import { createHash } from 'node:crypto'
import { copyFile, readFile, writeFile } from 'node:fs/promises'
import { basename, dirname, resolve } from 'node:path'
import process from 'node:process'
import { fileURLToPath } from 'node:url'
import pdfParse from 'pdf-parse/lib/pdf-parse.js'

const here = dirname(fileURLToPath(import.meta.url))
const outputDir = process.env.PDF_STUDY_OUTPUT_DIR
  ? resolve(process.env.PDF_STUDY_OUTPUT_DIR)
  : resolve(here, '../../public/research/pdf-extraction')
const cases = [
  {
    id: 'singleColumn',
    file: 'synthetic-single-column.pdf',
    layout: 'single-column text layer',
    expected: [
      'single-contact',
      'SUMMARY',
      'single-summary',
      'EXPERIENCE',
      'single-role',
      'single-impact',
      'EDUCATION',
      'single-school',
      'SKILLS',
      'single-skills',
    ],
  },
  {
    id: 'twoColumn',
    file: 'synthetic-two-column.pdf',
    layout: 'two-column text layer',
    expected: [
      'LEFT-SUMMARY',
      'left-focus',
      'LEFT-EXPERIENCE',
      'left-role',
      'left-result',
      'RIGHT-SKILLS',
      'right-skill SQL',
      'right-skill Forecasting',
      'RIGHT-EDUCATION',
      'right-school',
    ],
  },
  {
    id: 'korean',
    file: 'synthetic-korean.pdf',
    layout: 'Korean native text layer',
    expected: [
      '한글-이름',
      '한글-연락처',
      '경력',
      '한글-직무',
      '한글-성과',
      '기술',
      '한글-기술',
    ],
  },
  {
    id: 'mixedLanguage',
    file: 'synthetic-mixed-language.pdf',
    layout: 'Korean and English native text layer',
    expected: [
      '혼합-이름',
      'mixed-contact',
      'EXPERIENCE 경력',
      'mixed-role',
      '혼합-성과',
      'SKILLS 기술',
      'mixed-skills',
    ],
  },
  {
    id: 'tableLinks',
    file: 'synthetic-table-links.pdf',
    layout: 'table with visible URL and labeled link',
    expected: [
      'table-header-year',
      'table-header-role',
      'table-year-2024',
      'table-role Research lead',
      'table-year-2022',
      'table-role Analyst',
      'link-visible',
      'https://example.invalid/portfolio',
      'link-label',
    ],
    expectedLinks: [
      'https://example.invalid/portfolio',
      'https://example.invalid/case-study',
    ],
  },
  {
    id: 'imageOnly',
    file: 'synthetic-scanned-image.pdf',
    layout: 'image-only scan negative control',
    expected: [
      'scan-contact',
      'EXPERIENCE',
      'scan-role',
      'scan-result',
      'SKILLS',
      'scan-skills',
    ],
  },
]

const results = []
for (const testCase of cases) {
  const path = resolve(outputDir, testCase.file)
  const bytes = await readFile(path)
  const parsed = await pdfParse({ data: new Uint8Array(bytes) })
  const text = parsed.text.replace(/\r/g, '').trim()
  const found = testCase.expected.filter((token) => text.includes(token))
  const observedOrder = testCase.expected
    .map((token) => ({ token, index: text.indexOf(token) }))
    .filter(({ index }) => index >= 0)
    .sort((a, b) => a.index - b.index)
    .map(({ token }) => token)
  const expectedFoundOrder = testCase.expected.filter((token) =>
    found.includes(token)
  )
  results.push({
    id: testCase.id,
    file: testCase.file,
    layout: testCase.layout,
    bytes: bytes.length,
    sha256: createHash('sha256').update(bytes).digest('hex'),
    pages: parsed.numpages,
    extractedCharacters: text.length,
    expectedAnchorCount: testCase.expected.length,
    foundAnchorCount: found.length,
    missingAnchors: testCase.expected.filter((token) => !found.includes(token)),
    expectedAnchorOrder: testCase.expected,
    observedAnchorOrder: observedOrder,
    anchorOrderMatches:
      found.length === testCase.expected.length
        ? JSON.stringify(observedOrder) === JSON.stringify(expectedFoundOrder)
        : null,
    expectedLinkTargets: testCase.expectedLinks ?? [],
    extractedText: text,
  })
}

const report = {
  schemaVersion: 2,
  generatedAt: new Date().toISOString(),
  method: {
    parser: 'pdf-parse 1.1.1 default text extraction',
    fixtureTools: 'ReportLab 4.4.9 and Pillow 12.3.0',
    scope:
      'Native PDF text layer only; no OCR, ATS test, or resume field inference.',
    linkScope:
      'Visible link text is measured. Link annotation destinations are present in the fixture but are not evaluated by pdf-parse.',
    productionDifference:
      'This controlled parser test does not exercise or make claims about the refresh.cv production importer.',
    baseline: '/research/pdf-extraction/native-text-results-v1.json',
    reproductionFiles: [
      '/research/pdf-extraction/generate-fixtures.py.txt',
      '/research/pdf-extraction/measure-native-text.mjs.txt',
    ],
  },
  results,
}
await writeFile(
  resolve(outputDir, 'native-text-results.json'),
  `${JSON.stringify(report, null, 2)}\n`,
  'utf8'
)
await copyFile(
  fileURLToPath(import.meta.url),
  resolve(outputDir, 'measure-native-text.mjs.txt')
)
for (const result of results)
  console.log(
    `${basename(result.file)}: ${result.foundAnchorCount}/${result.expectedAnchorCount} anchors, ${result.extractedCharacters} chars, order=${result.anchorOrderMatches}`
  )