Add invariant tests

test/test_invariants.py

@@ -0,0 +1,116 @@
+"""
+Invariant tests for cterdam/iso_3166-2 dataset.
+
+Run from dataset root after cloning:
+    pytest test/test_invariants.py -v
+
+Verifies:
+- name field matches lang selection (name_en/ru/zh)
+- hash = SHA256(name)
+- lang matches parent region's language rule
+- hash format is 64-char hex string
+"""
+
+import hashlib
+import pytest
+from datasets import load_dataset
+
+SLAVIC_REGIONS = {
+    'RU', 'UA', 'BY', 'KZ', 'UZ', 'TJ', 'KG', 'TM', 'GE', 'AZ', 'AM', 'MD',
+    'LT', 'LV', 'EE', 'RS', 'ME', 'BA', 'HR', 'SI', 'SK', 'CZ', 'PL', 'BG', 'MK', 'FI'
+}
+
+SINOPHONE_REGIONS = {
+    'CN', 'TW', 'HK', 'MO', 'JP', 'KR', 'KP', 'MN', 'SG', 'VN', 'LA', 'TH', 'MM', 'KH', 'MY', 'BN'
+}
+
+
+def determine_expected_lang(region: str) -> str:
+    """Determine expected lang from parent region code."""
+    if region in SINOPHONE_REGIONS:
+        return 'zh'
+    elif region in SLAVIC_REGIONS:
+        return 'ru'
+    else:
+        return 'en'
+
+
+@pytest.fixture(scope="session")
+def dataset():
+    return load_dataset('.', split='train')
+
+
+def test_lang_values_valid(dataset):
+    """lang field must be one of: en, ru, zh."""
+    invalid = [(i, e['lang']) for i, e in enumerate(dataset) if e.get('lang') not in ('en', 'ru', 'zh')]
+    assert invalid == [], f"Invalid lang values: {invalid[:5]}"
+
+
+def test_hash_format_valid(dataset):
+    """hash field must be 64-char hex string (if name exists)."""
+    invalid = [
+        (i, e['hash'])
+        for i, e in enumerate(dataset)
+        if e.get('hash') is not None and not (
+            isinstance(e['hash'], str) and
+            len(e['hash']) == 64 and
+            all(c in '0123456789abcdef' for c in e['hash'])
+        )
+    ]
+    assert invalid == [], f"Invalid hash format: {invalid[:5]}"
+
+
+def test_name_matches_lang_field(dataset):
+    """name field must equal name_en/ru/zh based on lang."""
+    lang_to_field = {'en': 'name_en', 'ru': 'name_ru', 'zh': 'name_zh'}
+    mismatches = []
+    
+    for i, entry in enumerate(dataset):
+        name = entry.get('name')
+        lang = entry.get('lang')
+        if not name or not lang:
+            continue
+        
+        field_name = lang_to_field.get(lang)
+        lang_field_value = entry.get(field_name)
+        
+        if name != lang_field_value:
+            mismatches.append((i, entry.get('code'), lang, name[:20], lang_field_value[:20] if lang_field_value else None))
+            if len(mismatches) >= 5:
+                break
+    
+    assert mismatches == [], f"Name/lang field mismatches: {mismatches}"
+
+
+def test_hash_matches_name(dataset):
+    """hash field must equal SHA256(name)."""
+    mismatches = []
+    
+    for i, entry in enumerate(dataset):
+        name = entry.get('name')
+        stored_hash = entry.get('hash')
+        if name and stored_hash:
+            expected_hash = hashlib.sha256(name.encode()).hexdigest()
+            if stored_hash != expected_hash:
+                mismatches.append((i, entry.get('code'), stored_hash[:8], expected_hash[:8]))
+                if len(mismatches) >= 5:
+                    break
+    
+    assert mismatches == [], f"Hash mismatches: {mismatches}"
+
+
+def test_lang_matches_parent_region(dataset):
+    """lang field must match parent region's language rule."""
+    mismatches = []
+    
+    for i, entry in enumerate(dataset):
+        region = entry.get('region')
+        lang = entry.get('lang')
+        expected_lang = determine_expected_lang(region)
+        
+        if lang != expected_lang:
+            mismatches.append((i, entry.get('code'), region, lang, expected_lang))
+            if len(mismatches) >= 5:
+                break
+    
+    assert mismatches == [], f"Lang mismatches: {mismatches}"