Python code for checking that a document's slug is a sub-path of its parent's slug -- for all documents in all locales on MDN

hierarchy_check.py

from kuma.wiki.constants import EXPERIMENT_TITLE_PREFIX, LEGACY_MINDTOUCH_NAMESPACES, NOINDEX_SLUG_PREFIXES


def get_docs(locale):
    qs = Document.objects.filter(locale=locale, is_redirect=False).exclude(html="")
    qs = qs.only("id", "locale", "slug", "parent_topic")
    qs = qs.select_related('parent_topic')
    qs = qs.exclude(current_revision__isnull=True)
    q = Q(slug__startswith=EXPERIMENT_TITLE_PREFIX)
    for legacy_mindtouch_namespace in LEGACY_MINDTOUCH_NAMESPACES:
        q |= Q(slug__startswith=f"{legacy_mindtouch_namespace}:")
    for slug_start in NOINDEX_SLUG_PREFIXES:
        q |= Q(slug__startswith=slug_start)
    qs = qs.exclude(q)
    qs = qs.order_by("id")
    return list(qs)


def check(chunk):
    failures = []
    for i, doc in enumerate(chunk):
        parent = doc.parent_topic
        if parent and ('/'.join(doc.slug.split('/')[:-1]).lower() != parent.slug.lower()):
            failures.append((doc.slug, parent.slug))
    return failures


with open('hierarchy_check_failures.log', 'w') as log:
    for locale in settings.ENABLED_LOCALES:
        docs = get_docs(locale)
        failures = check(docs)
        header = f'locale = {locale}, {len(failures)} failures within {len(docs)} documents'
        print(header)
        dashes = '-' * len(header)
        print(dashes + '\n' + header + '\n' + dashes, file=log)
        for slug, parent_slug in failures:
            print(f'   doc.slug = {slug}', file=log)
            print(f'   doc.parent_topic.slug = {parent_slug}', file=log)
            print(f'   ---', file=log)