tehran_taj/apps/api/scripts/legacy-import/lib/context.ts

/**
 * Shared state for one import run: both connections, the id map and the report.
 */
import { copyFileSync, existsSync, mkdirSync, statSync } from 'node:fs';
import { join } from 'node:path';
import { PrismaClient } from '@prisma/client';
import { createPool, type Pool, type RowDataPacket } from 'mysql2/promise';

export interface Options {
  sourceUrl: string;
  targetUrl: string;
  /** Legacy system's `storage/` folder holding the image files (optional). */
  imagesDir?: string;
  /** Target UPLOAD_DIR the files are copied into (required with imagesDir). */
  uploadDir?: string;
  /** Run only these steps (by name); all when unset. */
  only?: Set<string>;
  reportPath: string;
}

/* ───────────── report ───────────── */

interface Section {
  counts: Map<string, number>;
  notes: string[];
}

/**
 * What the run did, step by step. Counts say how many rows went which way;
 * notes are the things a person must look at. Rendered to markdown at the end.
 */
export class Report {
  readonly startedAt = new Date();
  private readonly sections = new Map<string, Section>();

  private section(step: string): Section {
    let s = this.sections.get(step);
    if (!s) {
      s = { counts: new Map(), notes: [] };
      this.sections.set(step, s);
    }
    return s;
  }

  count(step: string, key: string, n = 1) {
    const s = this.section(step);
    s.counts.set(key, (s.counts.get(key) ?? 0) + n);
  }

  note(step: string, message: string) {
    this.section(step).notes.push(message);
  }

  get(step: string, key: string): number {
    return this.sections.get(step)?.counts.get(key) ?? 0;
  }

  render(header: string[]): string {
    const out: string[] = ['# legacy (oapi) → tehran-taj import report', '', ...header, ''];
    for (const [step, s] of this.sections) {
      out.push(`## ${step}`, '');
      if (s.counts.size) {
        out.push('| | |', '|---|---:|');
        for (const [k, v] of s.counts) out.push(`| ${k} | ${v} |`);
        out.push('');
      }
      for (const n of s.notes) out.push(`- ${n}`);
      if (s.notes.length) out.push('');
    }
    return out.join('\n');
  }
}

/* ───────────── id map ───────────── */

/**
 * Legacy id -> target id, per entity, persisted in the TARGET database under
 * its own schema `legacy_import` — outside the `public` schema Prisma manages,
 * so it is no app-schema change and drops cleanly after cutover
 * (`DROP SCHEMA legacy_import CASCADE`).
 *
 * It is what makes a re-run safe (a mapped row is updated, never re-created)
 * and what later redirects old links such as /product/123 (GO-LIVE-PLAN §7).
 * `how` records whether the row was created by the import or matched to a row
 * that already existed in the target (merge, §10.5).
 */
export type How = 'created' | 'matched' | 'merged';

export class IdMap {
  private readonly cache = new Map<string, Map<string, { target: string; how: How }>>();

  constructor(private readonly db: PrismaClient) {}

  static async ensureSchema(db: PrismaClient) {
    await db.$executeRawUnsafe(`CREATE SCHEMA IF NOT EXISTS legacy_import`);
    await db.$executeRawUnsafe(`
      CREATE TABLE IF NOT EXISTS legacy_import.id_map (
        entity    text        NOT NULL,
        legacy_id text        NOT NULL,
        target_id text        NOT NULL,
        how       text        NOT NULL,
        mapped_at timestamptz NOT NULL DEFAULT now(),
        PRIMARY KEY (entity, legacy_id)
      )`);
  }

  async load(entity: string) {
    const rows = await this.db.$queryRawUnsafe<{ legacy_id: string; target_id: string; how: How }[]>(
      `SELECT legacy_id, target_id, how FROM legacy_import.id_map WHERE entity = $1`,
      entity,
    );
    this.cache.set(entity, new Map(rows.map((r) => [r.legacy_id, { target: r.target_id, how: r.how }])));
  }

  get(entity: string, legacyId: unknown): string | undefined {
    if (legacyId === null || legacyId === undefined) return undefined;
    return this.cache.get(entity)?.get(String(legacyId))?.target;
  }

  how(entity: string, legacyId: unknown): How | undefined {
    return this.cache.get(entity)?.get(String(legacyId))?.how;
  }

  async set(entity: string, legacyId: unknown, targetId: string, how: How) {
    const key = String(legacyId);
    await this.db.$executeRawUnsafe(
      `INSERT INTO legacy_import.id_map (entity, legacy_id, target_id, how)
       VALUES ($1, $2, $3, $4)
       ON CONFLICT (entity, legacy_id) DO UPDATE SET target_id = EXCLUDED.target_id, how = EXCLUDED.how`,
      entity, key, targetId, how,
    );
    let m = this.cache.get(entity);
    if (!m) {
      m = new Map();
      this.cache.set(entity, m);
    }
    m.set(key, { target: targetId, how });
  }

  async drop(entity: string, legacyId: unknown) {
    await this.db.$executeRawUnsafe(
      `DELETE FROM legacy_import.id_map WHERE entity = $1 AND legacy_id = $2`,
      entity, String(legacyId),
    );
    this.cache.get(entity)?.delete(String(legacyId));
  }
}

/* ───────────── upsert discipline ───────────── */

/**
 * The one way a step writes a row (§10.5: this is a MERGE into a live target):
 *
 *  - already mapped        -> `update(id, how)`; the step decides what it may touch
 *  - matches a target row  -> mapped as 'matched', then `update(id, 'matched')`
 *  - otherwise             -> `create()`, mapped as 'created'
 *
 * A mapping whose target row has since been deleted is dropped and redone, so a
 * re-run heals instead of failing on a dangling id.
 */
export async function upsert(
  ctx: Ctx,
  entity: string,
  legacyId: unknown,
  ops: {
    exists: (id: string) => Promise<boolean>;
    find: () => Promise<{ id: string } | null>;
    create: () => Promise<{ id: string }>;
    update: (id: string, how: How) => Promise<void>;
  },
): Promise<{ id: string; how: How; isNew: boolean }> {
  const mapped = ctx.map.get(entity, legacyId);
  if (mapped) {
    if (await ops.exists(mapped)) {
      const how = ctx.map.how(entity, legacyId) ?? 'created';
      await ops.update(mapped, how);
      return { id: mapped, how, isNew: false };
    }
    await ctx.map.drop(entity, legacyId);
  }
  const found = await ops.find();
  if (found) {
    await ctx.map.set(entity, legacyId, found.id, 'matched');
    await ops.update(found.id, 'matched');
    return { id: found.id, how: 'matched', isNew: false };
  }
  const created = await ops.create();
  await ctx.map.set(entity, legacyId, created.id, 'created');
  return { id: created.id, how: 'created', isNew: true };
}

/**
 * For a row that already existed in the target (`matched`): only the fields
 * the target left empty are filled. The target's own values win — the owner
 * confirmed its data is real (§10.5).
 */
export function fillEmpty<T extends Record<string, unknown>>(current: Record<string, unknown>, patch: T): Partial<T> {
  const out: Partial<T> = {};
  for (const [k, v] of Object.entries(patch)) {
    const cur = current[k];
    if ((cur === null || cur === undefined || cur === '') && v !== null && v !== undefined) {
      (out as Record<string, unknown>)[k] = v;
    }
  }
  return out;
}

/* ───────────── context ───────────── */

export interface Ctx {
  src: Pool;
  db: PrismaClient;
  map: IdMap;
  report: Report;
  opt: Options;
  /** A read-only query against the legacy MySQL. */
  q<T = Record<string, unknown>>(sql: string, params?: unknown[]): Promise<T[]>;
  /** Public URL for a legacy `image` row, copying its file when a source dir is given. */
  imageUrl(imageId: unknown, step: string): Promise<string | null>;
  /**
   * An `UploadedFile` row for a legacy image — what avatars point at
   * (Customer/AdminUser.avatarFileId are FKs, not URLs). Mapped as 'file'.
   */
  uploadedFile(imageId: unknown, purpose: string, step: string): Promise<string | null>;
}

const MIME: Record<string, string> = {
  png: 'image/png', jpg: 'image/jpeg', jpeg: 'image/jpeg', gif: 'image/gif',
  webp: 'image/webp', svg: 'image/svg+xml', pdf: 'application/pdf',
};

export function openSource(url: string): Pool {
  return createPool({
    uri: url,
    connectionLimit: 4,
    charset: 'utf8mb4',
    // Timestamps come back as the wall-clock text legacy stored, and are
    // converted from Tehran time explicitly (normalize.tehran). The session
    // time zone is left at the server's own, which is what Laravel and the
    // Navicat dump both used.
    dateStrings: true,
    supportBigNumbers: true,
    bigNumberStrings: true,
  });
}

export function makeContext(opt: Options, src: Pool, db: PrismaClient): Ctx {
  const map = new IdMap(db);
  const report = new Report();
  const imageRows = new Map<string, string | null>();

  const ctx: Ctx = {
    src,
    db,
    map,
    report,
    opt,
    async q<T>(sql: string, params: unknown[] = []) {
      const [rows] = await src.query<RowDataPacket[]>(sql, params);
      return rows as unknown as T[];
    },
    async imageUrl(imageId: unknown, step: string) {
      if (imageId === null || imageId === undefined || imageId === '') return null;
      const key = String(imageId);
      if (!imageRows.has(key)) {
        const [row] = await ctx.q<{ image: string | null }>('SELECT image FROM image WHERE id = ?', [key]);
        imageRows.set(key, row?.image ? String(row.image).trim() : null);
      }
      const file = imageRows.get(key);
      if (!file) {
        report.count(step, 'image reference with no `image` row');
        return null;
      }
      copyImage(ctx, file, step);
      // Same shape the admin upload writes (admin-uploads.service.ts:61).
      return `/uploads/${file}`;
    },
    async uploadedFile(imageId: unknown, purpose: string, step: string) {
      if (imageId === null || imageId === undefined || imageId === '') return null;
      // Resolve first even when mapped: resolving copies the file (see productImages).
      const url = await ctx.imageUrl(imageId, step);
      const mapped = map.get('file', imageId);
      if (mapped) return mapped;
      if (!url) return null;
      const [row] = await ctx.q<{ original_name: string | null; ext: string | null }>(
        'SELECT original_name, ext FROM image WHERE id = ?', [String(imageId)],
      );
      const file = url.slice('/uploads/'.length);
      const ext = (row?.ext ?? file.split('.').pop() ?? '').toLowerCase().trim();
      const src = ctx.opt.imagesDir ? join(ctx.opt.imagesDir, file) : null;
      const created = await db.uploadedFile.create({
        data: {
          url,
          originalName: (row?.original_name ?? '').trim() || file,
          mimeType: MIME[ext] ?? 'application/octet-stream',
          sizeBytes: src && existsSync(src) ? statSync(src).size : 0,
          purpose,
        },
      });
      await map.set('file', imageId, created.id, 'created');
      return created.id;
    },
  };
  return ctx;
}

const copied = new Set<string>();

function copyImage(ctx: Ctx, file: string, step: string) {
  const { imagesDir, uploadDir } = ctx.opt;
  if (!imagesDir || !uploadDir || copied.has(file)) return;
  copied.add(file);
  const from = join(imagesDir, file);
  const to = join(uploadDir, file);
  if (existsSync(to)) {
    ctx.report.count('images', 'already in upload dir');
    return;
  }
  if (!existsSync(from) || !statSync(from).isFile()) {
    ctx.report.count('images', 'missing in legacy storage');
    ctx.report.note('images', `missing file ${file} (referenced from ${step})`);
    return;
  }
  mkdirSync(uploadDir, { recursive: true });
  copyFileSync(from, to);
  ctx.report.count('images', 'copied');
}