""" 查册文件检测器 对抗性测试套件 测试 python/company_particulars_detector.py 的各种场景 """ import json import sys import os import io import tempfile # Add project python dir to path sys.path.insert(0, os.path.join(os.path.dirname(os.path.dirname(__file__)), 'python')) from company_particulars_detector import CompanyParticularsDetector import fitz from PIL import Image PASS = 0 FAIL = 0 detector = CompanyParticularsDetector() def test(name, condition_fn): """Run a test case, print PASS/FAIL""" global PASS, FAIL try: condition_fn() print(f' [PASS] {name}') PASS += 1 except AssertionError as e: print(f' [FAIL] {name}: {e}') FAIL += 1 except Exception as e: print(f' [ERROR] {name}: {type(e).__name__}: {e}') FAIL += 1 # CJK font path for Chinese text tests (PyMuPDF default font doesn't support CJK) _CJK_FONT_PATH = None for _fp in ['C:/Windows/Fonts/simsun.ttc', 'C:/Windows/Fonts/msyh.ttc']: if os.path.exists(_fp): _CJK_FONT_PATH = _fp break def make_text_pdf(pages_content, use_cjk_font=False): """ Create a temp PDF. pages_content is a list where each element is either: - str: text to insert on the page - list: list of (x, y, text, fontsize) tuples Returns temp file path. """ fd, path = tempfile.mkstemp(suffix='.pdf') os.close(fd) doc = fitz.open() font_kwargs = {} if use_cjk_font and _CJK_FONT_PATH: font_kwargs = {'fontname': 'CJK', 'fontfile': _CJK_FONT_PATH} for content in pages_content: page = doc.new_page() if isinstance(content, str): page.insert_text((50, 50), content, fontsize=12, **font_kwargs) elif isinstance(content, list): for item in content: x, y, text = item[0], item[1], item[2] fs = item[3] if len(item) > 3 else 12 page.insert_text((x, y), text, fontsize=fs, **font_kwargs) else: # Empty page pass doc.save(path) doc.close() return path def make_image_page(): """Create a page with only an embedded image, no text""" img = Image.new('RGB', (100, 100), color='white') buf = io.BytesIO() img.save(buf, format='PNG') return buf.getvalue() def make_image_pdf(num_pages=2): """Create a PDF with only image pages (no text)""" fd, path = tempfile.mkstemp(suffix='.pdf') os.close(fd) doc = fitz.open() for _ in range(num_pages): page = doc.new_page() img_data = make_image_page() rect = fitz.Rect(0, 0, 100, 100) page.insert_image(rect, stream=img_data) doc.save(path) doc.close() return path def cleanup(*paths): for p in paths: try: if os.path.exists(p): os.unlink(p) except Exception: pass # ============================================================ print('=' * 60) print('查册文件检测器 对抗性测试套件') print('=' * 60) # ------------------------------------------------------- # Test 1: User's scanned PDF # ------------------------------------------------------- def t1(): r = detector.is_company_particulars('查册文件-CompanyParticulars.pdf') assert r['is_scanned'] == True, f'Expected scanned=True, got {r["is_scanned"]}' assert r['text_length'] < 50, f'Expected text_length<50, got {r["text_length"]}' assert r.get('image_only_pages', 0) > 0, f'Expected image_only_pages>0, got {r.get("image_only_pages")}' assert 'image_only_pages' in r, 'Missing image_only_pages key in result' test('扫描版查册文件 (user PDF): is_scanned=true, low text, has image pages', t1) # ------------------------------------------------------- # Test 2: Non-existent file # ------------------------------------------------------- def t2(): r = detector.is_company_particulars('__nonexistent_file__.pdf') assert r['is_company_particulars'] == False assert '不存在' in r.get('error', '') test('不存在的文件: error contains 文件不存在', t2) # ------------------------------------------------------- # Test 3: Text-based 查册文件 with keywords # ------------------------------------------------------- def t3(): path = make_text_pdf([[ (50, 50, 'Company Particulars Report', 14), (50, 100, 'Company Number: 12345678', 12), (50, 140, 'Registered Office Address: Room 123, Building A, HK', 12), (50, 180, 'Date of Incorporation: 2020-01-01', 12), (50, 220, 'Directors: John Doe, Jane Smith', 12), (50, 260, 'Share Capital: HKD 10,000', 12), (50, 300, 'Company Secretary: ABC Secretarial Ltd', 12), (50, 350, 'Padding text to ensure >50 chars total for proper detection flow', 10), ]]) try: r = detector.is_company_particulars(path) assert r['is_company_particulars'] == True, f'Expected True, got {r}' assert r['confidence'] >= 0.6, f'Confidence too low: {r["confidence"]}' assert len(r['matched_keywords']) >= 2, f'Too few matched: {r["matched_keywords"]}' finally: cleanup(path) test('文本型查册文件 (含 Company Particulars 等关键词): detected=true', t3) # ------------------------------------------------------- # Test 4: Non-查册文件 (regular text PDF) # ------------------------------------------------------- def t4(): path = make_text_pdf([[ (50, 50, 'Invoice Report', 14), (50, 100, 'This is a regular invoice document for payment', 12), (50, 140, 'Total amount: $1000.00 due in 30 days', 12), (50, 180, 'Customer: ABC Corp, Address: 123 Main St', 12), (50, 220, 'Extra padding to make sure we have more than 50 chars', 10), ]]) try: r = detector.is_company_particulars(path) assert r['is_company_particulars'] == False, f'Expected False, got {r}' assert r['is_scanned'] == False, f'Should not be scanned, got {r["is_scanned"]}' finally: cleanup(path) test('非查册文件 (普通发票文本): detected=false, not scanned', t4) # ------------------------------------------------------- # Test 5: Fully scanned PDF (all pages image-only) # ------------------------------------------------------- def t5(): path = make_image_pdf(2) try: r = detector.is_company_particulars(path) assert r['is_scanned'] == True, f'Expected scanned=True, got {r["is_scanned"]}' assert r.get('image_only_pages', 0) == 2, f'Expected 2 image-only pages, got {r.get("image_only_pages")}' assert r['text_length'] == 0, f'Expected 0 text, got {r["text_length"]}' finally: cleanup(path) test('全扫描PDF (2页纯图片无文本): is_scanned=true, image_only_pages=2', t5) # ------------------------------------------------------- # Test 6: Edge case - 49 chars of noise text # ------------------------------------------------------- def t6(): path = make_text_pdf(['ABCD EFGH IJKL MNOP QRST UVWX YZab cdef ghij klmn']) # ~49 chars try: r = detector.is_company_particulars(path) # Should be flagged as scanned (text too short) assert r['is_scanned'] == True, f'49 chars should be scanned, got {r["is_scanned"]}' finally: cleanup(path) test('边界值: 49字符噪音文本 -> is_scanned=true', t6) # ------------------------------------------------------- # Test 7: Edge case - 50 chars of text, no keywords # ------------------------------------------------------- def t7(): # exactly 50 chars of noise path = make_text_pdf(['ABCD EFGH IJKL MNOP QRST UVWX YZab cdef ghij klmn op']) try: r = detector.is_company_particulars(path) assert r['is_scanned'] == False, f'50 chars should NOT be scanned, got is_scanned={r["is_scanned"]}' assert r['is_company_particulars'] == False, 'No keywords, should be false' finally: cleanup(path) test('边界值: 50字符噪音文本 -> not scanned, not company particulars', t7) # ------------------------------------------------------- # Test 8: Mixed PDF (page1 image-only, page2 text with keywords) # ------------------------------------------------------- def t8(): fd, path = tempfile.mkstemp(suffix='.pdf') os.close(fd) doc = fitz.open() # Page 1: image only (封面) page1 = doc.new_page() img_data = make_image_page() rect = fitz.Rect(0, 0, 100, 100) page1.insert_image(rect, stream=img_data) # Page 2: text with 查册文件 keywords page2 = doc.new_page() page2.insert_text((50, 50), 'Company Particulars', fontsize=14) page2.insert_text((50, 100), 'Company Number: 99999999', fontsize=12) page2.insert_text((50, 150), 'Registered Office: Kowloon Tower, Flat B', fontsize=12) page2.insert_text((50, 200), 'Padding text to exceed fifty characters total across pages', fontsize=10) doc.save(path) doc.close() try: r = detector.is_company_particulars(path) # IMPORTANT: should NOT be flagged as scanned despite page1 being image-only # because page2 has extractable text with keywords assert r['is_scanned'] == False, f'FAIL: mixed PDF should NOT be scanned (page2 has text). Got is_scanned={r["is_scanned"]}' assert r['is_company_particulars'] == True, f'FAIL: should detect keywords on page2. Got {r["is_company_particulars"]}' assert 'Company Particulars' in r.get('matched_keywords', []), f'Expected Company Particulars in matched: {r.get("matched_keywords")}' finally: cleanup(path) test('混合PDF (封面图+文本正文): NOT scanned, keywords detected on page2', t8) # ------------------------------------------------------- # Test 9: Exclude keywords (CR证书) # ------------------------------------------------------- def t9(): path = make_text_pdf([[ (50, 50, 'Certificate of Incorporation', 14), (50, 100, 'Business Registration Certificate', 12), (50, 150, 'Company Number: 11111111', 12), (50, 200, 'This is a certificate of incorporation document for a Hong Kong company', 10), (50, 250, 'Extra text padding to exceed fifty characters to avoid the scanned detection', 10), ]]) try: r = detector.is_company_particulars(path) assert r['is_company_particulars'] == False, f'CR should be excluded, got {r["is_company_particulars"]}' assert r.get('error'), 'Should have exclusion error message' finally: cleanup(path) test('排除关键词: CR+BR 证书被正确排除', t9) # ------------------------------------------------------- # Test 10: CLI JSON output format # ------------------------------------------------------- def t10(): import subprocess result = subprocess.run( ['python', 'python/company_particulars_detector.py', '查册文件-CompanyParticulars.pdf'], capture_output=True, text=True, cwd=os.path.dirname(os.path.dirname(__file__)) ) output = json.loads(result.stdout) required_keys = [ 'is_company_particulars', 'matched_keywords', 'confidence', 'error', 'text_length', 'is_scanned', 'language', 'image_only_pages' ] for k in required_keys: assert k in output, f'Missing key "{k}" in CLI output' # Verify the critical field for PHP fallback assert output['is_scanned'] == True, 'CLI: scanned PDF should have is_scanned=true' assert output['image_only_pages'] >= 1, 'CLI: should have image_only_pages >= 1' test(f'CLI模式JSON输出: 所有 {8} 个必要字段存在, is_scanned=true', t10) # ------------------------------------------------------- # Test 11: Single keyword "公司资料" (Chinese) # ------------------------------------------------------- def t11(): path = make_text_pdf([[ (50, 50, '公司资料', 14), (50, 100, '公司编号: 88888888', 12), (50, 150, '注册办事处地址: 香港中环皇后大道1号', 12), (50, 200, 'Additional text to ensure the total character count exceeds fifty for proper non-scanned detection', 10), ]], use_cjk_font=True) try: r = detector.is_company_particulars(path) assert r['is_company_particulars'] == True, f'Chinese 公司资料 should match, got {r["is_company_particulars"]}' assert r['confidence'] >= 0.9, f'Confidence should be high for primary keyword: {r["confidence"]}' finally: cleanup(path) test('中文查册文件 (含"公司资料"主关键词): detected=true, high confidence', t11) # ------------------------------------------------------- # Test 12: 2 secondary keywords but no primary (>=3 needed for secondary-only) # ------------------------------------------------------- def t12(): path = make_text_pdf([[ (50, 50, 'Company Number: 77777777', 12), (50, 100, 'Date of Incorporation: 2019-06-15', 12), (50, 150, 'Some unrelated business document text for filler content to exceed the fifty character minimum threshold', 10), ]]) try: r = detector.is_company_particulars(path) # total_score = 2 (Company Number + Date of Incorporation) # has_primary_keyword = False # total_score < 3, so not detected (3-keyword threshold not met) # But total_score >= 2 and exclude_score == 0 → confidence 0.6 assert r['is_company_particulars'] == True, f'2 keywords + no exclude -> should be true (0.6 conf). Got {r}' assert r['confidence'] == 0.6, f'Expected 0.6 confidence, got {r["confidence"]}' finally: cleanup(path) test('2个次要关键词 (无主关键词, 无排除词): detected=true, confidence=0.6', t12) # ------------------------------------------------------- # Test 13: Empty PDF (no pages) # ------------------------------------------------------- def t13(): fd, path = tempfile.mkstemp(suffix='.pdf') os.close(fd) # Create a single empty page (PyMuPDF requires at least 1 page to save) doc = fitz.open() doc.new_page() doc.save(path) doc.close() try: r = detector.is_company_particulars(path) assert r['is_company_particulars'] == False # Empty page has 0 text -> should be flagged as scanned assert r['is_scanned'] == True, f'Empty page should be scanned: {r}' finally: cleanup(path) test('空PDF (单空白页): is_scanned=true (no extractable text)', t13) # ------------------------------------------------------- # Test 14: Defense-in-depth: PHP fallback fields always present # ------------------------------------------------------- def t14(): """Verify all fields that PHP uses for OCR fallback are always present""" # Test with various inputs cases = [ ('查册文件-CompanyParticulars.pdf', 'scanned user PDF'), ] for pdf_path, desc in cases: r = detector.is_company_particulars(pdf_path) for field in ['is_company_particulars', 'is_scanned', 'text_length', 'image_only_pages']: assert field in r, f'{desc}: missing field "{field}" for PHP fallback' test('PHP fallback 字段完整性: 所有必要字段始终存在', t14) # ============================================================ print('\n' + '=' * 60) print(f'RESULTS: {PASS} passed, {FAIL} failed, {PASS + FAIL} total') if FAIL > 0: print('*** SOME TESTS FAILED - DO NOT DEPLOY ***') sys.exit(1) else: print('All tests passed - safe to deploy') sys.exit(0)