/**
 * Conservative operator inference from a facility name. OpenStreetMap / Wikidata features often carry no
 * `operator` tag while the name starts with a canonical brand ("Equinix AM3", "Google Data Center, Elmina",
 * "Meta Odense Data Center"). Only the curated canonical table (names + aliases) is used, matched as a whole-token
 * prefix of the name — never fuzzy, never inside the name. Provenance method: `inferred:name-prefix`.
 */
import { normalizeName } from "@dci/core";
import { CANONICAL_OPERATORS, type CanonicalOperator } from "./canonical-operators.js";

export const OPERATOR_INFERENCE_METHOD = "inferred:name-prefix";

/**
 * Brand words that are also ordinary words or too short to stand alone ("Switch", "Meta", "Apple", "IO", "Spark"):
 * accepted only when the rest of the name looks like a facility (data-center vocabulary or a facility code).
 */
/** Ambiguous brands where data-center vocabulary in the raw name is enough ("Meta Odense Data Center", "Apple Data Center Viborg"). */
const AMBIGUOUS_VOCAB: ReadonlySet<string> = new Set(["meta", "apple", "switch", "aligned", "vantage", "compass", "lambda", "verne", "moro"]);
/** Ambiguous brands that additionally need a facility code ("Vantage WA11", "Switch Las Vegas 8"). */
const AMBIGUOUS_CODE: ReadonlySet<string> = new Set(["stack", "stc", "esr", "kio", "gds", "cdc", "adc", "att", "mci", "cht", "telx", "skt", "ctg", "hkt", "pdg", "gdh", "ngd", "dft", "dlr", "oci", "gcp", "aws", "ovh", "ntt", "amt", "apld", "bce", "yti", "ytl", "udcs", "jio", "sify", "inap"]);
/** Ordinary words: never inferred from a single token whatever follows ("Green Data Center Norway", "Orange County"). */
const NEVER_ALONE: ReadonlySet<string> = new Set(["green", "constant", "beyond", "stream", "bulk", "chief", "spark", "orange", "bell", "tract", "packet", "io", "tm", "kt", "bt", "tel", "tata", "reliance", "iliad", "scala", "elea", "echelon", "rowan", "takoda", "novva", "revera", "sabey", "ark", "virtus", "cirion", "nxtra", "khazna", "linode", "lumen"]);
const PLACE_WORDS = /^(county|parish|township|city|town|village|valley|street|st|road|rd|avenue|ave|park|lake|river|island|bay|beach|hills?|mountain|falls|creek)$/;
const FACILITY_HINT_RE = /\b(data ?cent(er|re)s?|datacent(er|re)s?|rechenzentrum|centre de donn[ée]es|centro de datos|campus|colocation|colo|dc|idc|hub|pop|site)\b/i;
const CODE_HINT_RE = /(^|[^A-Za-z0-9])[A-Z]{1,6}-?\d{1,3}[A-Z]?([^A-Za-z0-9]|$)|\b[A-Z][a-z]+ \d{1,2}\b/;

interface BrandEntry { tokens: string[]; op: CanonicalOperator; canonical: boolean }
let brands: BrandEntry[] | null = null;

function table(): BrandEntry[] {
  if (brands) return brands;
  const out: BrandEntry[] = [];
  for (const op of CANONICAL_OPERATORS) {
    const seen = new Set<string>();
    for (const [i, raw] of [op.name, ...op.aliases].entries()) {
      const tokens = normalizeName(raw).split(" ").filter(Boolean);
      const key = tokens.join(" ");
      if (!tokens.length || seen.has(key)) continue;
      // a single token shorter than 3 characters ("IO", "TM") is never a safe prefix
      if (tokens.length === 1 && tokens[0]!.length < 3) continue;
      seen.add(key);
      out.push({ tokens, op, canonical: i === 0 });
    }
  }
  // longest brand first so "Digital Realty" wins over "Digital"-prefixed shorter aliases
  out.sort((a, b) => b.tokens.length - a.tokens.length || b.tokens.join(" ").length - a.tokens.join(" ").length);
  brands = out;
  return out;
}

export interface InferredOperator { name: string; matched: string; method: string }

/**
 * Operator implied by the name's leading tokens, or null. Rules:
 * - the brand tokens must be the first tokens of the normalized name (noise words like "data center" already removed);
 * - the name must carry something beyond the brand (a site / code / vocabulary) unless it equals the canonical name exactly;
 * - ambiguous brands require data-center vocabulary or a facility code in the raw name.
 */
export function inferOperatorFromName(name: string | null | undefined): InferredOperator | null {
  if (!name) return null;
  const tokens = normalizeName(name).split(" ").filter(Boolean);
  if (!tokens.length) return null;
  for (const b of table()) {
    if (b.tokens.length > tokens.length) continue;
    if (!b.tokens.every((t, i) => tokens[i] === t)) continue;
    const rest = tokens.slice(b.tokens.length);
    const brandKey = b.tokens.join(" ");
    const nameOnly = rest.length === 0;
    if (nameOnly && !b.canonical) continue; // "AWS" alone is not a facility name
    if (b.tokens.length === 1) {
      if (NEVER_ALONE.has(brandKey)) continue;
      if (AMBIGUOUS_CODE.has(brandKey) && !CODE_HINT_RE.test(name)) continue;
      if (AMBIGUOUS_VOCAB.has(brandKey) && !(FACILITY_HINT_RE.test(name) || CODE_HINT_RE.test(name))) continue;
    }
    // the rest must look like a site, not a place name or a different business ("Orange County", "Apple Store")
    if (rest.length && (PLACE_WORDS.test(rest[0]!) || /^(store|shop|office|offices|headquarters|hq|fiber|retail|bank|insurance)\b/.test(rest.join(" ")))) continue;
    return { name: b.op.name, matched: brandKey, method: OPERATOR_INFERENCE_METHOD };
  }
  return null;
}