// Best-effort citekey generator for papers the user dropped into
// ~/Documents/Thesis that did NOT come through the Zotero/ZotLit pipeline.
// Matches the Zotero "firstAuthorTitleYear" style used by howl's vault
// (e.g. `bergPredictionMarketsDecision2003`). Deterministic + slug-safe.

import { basename } from 'node:path'

// Trim common subtitle separators so the citekey stays short.
const SUBTITLE_SEP = /[:—–-]\s+/

// Stop words we drop from the title when camel-casing.
const STOP = new Set([
  'a', 'an', 'and', 'as', 'at', 'but', 'by', 'for', 'from', 'in', 'into', 'is',
  'of', 'on', 'or', 'the', 'to', 'with', 'without', 'via', 'using', 'about',
  'over', 'per', 'are', 'be', 'been', 'this', 'that',
])

// Zotero sometimes names exports like "Author and Author - 2003 - Title.pdf"
// or "Author Title Year.pdf" — the regex below catches both.
const FILENAME_PARTS = /^(?<auth>[A-Za-z][\w.'’\s&,-]+?)\s*(?:-|—|–|_)\s*(?<year>\d{4})\s*(?:-|—|–|_)\s*(?<title>.+)$/

function firstAuthorFromAuth(auth: string): string | null {
  // Split on " and ", ",", or "&". First chunk → last word before punctuation.
  const first = auth.split(/\s+and\s+|,|&/)[0]?.trim()
  if (!first) return null
  // "Berg, J." → "Berg"; "Berg J. E." → "Berg"; "Smith John" → "Smith".
  const tokens = first.split(/\s+/).filter(t => /[A-Za-z]/.test(t))
  if (tokens.length === 0) return null
  const last = tokens[0]!.replace(/[^A-Za-z]/g, '')
  return last.length > 0 ? last.toLowerCase() : null
}

function camelTitle(title: string, maxChunks = 5): string {
  const primary = title.split(SUBTITLE_SEP)[0] ?? title
  const tokens = primary
    .replace(/\.pdf$/i, '')
    .replace(/[^\w\s]/g, ' ')
    .split(/\s+/)
    .filter(t => t.length > 0)
    .filter(t => !STOP.has(t.toLowerCase()))
    .slice(0, maxChunks)
  return tokens
    .map(t => t.charAt(0).toUpperCase() + t.slice(1).toLowerCase())
    .join('')
}

export function inferCitekey(filename: string, firstPageText?: string): string | null {
  const base = basename(filename).replace(/\.[^.]+$/, '')
  const m = FILENAME_PARTS.exec(base)
  let author: string | null = null
  let year: string | null = null
  let title: string | null = null

  if (m && m.groups) {
    author = firstAuthorFromAuth(m.groups.auth!)
    year = m.groups.year!
    title = m.groups.title!
  }

  if (!year && firstPageText) {
    const y = /(19|20)\d{2}/.exec(firstPageText)
    if (y) year = y[0]
  }
  if (!author && firstPageText) {
    const top = firstPageText.slice(0, 400)
    const by = /by\s+([A-Z][A-Za-z-]+)/.exec(top)
    const cap = /([A-Z][A-Za-z-]{2,})\s+(?:et\s+al|and)/.exec(top)
    author = by?.[1]?.toLowerCase() ?? cap?.[1]?.toLowerCase() ?? null
  }
  if (!title) title = base

  if (!author || !year) return null
  const camel = camelTitle(title)
  if (!camel) return null
  return `${author}${camel}${year}`
}

export function literaturePathFor(citekey: string): string {
  return `04_Notes/41_Literature/${citekey}.md`
}