--- /dev/null
+# -*- coding: utf-8 -*-
+#
+# This file is part of FNP-Redakcja, licensed under GNU Affero GPLv3 or later.
+# Copyright © Fundacja Nowoczesna Polska. See NOTICE for more information.
+#
+
+import sys
+from django.contrib.auth.models import User
+from lxml import etree
+from optparse import make_option
+
+from django.core.management import BaseCommand
+
+from catalogue.models import Book
+from librarian import DCNS
+
+
+class Command(BaseCommand):
+ option_list = BaseCommand.option_list + (
+ # make_option('-q', '--quiet', action='store_false', dest='verbose',
+ # default=True, help='Less output'),
+ # make_option('-d', '--dry-run', action='store_true', dest='dry_run',
+ # default=False, help="Don't actually touch anything"),
+ make_option(
+ '-u', '--username', dest='username', metavar='USER',
+ help='Assign commits to this user (required, preferably yourself).'),
+ )
+ args = 'exclude_file'
+
+ def handle(self, exclude_file, **options):
+ username = options.get('username')
+
+ if username:
+ user = User.objects.get(username=username)
+ else:
+ print 'Please provide a username.'
+ sys.exit(1)
+
+ excluded_slugs = [line.strip() for line in open(exclude_file, 'rb') if line.strip()]
+ books = Book.objects.exclude(slug__in=excluded_slugs)
+
+ for book in books:
+ if not book.is_published():
+ continue
+ print 'processing %s' % book.slug
+ chunk = book.chunk_set.first()
+ old_head = chunk.head
+ src = old_head.materialize()
+ tree = etree.fromstring(src)
+ audience_nodes = tree.findall('.//' + DCNS("audience"))
+ if not audience_nodes:
+ print '%s has no audience, skipping' % book.slug
+ continue
+
+ for node in audience_nodes:
+ node.getparent().remove(node)
+
+ chunk.commit(
+ etree.tostring(tree, encoding=unicode),
+ author=user,
+ description='automatyczne skasowanie audience',
+ publishable=old_head.publishable
+ )
+ print 'committed %s' % book.slug
+ if not old_head.publishable:
+ print 'Warning: %s not publishable, last head: %s, %s' % (
+ book.slug, old_head.author.username, old_head.description[:40].replace('\n', ' '))