src/lib/qc/mec-parser.ts

/**
 * Amazon MEC (Media Entertainment Core) XML metadata parser.
 * Parses mdmec:CoreMetadata XML and extracts text fields as QC lines
 * so they can be fed through the AI QC pipeline for spelling/grammar checks.
 */

import type { QCLine } from "./types";

interface MECField {
  fieldName: string;
  language: string;
  value: string;
}

/**
 * Detect if a buffer is an Amazon MEC XML file.
 */
export function isMECFile(buffer: Buffer, filename: string): boolean {
  if (!filename.match(/\.(xml|mec)$/i)) return false;
  const head = buffer.subarray(0, Math.min(500, buffer.length)).toString("utf-8");
  return head.includes("mdmec:CoreMetadata") || head.includes("movielabs.com/schema/mdmec");
}

/**
 * Parse MEC XML and extract all text fields as QCLines.
 * Each field becomes a QCLine with the field path as context.
 */
export function parseMECToLines(buffer: Buffer, targetLanguage?: string): {
  lines: QCLine[];
  metadata: MECMetadata;
} {
  const xml = buffer.toString("utf-8");
  const fields = extractMECFields(xml);
  const metadata = extractMECMetadata(xml);

  // Filter by language if specified
  const filtered = targetLanguage
    ? fields.filter(f => {
        // Match language prefix (e.g. "en" matches "en-US", "fil" matches "fil-PH")
        const langPrefix = targetLanguage.toLowerCase().split("-")[0];
        const fieldLangPrefix = f.language.toLowerCase().split("-")[0];
        return fieldLangPrefix === langPrefix || f.language === "" || f.language === "all";
      })
    : fields;

  const lines: QCLine[] = filtered.map((f, i) => ({
    id: i + 1,
    text: f.value,
    tcIn: undefined,
    tcOut: undefined,
    // Store field context in a way the QC prompts can use
    _fieldName: f.fieldName,
    _language: f.language,
  } as QCLine & { _fieldName: string; _language: string }));

  return { lines, metadata };
}

export interface MECMetadata {
  contentId: string;
  title: string;
  originalTitle: string;
  workType: string;
  releaseYear: string;
  releaseDate: string;
  languages: string[];
  country: string;
  genres: string[];
  castCount: number;
  crewCount: number;
}

function extractMECMetadata(xml: string): MECMetadata {
  const get = (tag: string): string => {
    const m = xml.match(new RegExp(`<[^>]*${tag}[^>]*>([^<]*)<`));
    return m ? m[1].trim() : "";
  };
  const getAttr = (tag: string, attr: string): string => {
    const m = xml.match(new RegExp(`<[^>]*${tag}[^>]*${attr}="([^"]*)"`));
    return m ? m[1].trim() : "";
  };

  const languages = new Set<string>();
  const langMatches = xml.matchAll(/language="([^"]+)"/g);
  for (const m of langMatches) languages.add(m[1]);

  const genresSet = new Set<string>();
  const genreMatches = xml.matchAll(/Genre\s+id="([^"]+)"/g);
  for (const m of genreMatches) genresSet.add(m[1]);

  const castCount = (xml.match(/Actor<\/md:JobFunction>/g) || []).length;
  const crewCount = (xml.match(/<md:JobFunction>(?!Actor)/g) || []).length;

  return {
    contentId: getAttr("Basic", "ContentID"),
    title: get("TitleDisplayUnlimited"),
    originalTitle: get("OriginalTitle"),
    workType: get("WorkType"),
    releaseYear: get("ReleaseYear"),
    releaseDate: get("ReleaseDate"),
    languages: [...languages],
    country: get("country"),
    genres: [...genresSet],
    castCount: castCount / 2, // dedupe fil-PH + en-US
    crewCount: crewCount / 2,
  };
}

function extractMECFields(xml: string): MECField[] {
  const fields: MECField[] = [];

  // Extract all LocalizedInfo blocks
  const localizedBlocks = xml.matchAll(
    /<md:LocalizedInfo\s+language="([^"]*)"[^>]*>([\s\S]*?)<\/md:LocalizedInfo>/g
  );

  for (const block of localizedBlocks) {
    const lang = block[1];
    const content = block[2];

    // Title
    const title = content.match(/<md:TitleDisplayUnlimited>([^<]*)<\/md:TitleDisplayUnlimited>/);
    if (title?.[1]?.trim()) {
      fields.push({ fieldName: `Title (${lang})`, language: lang, value: title[1].trim() });
    }

    // Original Title
    const origTitle = content.match(/<md:OriginalTitle>([^<]*)<\/md:OriginalTitle>/);
    if (origTitle?.[1]?.trim()) {
      fields.push({ fieldName: `OriginalTitle (${lang})`, language: lang, value: origTitle[1].trim() });
    }

    // Summary190
    const s190 = content.match(/<md:Summary190>([^<]*)<\/md:Summary190>/);
    if (s190?.[1]?.trim()) {
      fields.push({ fieldName: `Summary190 (${lang})`, language: lang, value: s190[1].trim() });
    }

    // Summary400
    const s400 = content.match(/<md:Summary400>([\s\S]*?)<\/md:Summary400>/);
    if (s400?.[1]?.trim()) {
      fields.push({ fieldName: `Summary400 (${lang})`, language: lang, value: s400[1].trim() });
    }

    // Summary4000
    const s4000 = content.match(/<md:Summary4000>([\s\S]*?)<\/md:Summary4000>/);
    if (s4000?.[1]?.trim()) {
      fields.push({ fieldName: `Summary4000 (${lang})`, language: lang, value: s4000[1].trim() });
    }
  }

  // Extract People names + characters
  const people = xml.matchAll(
    /<md:People>([\s\S]*?)<\/md:People>/g
  );

  const seenPeople = new Set<string>();
  for (const p of people) {
    const content = p[1];
    const job = content.match(/<md:JobFunction>([^<]*)<\/md:JobFunction>/)?.[1]?.trim();
    const name = content.match(/<md:DisplayName\s+language="([^"]*)"[^>]*>([^<]*)<\/md:DisplayName>/);
    const character = content.match(/<md:Character>([^<]*)<\/md:Character>/)?.[1]?.trim();

    if (name?.[2]?.trim()) {
      const lang = name[1];
      const nameVal = name[2].trim();
      const key = `${job}:${nameVal}:${lang}`;
      if (!seenPeople.has(key)) {
        seenPeople.add(key);
        const label = character
          ? `${job}: ${nameVal} as "${character}" (${lang})`
          : `${job}: ${nameVal} (${lang})`;
        fields.push({ fieldName: label, language: lang, value: nameVal });
        if (character) {
          fields.push({ fieldName: `Character name (${lang})`, language: lang, value: character });
        }
      }
    }
  }

  // CompanyDisplayCredit
  const company = xml.match(/<md:DisplayString>([^<]*)<\/md:DisplayString>/);
  if (company?.[1]?.trim()) {
    fields.push({ fieldName: "CompanyDisplayCredit", language: "all", value: company[1].trim() });
  }

  return fields;
}

/**
 * Format MEC lines for QC prompt — includes field context.
 */
export function formatMECLinesForPrompt(lines: (QCLine & { _fieldName?: string; _language?: string })[]): string {
  return lines.map(l => {
    const field = l._fieldName || `Field ${l.id}`;
    return `[LINE ${l.id}] [${field}] ${l.text}`;
  }).join("\n");
}