# This file is part of Wolnelektury, licensed under GNU Affero GPLv3 or later.
# Copyright © Fundacja Nowoczesna Polska. See NOTICE for more information.
#
-from django.contrib.contenttypes.models import ContentType
-from django.db.models import Count
+from django.conf import settings
from .models import Tag, Book
+from os.path import getmtime
+import cPickle
+from collections import defaultdict
BOOK_CATEGORIES = ('author', 'epoch', 'genre', 'kind')
+_COUNTERS = None
+_COUNTER_TIME = None
-def get_top_level_related_tags(tags=None, categories=BOOK_CATEGORIES):
+
+def get_top_level_related_tags(tags, categories=None):
"""
Finds tags related to given tags through books, and counts their usage.
Takes ancestry into account: if a tag is applied to a book, its
usage on the book's descendants is ignored.
+ """
+ global _COUNTERS, _COUNTER_TIME
+ # First, check that we have a valid and recent version of the counters.
+ if getmtime(settings.CATALOGUE_COUNTERS_FILE) > _COUNTER_TIME:
+ with open(settings.CATALOGUE_COUNTERS_FILE) as f:
+ _COUNTERS = cPickle.load(f)
- This is tested for PostgreSQL 9.1+, and might not work elsewhere.
- It particular, it uses raw SQL using WITH clause, which is
- supported in SQLite from v. 3.8.3, and is missing in MySQL.
- http://bugs.mysql.com/bug.php?id=16244
+ tagids = tuple(sorted(t.pk for t in tags))
+ try:
+ related_ids = _COUNTERS['next'][tagids]
+ except KeyError:
+ return
- """
- # First, find all tag relations of relevant books.
- bct = ContentType.objects.get_for_model(Book)
- relations = Tag.intermediary_table_model.objects.filter(
- content_type=bct)
- if tags is not None:
- tagged_books = Book.tagged.with_all(tags).only('pk')
- relations = relations.filter(
- object_id__in=tagged_books).exclude(
- tag_id__in=[tag.pk for tag in tags])
-
- rel_sql, rel_params = relations.query.sql_with_params()
-
- # Exclude those relations between a book and a tag,
- # for which there is a relation between the book's ancestor
- # and the tag and
-
- return Tag.objects.raw('''
- WITH AllTagged AS (''' + rel_sql + ''')
- SELECT catalogue_tag.*, COUNT(catalogue_tag.id) AS count
- FROM catalogue_tag, AllTagged
- WHERE catalogue_tag.id=AllTagged.tag_id
- AND catalogue_tag.category IN %s
- AND NOT EXISTS (
- SELECT AncestorTagged.id
- FROM catalogue_book_ancestor Ancestor,
- AllTagged AncestorTagged
- WHERE Ancestor.from_book_id=AllTagged.object_id
- AND AncestorTagged.content_type_id=%s
- AND AncestorTagged.object_id=Ancestor.to_book_id
- AND AncestorTagged.tag_id=AllTagged.tag_id
- )
- GROUP BY catalogue_tag.id
- ORDER BY sort_key''', rel_params + (categories, bct.pk))
+ related = Tag.objects.filter(pk__in=related_ids)
+
+ # TODO: do we really need that?
+ if categories is not None:
+ related = related.filter(category__in=categories)
+
+ for tag in related:
+ tag.count = _COUNTERS['count'][tuple(sorted(tagids + (tag.pk,)))]
+ yield tag
+
+
+def update_counters():
+ def combinations(things):
+ if len(things):
+ for c in combinations(things[1:]):
+ yield c
+ yield (things[0],) + c
+ else:
+ yield ()
+
+ def count_for_book(book, count_by_combination=None, parent_combinations=None):
+ if not parent_combinations:
+ parent_combinations = set()
+ tags = sorted(tuple(t.pk for t in book.tags.filter(category__in=('author', 'genre', 'epoch', 'kind'))))
+ combs = list(combinations(tags))
+ for c in combs:
+ if c not in parent_combinations:
+ count_by_combination[c] += 1
+ combs_for_child = set(list(parent_combinations) + combs)
+ for child in book.children.all():
+ count_for_book(child, count_by_combination, combs_for_child)
+
+ count_by_combination = defaultdict(lambda: 0)
+ for b in Book.objects.filter(parent=None):
+ count_for_book(b, count_by_combination)
+
+ next_combinations = defaultdict(set)
+ # Now build an index of all combinations.
+ for c in count_by_combination.keys():
+ if not c:
+ continue
+ for n in c:
+ rest = tuple(x for x in c if x != n)
+ next_combinations[rest].add(n)
+
+ counters = {
+ "count": dict(count_by_combination),
+ "next": dict(next_combinations),
+ }
+
+ with open(settings.CATALOGUE_COUNTERS_FILE, 'w') as f:
+ cPickle.dump(counters, f)
fnp / wolnelektury.git / blobdiff