UnisKB/apps/system_manage/test_heading_structure.py

153 lines
8.5 KiB
Python
Raw Blame History

This file contains ambiguous Unicode characters!

This file contains ambiguous Unicode characters that may be confused with others in your current locale. If your use case is intentional and legitimate, you can safely ignore this warning. Use the Escape button to highlight these characters.

import copy
import json
import unittest
from types import SimpleNamespace
from system_manage.heading_structure import analyze, manual_edit, model_repair, protect_confirmed, validate_manual
class HeadingRulesTests(unittest.TestCase):
def test_flat_numbered_outline_preserves_body_media_tables_and_code(self):
body = '\nBody 42\n\n![figure](/image.png)\n\n| A | B |\n|---|---|\n| 1 | 2 |\n\n```\n# 1.9 Code\n```\n'
text = '# 1 Root\n# 1.1 Section\n# 1.1.1 Item\n' + body + '# 2 End\n'
result, report = analyze(text, {'engine': 'mineru'})
self.assertIn('## 1.1 Section', result)
self.assertIn('### 1.1.1 Item', result)
self.assertIn(body, result)
self.assertEqual([r['level'] for r in report['headings']], [1, 2, 3, 1])
def test_original_markdown_and_html_levels_remain_authoritative(self):
text = '# 1 Root\n#### 1.1 Deliberate\n'
for fmt in ('md', 'html', 'docx'):
self.assertEqual(analyze(text, {'format': fmt})[0], text)
def test_fourth_to_sixth_decimal_levels_are_not_mistaken_for_ip_addresses(self):
numbers = ['1', '1.1', '1.1.1', '1.1.1.1', '1.1.1.1.1', '1.1.1.1.1.1']
text = '\n'.join('# ' + number + ' Section' for number in numbers)
_, report = analyze(text)
self.assertEqual([r['level'] for r in report['headings']], [1, 2, 3, 4, 5, 6])
def test_unstructured_chinese_enumeration_is_not_promoted_without_context(self):
text = '一、选项甲\n二、选项乙\n'
self.assertEqual(analyze(text)[0], text)
def test_identical_titles_keep_distinct_manual_locks_after_reordering(self):
text, report = analyze('# Root\n## Same\nfirst body\n## Same\nsecond body\n# Other\n', {'format': 'md'})
second = [r for r in report['headings'] if r['text'] == 'Same'][1]
target = report['headings'][-1]
output, moved = manual_edit(text, report, {'id': second['id'], 'kind': 'parent', 'parent_id': target['id']})
_, restored = analyze(output, previous=report, repair=False)
validate_manual(restored, moved)
self.assertIn(second['id'], moved['manual'])
def test_branch_moving_does_not_merge_an_unterminated_final_line(self):
text, report = analyze('# A\nbody A\n# B\nbody B', {'format': 'md'})
output, _ = manual_edit(text, report, {'id': report['headings'][0]['id'], 'kind': 'parent', 'parent_id': report['headings'][1]['id']})
self.assertIn('body B\n## A\n', output)
def test_dates_ips_versions_and_ordinary_lists_are_not_promoted(self):
text = '2026 Annual report\n\n192.168.1.1 Gateway\n\n1.0 version\n\n1. First task\n2. Second task\n\n- 1.1 List detail\n'
result, report = analyze(text)
self.assertEqual(result, text)
self.assertFalse(any(r['level'] for r in report['headings']))
def test_mixed_chinese_numbering_is_relative_to_observed_schemes(self):
text = '# 第一章 总则\n# 一、目标\n# (一)范围\n# 1 条件\n# (1)细则\n'
result, report = analyze(text)
self.assertEqual([r['level'] for r in report['headings']], [1, 2, 3, 4, 5])
def test_bookmark_requires_matching_title_and_page(self):
source = {'bookmarks': [{'text': 'Section', 'page': 1, 'level': 3}],
'blocks': [{'text': 'Section', 'page': 1, 'kind': 'title'}]}
result, report = analyze('# Section\n', source)
self.assertEqual(result, '### Section\n')
self.assertEqual(report['headings'][0]['source'], 'pdf_bookmark')
source['bookmarks'][0]['page'] = 2
self.assertEqual(analyze('# Section\n', source)[0], '# Section\n')
def test_layout_uses_document_wide_size_groups(self):
source = {'blocks': [{'text': 'Root', 'kind': 'title', 'font_size': 24},
{'text': 'Child', 'kind': 'title', 'font_size': 18}, {'text': 'Body.', 'kind': 'text', 'font_size': 10}]}
result, report = analyze('# Root\n# Child\nBody.\n', source)
self.assertEqual([r['level'] for r in report['headings']], [1, 2])
self.assertIn('## Child', result)
def test_repeated_margin_lines_are_hidden_without_deleting_source(self):
source = {'blocks': [{'text': 'Repeated header', 'page': i, 'bbox': [0, 1, 100, 9], 'page_height': 800, 'kind': 'text'} for i in range(3)]}
text = 'Repeated header\n\nRepeated header\n\nRepeated header\n'
output, report = analyze(text, source)
self.assertEqual(output, text)
self.assertEqual(report['excluded_lines'], [0, 2, 4])
def test_toc_requires_directory_context_and_does_not_hide_real_body_title(self):
text = '# Contents\nSection A .... 1\nSection B .... 3\n\n# Section A\nBody\n'
output, report = analyze(text)
self.assertEqual(output, text)
self.assertEqual(report['excluded_lines'], [1, 2])
self.assertEqual(analyze('Conclusion .... 42\n')[1]['excluded_lines'], [])
def test_manual_branch_move_preserves_text_and_locks_descendants(self):
text, report = analyze('# A\n## B\n### C\nbody42\n# D\n', {'format': 'md'})
by_text = {h['text']: h for h in report['headings']}
output, report = manual_edit(text, report, {'id': by_text['B']['id'], 'kind': 'parent', 'parent_id': by_text['D']['id']})
self.assertEqual(output, '# A\n# D\n## B\n### C\nbody42\n')
again, fresh = analyze(output, previous=report)
self.assertEqual(again, output)
self.assertEqual(len(fresh['manual']), 2)
self.assertEqual(next(h for h in fresh['headings'] if h['text'] == 'B')['id'], by_text['B']['id'])
def test_cycles_and_subtree_depth_over_six_are_rejected(self):
text, report = analyze('# A\n## B\n###### C\n', {'format': 'md'})
with self.assertRaises(ValueError):
manual_edit(text, report, {'id': report['headings'][0]['id'], 'kind': 'parent', 'parent_id': report['headings'][1]['id']})
with self.assertRaises(ValueError):
manual_edit(text, report, {'id': report['headings'][1]['id'], 'kind': 'demote'})
def test_confirmed_outline_survives_rules_and_model_repair(self):
text, report = analyze('# Root\n## Child\n', {'format': 'md'})
text, report = manual_edit(text, report, {'id': report['headings'][1]['id'], 'kind': 'confirm'})
out, fresh = model_repair(text, {}, report, lambda instruction, payload: json.dumps([
{'id': r['id'], 'level': 1} for r in json.loads(payload)['batch']]))
self.assertIn('## Child', out)
self.assertTrue(next(h for h in fresh['headings'] if h['text'] == 'Child')['confirmed'])
def test_model_only_returns_ids_and_levels_and_never_rewrites_body(self):
text = '# Root\n# Child\n# Detail\nBody 42\n\n![image](/image.png)\n'
def generate(instruction, payload):
data = json.loads(payload)
self.assertNotIn('Body 42', payload)
return json.dumps([{'id': item['id'], 'level': i + 1} for i, item in enumerate(data['batch'])])
result, report = model_repair(text, {}, {}, generate)
self.assertEqual(result, '# Root\n## Child\n### Detail\nBody 42\n\n![image](/image.png)\n')
self.assertEqual(report['model']['status'], 'applied')
def test_bad_or_failed_model_responses_fall_back_atomically(self):
text = '# Root\n# Child\nBody42\n'
responses = ['[]', '[{"id":"invented","level":2}]', 'not json']
for response in responses:
out, report = model_repair(text, {}, {}, lambda *args: response)
self.assertEqual(out, text)
self.assertEqual(report['model']['status'], 'fallback')
def failure(*args):
raise RuntimeError('secret-provider-details')
out, report = model_repair(text, {}, {}, failure)
self.assertEqual(out, text)
self.assertNotIn('secret-provider-details', json.dumps(report))
def test_wrong_manual_ids_and_modified_titles_are_not_trusted(self):
_, report = analyze('# A\n')
with self.assertRaises(ValueError):
validate_manual(report, {'manual': {'unknown': {'text': 'A', 'level': 2}}})
def test_summary_cannot_drop_confirmed_title(self):
text, report = analyze('# Root\n## Child\n', {'format': 'md'})
text, report = manual_edit(text, report, {'id': report['headings'][1]['id'], 'kind': 'confirm'})
with self.assertRaises(ValueError):
protect_confirmed(text, '# Summary\n', report)
out, _ = protect_confirmed(text, '# Root\n# Child\nSummary body\n', report)
self.assertIn('## Child', out)
if __name__ == '__main__':
unittest.main()