worldcup/scripts/scrape-wikipedia.ts

import { load } from 'cheerio'
import type { CheerioAPI } from 'cheerio'
import type { Cheerio } from 'cheerio'
import type { Element } from 'domhandler'
import { mkdirSync, writeFileSync } from 'fs'
import path from 'path'
import { fileURLToPath } from 'url'

const __dirname = path.dirname(fileURLToPath(import.meta.url))
const DATA_DIR = path.join(__dirname, '../app/data/wikipedia')

const YEARS = [
  1930,1934,1938,1950,1954,1958,1962,1966,1970,1974,
  1978,1982,1986,1990,1994,1998,2002,2006,2010,2014,2018,2022,
]

const delay = (ms: number) => new Promise(r => setTimeout(r, ms))

// ── Types ──────────────────────────────────────────────────────────────────

type Goal = {
  name: string
  minute?: number
  offset?: number
  penalty?: boolean
  owngoal?: boolean
}

type ScoreObj = {
  ft?: [number, number]
  et?: [number, number]
  p?: [number, number]
}

type Match = {
  round: string
  group?: string
  date?: string
  time?: string
  team1: string
  team2: string
  score?: ScoreObj
  goals1?: Goal[]
  goals2?: Goal[]
  ground?: string
}

type Stadium = { name: string; city: string }
type Player = { name: string; number?: number; pos?: string; date_of_birth?: string }
type Squad = { name: string; players: Player[] }
type Group = { name: string; teams: string[] }

// ── Fetch ──────────────────────────────────────────────────────────────────

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 (github.com/worldcup)' } })
      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
}

// ── Score parsing ──────────────────────────────────────────────────────────

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])]
}

// ── Team name extraction ───────────────────────────────────────────────────

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
}

// ── Goal parsing ───────────────────────────────────────────────────────────

function parseGoals($: CheerioAPI, $td: Cheerio<Element>): Goal[] {
  const goals: Goal[] = []

  $td.find('li').each((_, li) => {
    const $li = $(li)

    // Player name: first <a> NOT inside .fb-goal
    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

    // Each direct child <span> inside .fb-goal (excluding image wrapper)
    $fbGoal.children('span').each((_, span) => {
      const $span = $(span)
      if ($span.attr('typeof')) return // image wrapper

      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 isPen = text.includes('pen.')
      const isOG = text.includes('o.g.')

      const goal: Goal = { name: playerName }
      if (!isNaN(minute)) goal.minute = minute
      if (offset) goal.offset = offset
      if (isPen) goal.penalty = true
      if (isOG) goal.owngoal = true
      goals.push(goal)
    })
  })

  return goals
}

// ── Ground extraction ──────────────────────────────────────────────────────

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: '' }
}

// ── Footballbox parsing ────────────────────────────────────────────────────

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 timeMatch = timeText.match(/(\d{2}:\d{2})/)
  const timeStr = timeMatch?.[1]

  const scoreText = $box.find('.fscore').first().text().trim()
  const hasAET = scoreText.toLowerCase().includes('a.e.t.')
  const scoreArr = parseScoreText(scoreText)

  // Use first fgoals row only (exclude penalty shootout row)
  const $regularRow = $box.find('tr.fgoals').first()
  const goals1 = parseGoals($, $regularRow.find('.fhgoal'))
  const goals2 = parseGoals($, $regularRow.find('.fagoal'))

  // Penalty shootout score: row after "Penalties" header tr
  let penScore: [number, number] | undefined
  $box.find('tr').each((_, tr) => {
    const $tr = $(tr)
    if ($tr.find('th[colspan]').text().toLowerCase().includes('penalt')) {
      const penText = $tr.next('tr').find('th').not('.fhome,.faway').first().text().trim()
      const ps = parseScoreText(penText)
      if (ps) penScore = ps
      return false
    }
  })

  let score: ScoreObj | undefined
  if (scoreArr) {
    if (hasAET) {
      // scoreArr is ET total; compute FT from goals in ≤90 min
      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
      const ftHome = ftGoals(goals1) + ftGoals(goals2, true)
      const ftAway = ftGoals(goals2) + ftGoals(goals1, true)
      score = { ft: [ftHome, ftAway], 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 } : {}),
  }
}

// ── Collect matches from a pre-loaded page ─────────────────────────────────

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
}

// ── Section heading state machine ──────────────────────────────────────────

type State = {
  active: boolean
  round: string
  group: string | null
}

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
    }
    // bracket, draw, seeding, replay → keep current state
  }
}

// ── Main year scraper ──────────────────────────────────────────────────────

type YearResult = {
  matches: Match[]
  stadiums: Map<string, Stadium>
  groups: Map<string, Set<string>>
}

async function scrapeYear(year: number, mainHtml: 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 }

  // Maps group name → sub-page to fetch (if main page has no matches for that group)
  const groupSubpages = new Map<string, string>()
  // Groups that got at least one match from the main page
  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)
    }
  }

  // Walk elements in document order: headings, hatnotes, footballboxes
  $('.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')
      const text = $h.text().replace(/\[edit\]/g, '').trim()
      processHeading(text, level, state)

    } else if ($el.hasClass('hatnote') && $el.text().includes('Main article')) {
      // Record sub-page link for current group context (for fallback if no main-page matches)
      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 ($el.hasClass('footballbox')) {
      if (!state.active) return
      const round = state.round || state.group || 'Unknown'
      const m = parseBox($, $el, round, state.group)
      if (m) recordMatch(m)
    }
  })

  // Fetch group sub-pages for any group that got 0 matches from main page
  for (const [group, page] of groupSubpages) {
    if (groupsOnMainPage.has(group)) continue

    await delay(1200)
    const subHtml = await fetchWikiHtml(page)
    if (!subHtml) { process.stdout.write(`(failed: ${page}) `); continue }

    // Determine the round for this group from the state machine result
    // (we'll reconstruct from the main-page walk state — use the round that was active when this group was seen)
    // Since we can't easily recover state here, we re-walk to find the round for this group
    let round = 'Group stage'
    let foundGroup = false
    const stateTemp: State = { active: false, round: '', group: null }
    $('.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')
      const text = $h.text().replace(/\[edit\]/g, '').trim()
      processHeading(text, level, stateTemp)
      if (stateTemp.group === group) {
        round = stateTemp.round || 'Group stage'
        foundGroup = true
        return false
      }
    })

    const $sub = load(subHtml)
    const subMatches = collectBoxes($sub, round || 'Group stage', group)
    for (const m of subMatches) {
      recordMatch(m)
    }
    process.stdout.write(`[+${page.slice(-8)}] `)
  }

  return { matches, stadiums, groups }
}

// ── Squad page scraper ─────────────────────────────────────────────────────

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
      const level = parseInt($h.prop('tagName')?.slice(1) ?? '9')
      if (level !== 3) return
      const name = $h.text().replace(/\[edit\]/g, '').trim()
      if (/^group /i.test(name)) return // skip group headers
      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 posLink = $td.find('a').first().text().trim()
        if (['GK', 'DF', 'MF', 'FW'].includes(posLink)) pos = posLink
      }

      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
}

// ── Output ─────────────────────────────────────────────────────────────────

function writeOutput(
  year: number,
  matches: Match[],
  stadiums: Map<string, Stadium>,
  groups: Map<string, Set<string>>,
  squads: Squad[],
): void {
  const dir = path.join(DATA_DIR, String(year))
  mkdirSync(dir, { recursive: true })

  writeFileSync(
    path.join(dir, 'worldcup.json'),
    JSON.stringify({ matches }, null, 2),
    'utf-8',
  )

  if (stadiums.size > 0) {
    writeFileSync(
      path.join(dir, 'worldcup.stadiums.json'),
      JSON.stringify({ stadiums: Array.from(stadiums.values()) }, null, 2),
      'utf-8',
    )
  }

  const groupList: Group[] = []
  groups.forEach((teams, name) => {
    groupList.push({ name, teams: Array.from(teams) })
  })
  if (groupList.length > 0) {
    writeFileSync(
      path.join(dir, 'worldcup.groups.json'),
      JSON.stringify({ groups: groupList }, null, 2),
      'utf-8',
    )
  }

  if (squads.length > 0) {
    writeFileSync(
      path.join(dir, 'worldcup.squads.json'),
      JSON.stringify(squads, null, 2),
      'utf-8',
    )
  }
}

// ── Entry point ────────────────────────────────────────────────────────────

async function main() {
  const onlyYear = process.argv[2] ? parseInt(process.argv[2]) : null
  const yearsToScrape = onlyYear ? [onlyYear] : YEARS

  console.log(`Scraping ${yearsToScrape.length} World Cup(s) from Wikipedia...`)

  for (const year of yearsToScrape) {
    process.stdout.write(`  ${year}... `)

    const mainHtml = await fetchWikiHtml(`${year}_FIFA_World_Cup`)
    if (!mainHtml) { console.log('FAILED'); continue }

    const { matches, stadiums, groups } = await scrapeYear(year, mainHtml)

    await delay(600)

    const squadHtml = await fetchWikiHtml(`${year}_FIFA_World_Cup_squads`)
    const squads = squadHtml ? scrapeSquads(squadHtml) : []

    writeOutput(year, matches, stadiums, groups, squads)

    console.log(`${matches.length} matches, ${stadiums.size} stadiums, ${groups.size} groups, ${squads.length} teams`)

    await delay(600)
  }

  console.log('\nDone! Files written to app/data/wikipedia/{year}/')
}

main().catch(e => { console.error(e); process.exit(1) })