = null
|
||||
} else if (/third.place|match for third|play.off for third/i.test(t)) {
|
||||
state.active = true; state.round = 'Third-place match'; state.group = null
|
||||
} else if (t === 'final') {
|
||||
state.active = true; state.round = 'Final'; state.group = null
|
||||
} else {
|
||||
state.active = false
|
||||
}
|
||||
return
|
||||
}
|
||||
if (!state.active) return
|
||||
if (level === 3 || level === 4) {
|
||||
if (/^group [a-h1-9]+$/i.test(t)) {
|
||||
state.group = text.trim()
|
||||
} else if (/round of 32/i.test(t)) {
|
||||
state.round = 'Round of 32'; state.group = null
|
||||
} else if (/round of 16/i.test(t)) {
|
||||
state.round = 'Round of 16'; state.group = null
|
||||
} else if (/quarter.final/i.test(t)) {
|
||||
state.round = 'Quarter-finals'; state.group = null
|
||||
} else if (/semi.final/i.test(t)) {
|
||||
state.round = 'Semi-finals'; state.group = null
|
||||
} else if (/third.place|match for third|play.off for third/i.test(t)) {
|
||||
state.round = 'Third-place match'; state.group = null
|
||||
} else if (t === 'final') {
|
||||
state.round = 'Final'; state.group = null
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// ── Infobox ────────────────────────────────────────────────────────────────
|
||||
|
||||
function parseInfobox($: CheerioAPI): Partial<Meta> {
|
||||
const result: Partial<Meta> = {}
|
||||
|
||||
function tdText($td: Cheerio<Element>): string {
|
||||
const $clone = $td.clone()
|
||||
$clone.find('br').replaceWith(' / ')
|
||||
$clone.find('sup, img').remove()
|
||||
return $clone.text().replace(/\[\d+\]/g, '').replace(/\s+/g, ' ').trim()
|
||||
}
|
||||
function tdFirstLink($td: Cheerio<Element>): string | null {
|
||||
let name: string | null = null
|
||||
$td.find('a').each((_, a) => {
|
||||
const t = $(a).clone().find('img').remove().end().text().trim()
|
||||
if (t && !/\[\d+\]/.test(t)) { name = t; return false }
|
||||
})
|
||||
return name ?? (tdText($td) || null)
|
||||
}
|
||||
function tdAllLinks($td: Cheerio<Element>): string {
|
||||
const names: string[] = []
|
||||
$td.find('a').each((_, a) => {
|
||||
const t = $(a).clone().find('img').remove().end().text().trim()
|
||||
if (t && !/\[\d+\]/.test(t)) names.push(t)
|
||||
})
|
||||
return names.length ? names.join(' / ') : tdText($td)
|
||||
}
|
||||
|
||||
$('table.infobox').first().find('tr').each((_, tr) => {
|
||||
const $tr = $(tr)
|
||||
const label = $tr.find('th').text().trim().toLowerCase()
|
||||
const $td = $tr.find('td').first()
|
||||
if (!$td.length) return
|
||||
if (/host countr/i.test(label)) result.host = tdAllLinks($td)
|
||||
else if (/^teams$/i.test(label)) { const m = $td.text().match(/\d+/); if (m) result.teams_count = parseInt(m[0]) }
|
||||
else if (/champion/i.test(label)) result.winner = tdFirstLink($td)
|
||||
else if (/runners?.up/i.test(label)) result.runner_up = tdFirstLink($td)
|
||||
else if (/third.place/i.test(label)) result.third_place = tdFirstLink($td)
|
||||
else if (/fourth.place/i.test(label)) result.fourth_place = tdFirstLink($td)
|
||||
})
|
||||
|
||||
return result
|
||||
}
|
||||
|
||||
function derivePlacements(matches: Match[]): Pick<Meta, 'winner' | 'runner_up' | 'third_place' | 'fourth_place'> {
|
||||
function matchWinner(m: Match): [string, string] | null {
|
||||
if (!m.score) return null
|
||||
const [h, a] = m.score.et ?? m.score.ft ?? [0, 0]
|
||||
if (h > a) return [m.team1, m.team2]
|
||||
if (a > h) return [m.team2, m.team1]
|
||||
if (m.score.p) {
|
||||
const [ph, pa] = m.score.p
|
||||
if (ph > pa) return [m.team1, m.team2]
|
||||
if (pa > ph) return [m.team2, m.team1]
|
||||
}
|
||||
return null
|
||||
}
|
||||
let winner: string | null = null, runner_up: string | null = null
|
||||
let third_place: string | null = null, fourth_place: string | null = null
|
||||
for (const m of matches) {
|
||||
if (m.round === 'Final') {
|
||||
const r = matchWinner(m); if (r) [winner, runner_up] = r
|
||||
} else if (m.round === 'Third-place match') {
|
||||
const r = matchWinner(m); if (r) [third_place, fourth_place] = r
|
||||
}
|
||||
}
|
||||
return { winner, runner_up, third_place, fourth_place }
|
||||
}
|
||||
|
||||
// ── Main year scraper ──────────────────────────────────────────────────────
|
||||
|
||||
export async function scrapeYear(
|
||||
year: number,
|
||||
mainHtml: string,
|
||||
opts?: { skipGroups?: Set<string> },
|
||||
): Promise<YearResult> {
|
||||
const $ = load(mainHtml)
|
||||
const matches: Match[] = []
|
||||
const stadiums = new Map<string, Stadium>()
|
||||
const groups = new Map<string, Set<string>>()
|
||||
const state: State = { active: false, round: '', group: null }
|
||||
const groupSubpages = new Map<string, string>()
|
||||
const groupsOnMainPage = new Set<string>()
|
||||
|
||||
function recordMatch(m: Match) {
|
||||
matches.push(m)
|
||||
if (m.group) groupsOnMainPage.add(m.group)
|
||||
if (m.ground) {
|
||||
const { name, city } = parseGroundParts(m.ground)
|
||||
if (name && !stadiums.has(name)) stadiums.set(name, { name, city })
|
||||
}
|
||||
if (m.group) {
|
||||
if (!groups.has(m.group)) groups.set(m.group, new Set())
|
||||
groups.get(m.group)!.add(m.team1)
|
||||
groups.get(m.group)!.add(m.team2)
|
||||
}
|
||||
}
|
||||
|
||||
$('.mw-parser-output').find('div.mw-heading, .footballbox, .hatnote').each((_, el) => {
|
||||
const $el = $(el)
|
||||
if ($el.hasClass('mw-heading')) {
|
||||
const $h = $el.find('h2, h3, h4').first()
|
||||
const level = parseInt($h.prop('tagName')?.slice(1) ?? '9')
|
||||
processHeading($h.text().replace(/\[edit\]/g, '').trim(), level, state)
|
||||
} else if ($el.hasClass('hatnote') && $el.text().includes('Main article')) {
|
||||
if (state.active && state.group) {
|
||||
const link = $el.find('a[href^="/wiki/"]').first().attr('href')
|
||||
if (link) {
|
||||
const page = link.replace('/wiki/', '').split('#')[0]
|
||||
if (/World_Cup_Group/i.test(page) && !groupSubpages.has(state.group))
|
||||
groupSubpages.set(state.group, page)
|
||||
}
|
||||
}
|
||||
} else if ($el.hasClass('footballbox')) {
|
||||
if (!state.active) return
|
||||
const m = parseBox($, $el, state.round || state.group || 'Unknown', state.group)
|
||||
if (m) recordMatch(m)
|
||||
}
|
||||
})
|
||||
|
||||
for (const [group, page] of groupSubpages) {
|
||||
if (groupsOnMainPage.has(group)) continue
|
||||
if (opts?.skipGroups?.has(group)) {
|
||||
process.stdout.write(`[skip ${group}] `)
|
||||
continue
|
||||
}
|
||||
await delay(1200)
|
||||
const subHtml = await fetchWikiHtml(page)
|
||||
if (!subHtml) { process.stdout.write(`(failed: ${page}) `); continue }
|
||||
|
||||
const stateTemp: State = { active: false, round: '', group: null }
|
||||
let round = 'Group stage'
|
||||
$('.mw-parser-output').find('div.mw-heading').each((_, el) => {
|
||||
const $h = $(el).find('h2, h3, h4').first()
|
||||
const level = parseInt($h.prop('tagName')?.slice(1) ?? '9')
|
||||
processHeading($h.text().replace(/\[edit\]/g, '').trim(), level, stateTemp)
|
||||
if (stateTemp.group === group) { round = stateTemp.round || 'Group stage'; return false }
|
||||
})
|
||||
|
||||
const $sub = load(subHtml)
|
||||
for (const m of collectBoxes($sub, round, group)) recordMatch(m)
|
||||
process.stdout.write(`[+${page.slice(-8)}] `)
|
||||
}
|
||||
|
||||
const infobox = parseInfobox($)
|
||||
const placements = derivePlacements(matches)
|
||||
const meta: Meta = {
|
||||
host: infobox.host ?? '',
|
||||
teams_count: infobox.teams_count ?? null,
|
||||
winner: placements.winner ?? infobox.winner ?? null,
|
||||
runner_up: placements.runner_up ?? infobox.runner_up ?? null,
|
||||
third_place: placements.third_place ?? infobox.third_place ?? null,
|
||||
fourth_place: placements.fourth_place?? infobox.fourth_place?? null,
|
||||
}
|
||||
|
||||
return { matches, stadiums, groups, meta }
|
||||
}
|
||||
|
||||
// ── Squad page scraper ─────────────────────────────────────────────────────
|
||||
|
||||
export function scrapeSquads(html: string): Squad[] {
|
||||
const $ = load(html)
|
||||
const squads: Squad[] = []
|
||||
let currentTeam: Squad | null = null
|
||||
|
||||
$('.mw-parser-output').find('div.mw-heading, tr.nat-fs-player').each((_, el) => {
|
||||
const $el = $(el)
|
||||
|
||||
if ($el.hasClass('mw-heading')) {
|
||||
const $h = $el.find('h3, h4').first()
|
||||
if (!$h.length) return
|
||||
if (parseInt($h.prop('tagName')?.slice(1) ?? '9') !== 3) return
|
||||
const name = $h.text().replace(/\[edit\]/g, '').trim()
|
||||
if (/^group /i.test(name)) return
|
||||
currentTeam = { name, players: [] }
|
||||
squads.push(currentTeam)
|
||||
return
|
||||
}
|
||||
|
||||
if (!currentTeam) return
|
||||
|
||||
let number: number | undefined
|
||||
let pos: string | undefined
|
||||
let playerName = ''
|
||||
let dob: string | undefined
|
||||
|
||||
$el.find('td, th[scope="row"]').each((i, td) => {
|
||||
const $td = $(td)
|
||||
const text = $td.text().trim()
|
||||
if ($td.is('th[scope="row"]')) {
|
||||
playerName = $td.find('a').first().text().trim() || text
|
||||
} else if (i === 0 && !playerName) {
|
||||
const n = parseInt(text); if (!isNaN(n)) number = n
|
||||
} else if (i === 1 && !playerName && !pos) {
|
||||
const p = $td.find('a').first().text().trim()
|
||||
if (['GK', 'DF', 'MF', 'FW'].includes(p)) pos = p
|
||||
}
|
||||
const $bday = $td.find('.bday')
|
||||
if ($bday.length) dob = $bday.text().trim()
|
||||
})
|
||||
|
||||
if (!playerName) return
|
||||
const player: Player = { name: playerName }
|
||||
if (number !== undefined) player.number = number
|
||||
if (pos) player.pos = pos
|
||||
if (dob) player.date_of_birth = dob
|
||||
currentTeam.players.push(player)
|
||||
})
|
||||
|
||||
return squads.filter(s => s.players.length > 0)
|
||||
}
|
||||
Reference in New Issue
Block a user