Übersetzungskataloge, getrennte Sprachpräferenzen und einen versionierten Aufgabenpilot ergänzen. Historische Aufgaben und Bewertungen bleiben erhalten; äquivalente Ansichten teilen Nachweise. Lokal mit Race-Tests und PostgreSQL, Vet, Web-Build, Bank-Audits und Browser-Fixtures geprüft. Dieser Stand dient als Ausgangspunkt für die anschließend beauftragte Zusammenführung zur Muttersprache.
52 lines
3.8 KiB
Python
52 lines
3.8 KiB
Python
"""Offline regression checks against separately reviewed grammar solutions."""
|
|
from build import lines,KEYS,DEST,OPEN
|
|
import json,collections
|
|
EXPECTED={
|
|
'introductions':['am','are','is','is','is','My','from','is','are','are','am','are','His','Her','What','Where','am','is','your','Who','am','is','are','is'],
|
|
'family':['have','has','has','She','He','have','brothers','My','have','Has','Have','has not','They','Her','His','has','has','have not','children','are','has','He','have','Has'],
|
|
'numbers_time':['is','at','on','in','days','is','time','day','past','half','to','on','in','first','second','minutes','How','When','after','are','at','on','in','third'],
|
|
'daily_life':['eat','plays','goes','read','play','do','Do','Does','does','play','does','always','at','on','studies','washes','do','does not','When','finishes','clean','do','cooks','Does'],
|
|
'food_shopping':['an','a','apples','some','is','are','would','any','any','many','much','of','Can','tomatoes','This','These','do','costs','or','some','an','many','of','are'],
|
|
'home_city':['is','are','in','on','under','next to','Where','left','right','straight','is','Is','Are','opposite','between','behind','in front of','rooms','our','is','are','Where','next to','are']}
|
|
for module in KEYS[:6]:
|
|
for index,row in enumerate(lines(module,'grammar')):
|
|
assert row[1].count('___')==1,(module,index,'blank')
|
|
assert row[2]==EXPECTED[module][index],(module,index,'reviewed grammar key')
|
|
assert len(set(x.casefold() for x in row[2:5]))==3,(module,index,'options')
|
|
for index,row in enumerate(lines(module,'vocabulary')):
|
|
word,answer,b,c,example,goal=row
|
|
assert word.casefold() in example.casefold(),(module,index,'missing word context')
|
|
assert len(set(x.casefold() for x in [answer,b,c]))==3,(module,index,'translations')
|
|
for index,row in enumerate(lines(module,'stories')):
|
|
assert len(row)==11,(module,index,'story format')
|
|
assert row[1].endswith('?') and row[5].endswith('?'),(module,index,'questions')
|
|
assert len(set(x.casefold() for x in row[2:5]))==3 and len(set(x.casefold() for x in row[6:9]))==3,(module,index,'story options')
|
|
assert not any(w in row[0].lower() for w in ['look at the picture','see the photo','http','<']), (module,index,'external context')
|
|
for module in KEYS[:6]:
|
|
assert len(OPEN[module])==14,(module,'open instruction coverage')
|
|
for instruction,criterion in OPEN[module]:
|
|
assert instruction and criterion and 'Situation' not in instruction,(module,instruction)
|
|
entries=[]
|
|
for path in sorted(DEST.glob('*.json')):
|
|
if path.name!='manifest.json':entries+=json.loads(path.read_text())
|
|
assert len(entries)==840
|
|
counts=collections.Counter((e['task']['module_id'],e['task']['skill']) for e in entries)
|
|
assert len(counts)==42 and set(counts.values())=={20}
|
|
for key in counts:
|
|
bucket=[e for e in entries if (e['task']['module_id'],e['task']['skill'])==key]
|
|
assert collections.Counter(e['difficulty'] for e in bucket)=={1:8,2:8,3:4}
|
|
if key[1] in ['grammar','vocabulary','reading','listening']:
|
|
assert sorted(collections.Counter(e['task']['answer_key']['correct_option_id'] for e in bucket).values())==[6,7,7]
|
|
assert len({e['id'] for e in entries})==840
|
|
changed=[e for e in entries if e['task']['kind'] in ['free_speech','free_text']]
|
|
assert len(changed)==140 and all(e['id'].startswith('a1bank_en_v2_') for e in changed)
|
|
for e in changed:
|
|
t=e['task']
|
|
assert 'Situation' not in t['prompt'] and 'eigene oder erfundene Angaben' in t['prompt'],e['id']
|
|
assert len(t['answer_key']['rubric'])>=4 and 'Werte aus der Beispielantwort sind nicht verbindlich.' in ' '.join(t['answer_key']['rubric']),e['id']
|
|
assert all(e['id'].startswith('a1bank_en_v1_') for e in entries if e not in changed)
|
|
print('PASS: 144 reviewed grammar keys, 144 contextual vocabulary items, 144 factual scenes; 840 outputs, 42 balanced buckets, media completeness and balanced choice positions.')
|
|
|
|
from variants import audit as audit_variants
|
|
audit_variants()
|