fixed multithreading bug with attaching to jvm
[wolnelektury.git] / apps / catalogue / models.py
index ab480d7..0863d5b 100644 (file)
@@ -2,8 +2,12 @@
 # This file is part of Wolnelektury, licensed under GNU Affero GPLv3 or later.
 # Copyright © Fundacja Nowoczesna Polska. See NOTICE for more information.
 #
+from datetime import datetime
+
 from django.db import models
 from django.db.models import permalink, Q
+import django.dispatch
+from django.core.cache import cache
 from django.utils.translation import ugettext_lazy as _
 from django.contrib.auth.models import User
 from django.core.files import File
@@ -17,12 +21,17 @@ from django.conf import settings
 
 from newtagging.models import TagBase, tags_updated
 from newtagging import managers
-from catalogue.fields import JSONField
+from catalogue.fields import JSONField, OverwritingFileField
+from catalogue.utils import ExistingFile, ORMDocProvider, create_zip, remove_zip
 
 from librarian import dcparser, html, epub, NoDublinCore
+import mutagen
 from mutagen import id3
 from slughifi import slughifi
+from sortify import sortify
+from os import unlink
 
+import search
 
 TAG_CATEGORIES = (
     ('author', _('author')),
@@ -41,6 +50,10 @@ MEDIA_FORMATS = (
     ('daisy', _('DAISY file')), 
 )
 
+# not quite, but Django wants you to set a timeout
+CACHE_FOREVER = 2419200  # 28 days
+
+
 class TagSubcategoryManager(models.Manager):
     def __init__(self, subcategory):
         super(TagSubcategoryManager, self).__init__()
@@ -64,6 +77,12 @@ class Tag(TagBase):
     gazeta_link = models.CharField(blank=True, max_length=240)
     wiki_link = models.CharField(blank=True, max_length=240)
 
+    created_at    = models.DateTimeField(_('creation date'), auto_now_add=True, db_index=True)
+    changed_at    = models.DateTimeField(_('creation date'), auto_now=True, db_index=True)
+
+    class UrlDeprecationWarning(DeprecationWarning):
+        pass
+
     categories_rev = {
         'autor': 'author',
         'epoka': 'epoch',
@@ -122,19 +141,20 @@ class Tag(TagBase):
             real_tags = []
             ambiguous_slugs = []
             category = None
+            deprecated = False
             tags_splitted = tags.split('/')
-            for index, name in enumerate(tags_splitted):
-                if name in Tag.categories_rev:
+            for name in tags_splitted:
+                if category:
+                    real_tags.append(Tag.objects.get(slug=name, category=category))
+                    category = None
+                elif name in Tag.categories_rev:
                     category = Tag.categories_rev[name]
                 else:
-                    if category:
-                        real_tags.append(Tag.objects.get(slug=name, category=category))
-                        category = None
-                    else:
-                        try:
-                            real_tags.append(Tag.objects.exclude(category='book').get(slug=name))
-                        except Tag.MultipleObjectsReturned, e:
-                            ambiguous_slugs.append(name)
+                    try:
+                        real_tags.append(Tag.objects.exclude(category='book').get(slug=name))
+                        deprecated = True 
+                    except Tag.MultipleObjectsReturned, e:
+                        ambiguous_slugs.append(name)
 
             if category:
                 # something strange left off
@@ -145,8 +165,11 @@ class Tag(TagBase):
                 e.tags = real_tags
                 e.ambiguous_slugs = ambiguous_slugs
                 raise e
-            else:
-                return real_tags
+            if deprecated:
+                e = Tag.UrlDeprecationWarning()
+                e.tags = real_tags
+                raise e
+            return real_tags
         else:
             return TagBase.get_tag_list(tags)
 
@@ -156,25 +179,30 @@ class Tag(TagBase):
 
 
 # TODO: why is this hard-coded ?
-def book_upload_path(ext=None):
+def book_upload_path(ext=None, maxlen=100):
     def get_dynamic_path(media, filename, ext=ext):
         # how to put related book's slug here?
         if not ext:
-            ext = media.type
+            if media.type == 'daisy':
+                ext = 'daisy.zip'
+            else:
+                ext = media.type
         if not media.name:
             name = slughifi(filename.split(".")[0])
         else:
             name = slughifi(media.name)
-        return 'lektura/%s.%s' % (name, ext)
+        return 'book/%s/%s.%s' % (ext, name[:maxlen-len('book/%s/.%s' % (ext, ext))-4], ext)
     return get_dynamic_path
 
 
 class BookMedia(models.Model):
     type        = models.CharField(_('type'), choices=MEDIA_FORMATS, max_length="100")
-    name        = models.CharField(_('name'), max_length="100", blank=True)
-    file        = models.FileField(_('file'), upload_to=book_upload_path(), blank=True)
+    name        = models.CharField(_('name'), max_length="100")
+    file        = OverwritingFileField(_('file'), upload_to=book_upload_path())
     uploaded_at = models.DateTimeField(_('creation date'), auto_now_add=True, editable=False)
-    extra_info  = JSONField(_('extra information'), default='{}')
+    extra_info  = JSONField(_('extra information'), default='{}', editable=False)
+    book = models.ForeignKey('Book', related_name='media')
+    source_sha1 = models.CharField(null=True, blank=True, max_length=40, editable=False)
 
     def __unicode__(self):
         return "%s (%s)" % (self.name, self.file.name.split("/")[-1])
@@ -184,59 +212,108 @@ class BookMedia(models.Model):
         verbose_name        = _('book media')
         verbose_name_plural = _('book media')
 
-    def save(self, force_insert=False, force_update=False, **kwargs):
-        media = super(BookMedia, self).save(force_insert, force_update, **kwargs)
-        if self.type == 'mp3':
-            file = self.file
-            extra_info = self.get_extra_info_value()
-            extra_info.update(self.get_mp3_info())
-            self.set_extra_info_value(extra_info)
-            media = super(BookMedia, self).save(force_insert, force_update, **kwargs)
-        return media
-
-    def get_mp3_info(self):
-        """Retrieves artist and director names from audio ID3 tags."""
+    def save(self, *args, **kwargs):
         try:
-            audio = id3.ID3(self.file.path)
-            artist_name = ', '.join(', '.join(tag.text) for tag in audio.getall('TPE1'))
-            director_name = ', '.join(', '.join(tag.text) for tag in audio.getall('TPE3'))
-        except:
-            artist_name = director_name = ''
-        return {'artist_name': artist_name, 'director_name': director_name}
+            old = BookMedia.objects.get(pk=self.pk)
+        except BookMedia.DoesNotExist, e:
+            pass
+        else:
+            # if name changed, change the file name, too
+            if slughifi(self.name) != slughifi(old.name):
+                self.file.save(None, ExistingFile(self.file.path), save=False, leave=True)
+
+        super(BookMedia, self).save(*args, **kwargs)
+
+        # remove the zip package for book with modified media
+        remove_zip(self.book.slug)
+
+        extra_info = self.get_extra_info_value()
+        extra_info.update(self.read_meta())
+        self.set_extra_info_value(extra_info)
+        self.source_sha1 = self.read_source_sha1(self.file.path, self.type)
+        return super(BookMedia, self).save(*args, **kwargs)
+
+    def read_meta(self):
+        """
+            Reads some metadata from the audiobook.
+        """
+
+        artist_name = director_name = project = funded_by = ''
+        if self.type == 'mp3':
+            try:
+                audio = id3.ID3(self.file.path)
+                artist_name = ', '.join(', '.join(tag.text) for tag in audio.getall('TPE1'))
+                director_name = ', '.join(', '.join(tag.text) for tag in audio.getall('TPE3'))
+                project = ", ".join([t.data for t in audio.getall('PRIV') 
+                        if t.owner=='wolnelektury.pl?project'])
+                funded_by = ", ".join([t.data for t in audio.getall('PRIV') 
+                        if t.owner=='wolnelektury.pl?funded_by'])
+            except:
+                pass
+        elif self.type == 'ogg':
+            try:
+                audio = mutagen.File(self.file.path)
+                artist_name = ', '.join(audio.get('artist', []))
+                director_name = ', '.join(audio.get('conductor', []))
+                project = ", ".join(audio.get('project', []))
+                funded_by = ", ".join(audio.get('funded_by', []))
+            except:
+                pass
+        else:
+            return {}
+        return {'artist_name': artist_name, 'director_name': director_name,
+                'project': project, 'funded_by': funded_by}
+
+    @staticmethod
+    def read_source_sha1(filepath, filetype):
+        """
+            Reads source file SHA1 from audiobok metadata.
+        """
+
+        if filetype == 'mp3':
+            try:
+                audio = id3.ID3(filepath)
+                return [t.data for t in audio.getall('PRIV') 
+                        if t.owner=='wolnelektury.pl?flac_sha1'][0]
+            except:
+                return None
+        elif filetype == 'ogg':
+            try:
+                audio = mutagen.File(filepath)
+                return audio.get('flac_sha1', [None])[0] 
+            except:
+                return None
+        else:
+            return None
 
 
 class Book(models.Model):
     title         = models.CharField(_('title'), max_length=120)
+    sort_key = models.CharField(_('sort key'), max_length=120, db_index=True, editable=False)
     slug          = models.SlugField(_('slug'), max_length=120, unique=True, db_index=True)
     description   = models.TextField(_('description'), blank=True)
-    created_at    = models.DateTimeField(_('creation date'), auto_now_add=True)
-    _short_html   = models.TextField(_('short HTML'), editable=False)
+    created_at    = models.DateTimeField(_('creation date'), auto_now_add=True, db_index=True)
+    changed_at    = models.DateTimeField(_('creation date'), auto_now=True, db_index=True)
     parent_number = models.IntegerField(_('parent number'), default=0)
-    extra_info    = JSONField(_('extra information'))
+    extra_info    = JSONField(_('extra information'), default='{}')
     gazeta_link   = models.CharField(blank=True, max_length=240)
     wiki_link     = models.CharField(blank=True, max_length=240)
     # files generated during publication
-    xml_file      = models.FileField(_('XML file'), upload_to=book_upload_path('xml'), blank=True)
-    html_file     = models.FileField(_('HTML file'), upload_to=book_upload_path('html'), blank=True)
-    pdf_file      = models.FileField(_('PDF file'), upload_to=book_upload_path('pdf'), blank=True)
-    epub_file     = models.FileField(_('EPUB file'), upload_to=book_upload_path('epub'), blank=True)    
-    txt_file      = models.FileField(_('TXT file'), upload_to=book_upload_path('txt'), blank=True)        
-    # other files
-    medias        = models.ManyToManyField(BookMedia, blank=True)
+
+    file_types = ['epub', 'html', 'mobi', 'pdf', 'txt', 'xml']
     
     parent        = models.ForeignKey('self', blank=True, null=True, related_name='children')
     objects  = models.Manager()
     tagged   = managers.ModelTaggedItemManager(Tag)
     tags     = managers.TagDescriptor(Tag)
 
-    _tag_counter = JSONField(null=True, editable=False)
-    _theme_counter = JSONField(null=True, editable=False)
+    html_built = django.dispatch.Signal()
 
     class AlreadyExists(Exception):
         pass
 
     class Meta:
-        ordering = ('title',)
+        ordering = ('sort_key',)
         verbose_name = _('book')
         verbose_name_plural = _('books')
 
@@ -244,16 +321,14 @@ class Book(models.Model):
         return self.title
 
     def save(self, force_insert=False, force_update=False, reset_short_html=True, **kwargs):
+        self.sort_key = sortify(self.title)
+
+        ret = super(Book, self).save(force_insert, force_update)
+
         if reset_short_html:
-            # Reset _short_html during save
-            update = {}
-            for key in filter(lambda x: x.startswith('_short_html'), self.__dict__):
-                update[key] = ''
-                self.__setattr__(key, '')
-            # Fragment.short_html relies on book's tags, so reset it here too
-            self.fragments.all().update(**update)
+            self.reset_short_html()
 
-        return super(Book, self).save(force_insert, force_update)
+        return ret
 
     @permalink
     def get_absolute_url(self):
@@ -276,51 +351,17 @@ class Book(models.Model):
         return book_tag
 
     def has_media(self, type):
-        if   type == 'xml':
-            if self.xml_file:
-                return True
-            else:
-                return False
-        elif type == 'html':
-            if self.html_file:
-                return True
-            else:
-                return False        
-        elif type == 'txt':
-            if self.txt_file:
-                return True
-            else:
-                return False        
-        elif type == 'pdf':
-            if self.pdf_file:
-                return True
-            else:
-                return False  
-        elif type == 'epub':
-            if self.epub_file:
-                return True
-            else:
-                return False                          
+        if type in Book.file_types:
+            return bool(getattr(self, "%s_file" % type))
         else:
-            if self.medias.filter(book=self, type=type).count() > 0:
-                return True
-            else:
-                return False
+            return self.media.filter(type=type).exists()
 
     def get_media(self, type):
         if self.has_media(type):
-            if   type == "xml":
-                return self.xml_file
-            elif type == "html":
-                return self.html_file
-            elif type == "epub":
-                return self.epub_file
-            elif type == "txt":
-                return self.txt_file
-            elif type == "pdf":
-                return self.pdf_file
+            if type in Book.file_types:
+                return getattr(self, "%s_file" % type)
             else:                                             
-                return self.medias.filter(book=self, type=type)
+                return self.media.filter(type=type)
         else:
             return None
 
@@ -333,37 +374,54 @@ class Book(models.Model):
     def get_daisy(self):
         return self.get_media("daisy")                       
 
+    def reset_short_html(self):
+        if self.id is None:
+            return
+
+        cache_key = "Book.short_html/%d/%s"
+        for lang, langname in settings.LANGUAGES:
+            cache.delete(cache_key % (self.id, lang))
+        # Fragment.short_html relies on book's tags, so reset it here too
+        for fragm in self.fragments.all():
+            fragm.reset_short_html()
+
     def short_html(self):
-        key = '_short_html_%s' % get_language()
-        short_html = getattr(self, key)
+        if self.id:
+            cache_key = "Book.short_html/%d/%s" % (self.id, get_language())
+            short_html = cache.get(cache_key)
+        else:
+            short_html = None
 
-        if short_html and len(short_html):
+        if short_html is not None:
             return mark_safe(short_html)
         else:
             tags = self.tags.filter(~Q(category__in=('set', 'theme', 'book')))
             tags = [mark_safe(u'<a href="%s">%s</a>' % (tag.get_absolute_url(), tag.name)) for tag in tags]
 
             formats = []
-            # files generated during publication               
+            # files generated during publication
             if self.has_media("html"):
                 formats.append(u'<a href="%s">%s</a>' % (reverse('book_text', kwargs={'slug': self.slug}), _('Read online')))
             if self.has_media("pdf"):
                 formats.append(u'<a href="%s">PDF</a>' % self.get_media('pdf').url)
+            if self.has_media("mobi"):
+                formats.append(u'<a href="%s">MOBI</a>' % self.get_media('mobi').url)
             if self.root_ancestor.has_media("epub"):
                 formats.append(u'<a href="%s">EPUB</a>' % self.root_ancestor.get_media('epub').url)
             if self.has_media("txt"):
                 formats.append(u'<a href="%s">TXT</a>' % self.get_media('txt').url)
             # other files
-            for m in self.medias.order_by('type'):
+            for m in self.media.order_by('type'):
                 formats.append(u'<a href="%s">%s</a>' % (m.file.url, m.type.upper()))
 
             formats = [mark_safe(format) for format in formats]
 
-            setattr(self, key, unicode(render_to_string('catalogue/book_short.html',
-                {'book': self, 'tags': tags, 'formats': formats})))
-            self.save(reset_short_html=False)
-            return mark_safe(getattr(self, key))
+            short_html = unicode(render_to_string('catalogue/book_short.html',
+                {'book': self, 'tags': tags, 'formats': formats}))
 
+            if self.id:
+                cache.set(cache_key, short_html, CACHE_FOREVER)
+            return mark_safe(short_html)
 
     @property
     def root_ancestor(self):
@@ -383,26 +441,6 @@ class Book(models.Model):
     has_description.boolean = True
 
     # ugly ugly ugly
-    def has_pdf_file(self):
-        return bool(self.pdf_file)
-    has_pdf_file.short_description = 'PDF'
-    has_pdf_file.boolean = True
-
-    def has_epub_file(self):
-        return bool(self.epub_file)
-    has_epub_file.short_description = 'EPUB'
-    has_epub_file.boolean = True
-
-    def has_txt_file(self):
-        return bool(self.txt_file)
-    has_txt_file.short_description = 'HTML'
-    has_txt_file.boolean = True
-
-    def has_html_file(self):
-        return bool(self.html_file)
-    has_html_file.short_description = 'HTML'
-    has_html_file.boolean = True
-
     def has_odt_file(self):
         return bool(self.has_media("odt"))
     has_odt_file.short_description = 'ODT'
@@ -417,34 +455,64 @@ class Book(models.Model):
         return bool(self.has_media("ogg"))
     has_ogg_file.short_description = 'OGG'
     has_ogg_file.boolean = True
-    
+
     def has_daisy_file(self):
         return bool(self.has_media("daisy"))
     has_daisy_file.short_description = 'DAISY'
-    has_daisy_file.boolean = True    
-    
+    has_daisy_file.boolean = True
+
+    def build_pdf(self):
+        """ (Re)builds the pdf file.
+
+        """
+        from librarian import pdf
+        from tempfile import NamedTemporaryFile
+        import os
+
+        try:
+            pdf_file = NamedTemporaryFile(delete=False)
+            pdf.transform(ORMDocProvider(self),
+                      file_path=str(self.xml_file.path),
+                      output_file=pdf_file,
+                      )
+
+            self.pdf_file.save('%s.pdf' % self.slug, File(open(pdf_file.name)))
+        finally:
+            unlink(pdf_file.name)
+
+        # remove zip with all pdf files
+        remove_zip(settings.ALL_PDF_ZIP)
+
+    def build_mobi(self):
+        """ (Re)builds the MOBI file.
+
+        """
+        from librarian import mobi
+        from tempfile import NamedTemporaryFile
+        import os
+
+        try:
+            mobi_file = NamedTemporaryFile(suffix='.mobi', delete=False)
+            mobi.transform(ORMDocProvider(self), verbose=1,
+                      file_path=str(self.xml_file.path),
+                      output_file=mobi_file.name,
+                      )
+
+            self.mobi_file.save('%s.mobi' % self.slug, File(open(mobi_file.name)))
+        finally:
+            unlink(mobi_file.name)
+
+        # remove zip with all mobi files
+        remove_zip(settings.ALL_MOBI_ZIP)
+
     def build_epub(self, remove_descendants=True):
         """ (Re)builds the epub file.
             If book has a parent, does nothing.
             Unless remove_descendants is False, descendants' epubs are removed.
         """
-    
         from StringIO import StringIO
         from hashlib import sha1
         from django.core.files.base import ContentFile
-        from librarian import DocProvider
-
-        class BookImportDocProvider(DocProvider):
-            """ used for joined EPUBs """
-
-            def __init__(self, book):
-                self.book = book
-
-            def by_slug(self, slug):
-                if slug == self.book.slug:
-                    return self.book.xml_file
-                else:
-                    return Book.objects.get(slug=slug).xml_file
 
         if self.parent:
             # don't need an epub
@@ -452,7 +520,7 @@ class Book(models.Model):
 
         epub_file = StringIO()
         try:
-            epub.transform(BookImportDocProvider(self), self.slug, output_file=epub_file)
+            epub.transform(ORMDocProvider(self), self.slug, output_file=epub_file)
             self.epub_file.save('%s.epub' % self.slug, ContentFile(epub_file.getvalue()))
             FileRecord(slug=self.slug, type='epub', sha1=sha1(epub_file.getvalue()).hexdigest()).save()
         except NoDublinCore:
@@ -467,6 +535,9 @@ class Book(models.Model):
             child_book.save()
             book_descendants += list(child_book.children.all())
 
+        # remove zip package with all epub files
+        remove_zip(settings.ALL_EPUB_ZIP)
+
     def build_txt(self):
         from StringIO import StringIO
         from django.core.files.base import ContentFile
@@ -475,9 +546,93 @@ class Book(models.Model):
         out = StringIO()
         text.transform(open(self.xml_file.path), out)
         self.txt_file.save('%s.txt' % self.slug, ContentFile(out.getvalue()))
-        self.save()
 
 
+    def build_html(self):
+        from tempfile import NamedTemporaryFile
+        from markupstring import MarkupString
+
+        meta_tags = list(self.tags.filter(
+            category__in=('author', 'epoch', 'genre', 'kind')))
+        book_tag = self.book_tag()
+
+        html_file = NamedTemporaryFile()
+        if html.transform(self.xml_file.path, html_file, parse_dublincore=False):
+            self.html_file.save('%s.html' % self.slug, File(html_file))
+
+            # get ancestor l-tags for adding to new fragments
+            ancestor_tags = []
+            p = self.parent
+            while p:
+                ancestor_tags.append(p.book_tag())
+                p = p.parent
+
+            # Delete old fragments and create them from scratch
+            self.fragments.all().delete()
+            # Extract fragments
+            closed_fragments, open_fragments = html.extract_fragments(self.html_file.path)
+            for fragment in closed_fragments.values():
+                try:
+                    theme_names = [s.strip() for s in fragment.themes.split(',')]
+                except AttributeError:
+                    continue
+                themes = []
+                for theme_name in theme_names:
+                    if not theme_name:
+                        continue
+                    tag, created = Tag.objects.get_or_create(slug=slughifi(theme_name), category='theme')
+                    if created:
+                        tag.name = theme_name
+                        tag.sort_key = theme_name.lower()
+                        tag.save()
+                    themes.append(tag)
+                if not themes:
+                    continue
+
+                text = fragment.to_string()
+                short_text = ''
+                if (len(MarkupString(text)) > 240):
+                    short_text = unicode(MarkupString(text)[:160])
+                new_fragment = Fragment.objects.create(anchor=fragment.id, book=self,
+                    text=text, short_text=short_text)
+
+                new_fragment.save()
+                new_fragment.tags = set(meta_tags + themes + [book_tag] + ancestor_tags)
+            self.save()
+            self.html_built.send(sender=self)
+            return True
+        return False
+
+    @staticmethod
+    def zip_format(format_):
+        def pretty_file_name(book):
+            return "%s/%s.%s" % (
+                b.get_extra_info_value()['author'],
+                b.slug,
+                format_)
+
+        field_name = "%s_file" % format_
+        books = Book.objects.filter(parent=None).exclude(**{field_name: ""})
+        paths = [(pretty_file_name(b), getattr(b, field_name).path)
+                    for b in books]
+        result = create_zip.delay(paths,
+                    getattr(settings, "ALL_%s_ZIP" % format_.upper()))
+        return result.wait()
+
+    def zip_audiobooks(self):
+        bm = BookMedia.objects.filter(book=self, type='mp3')
+        paths = map(lambda bm: (None, bm.file.path), bm)
+        result = create_zip.delay(paths, self.slug)
+        return result.wait()
+
+    def search_index(self):
+        idx = search.ReusableIndex()
+        idx.open()
+        try:
+            idx.index_book(self)
+        finally:
+            idx.close()
+
     @classmethod
     def from_xml_file(cls, xml_file, **kwargs):
         # use librarian to parse meta-data
@@ -492,11 +647,10 @@ class Book(models.Model):
             xml_file.close()
 
     @classmethod
-    def from_text_and_meta(cls, raw_file, book_info, overwrite=False, build_epub=True, build_txt=True):
+    def from_text_and_meta(cls, raw_file, book_info, overwrite=False,
+            build_epub=True, build_txt=True, build_pdf=True, build_mobi=True,
+            search_index=True):
         import re
-        from tempfile import NamedTemporaryFile
-        from markupstring import MarkupString
-        from django.core.files.storage import default_storage
 
         # check for parts before we do anything
         children = []
@@ -525,10 +679,9 @@ class Book(models.Model):
 
         book.title = book_info.title
         book.set_extra_info_value(book_info.to_dict())
-        book._short_html = ''
         book.save()
 
-        book_tags = []
+        meta_tags = []
         categories = (('kinds', 'kind'), ('genres', 'genre'), ('authors', 'author'), ('epochs', 'epoch'))
         for field_name, category in categories:
             try:
@@ -543,11 +696,11 @@ class Book(models.Model):
                 tag, created = Tag.objects.get_or_create(slug=slughifi(tag_name), category=category)
                 if created:
                     tag.name = tag_name
-                    tag.sort_key = tag_sort_key.lower()
+                    tag.sort_key = sortify(tag_sort_key.lower())
                     tag.save()
-                book_tags.append(tag)
+                meta_tags.append(tag)
 
-        book.tags = book_tags + book_shelves
+        book.tags = set(meta_tags + book_shelves)
 
         book_tag = book.book_tag()
 
@@ -562,52 +715,21 @@ class Book(models.Model):
         # delete old fragments when overwriting
         book.fragments.all().delete()
 
-        html_file = NamedTemporaryFile()
-        if html.transform(book.xml_file.path, html_file, parse_dublincore=False):
-            book.html_file.save('%s.html' % book.slug, File(html_file), save=False)
-
-            # get ancestor l-tags for adding to new fragments
-            ancestor_tags = []
-            p = book.parent
-            while p:
-                ancestor_tags.append(p.book_tag())
-                p = p.parent
-
-            # Extract fragments
-            closed_fragments, open_fragments = html.extract_fragments(book.html_file.path)
-            for fragment in closed_fragments.values():
-                try:
-                    theme_names = [s.strip() for s in fragment.themes.split(',')]
-                except AttributeError:
-                    continue
-                themes = []
-                for theme_name in theme_names:
-                    if not theme_name:
-                        continue
-                    tag, created = Tag.objects.get_or_create(slug=slughifi(theme_name), category='theme')
-                    if created:
-                        tag.name = theme_name
-                        tag.sort_key = theme_name.lower()
-                        tag.save()
-                    themes.append(tag)
-                if not themes:
-                    continue
+        if book.build_html():
+            if not settings.NO_BUILD_TXT and build_txt:
+                book.build_txt()
 
-                text = fragment.to_string()
-                short_text = ''
-                if (len(MarkupString(text)) > 240):
-                    short_text = unicode(MarkupString(text)[:160])
-                new_fragment, created = Fragment.objects.get_or_create(anchor=fragment.id, book=book,
-                    defaults={'text': text, 'short_text': short_text})
+        if not settings.NO_BUILD_EPUB and build_epub:
+            book.root_ancestor.build_epub()
 
-                new_fragment.save()
-                new_fragment.tags = set(book_tags + themes + [book_tag] + ancestor_tags)
+        if not settings.NO_BUILD_PDF and build_pdf:
+            book.root_ancestor.build_pdf()
 
-        if not settings.NO_BUILD_TXT and build_txt:
-            book.build_txt()
+        if not settings.NO_BUILD_MOBI and build_mobi:
+            book.build_mobi()
 
-        if not settings.NO_BUILD_EPUB and build_epub:
-            book.root_ancestor.build_epub()
+        if not settings.NO_SEARCH_INDEX and search_index:
+            book.search_index()
 
         book_descendants = list(book.children.all())
         # add l-tag to descendants and their fragments
@@ -620,57 +742,69 @@ class Book(models.Model):
                 fragment.tags = set(list(fragment.tags) + [book_tag])
             book_descendants += list(child_book.children.all())
 
+        book.save()
+
         # refresh cache
         book.reset_tag_counter()
         book.reset_theme_counter()
 
-        book.save()
         return book
 
-
-    def refresh_tag_counter(self):
-        tags = {}
-        for child in self.children.all().order_by():
-            for tag_pk, value in child.tag_counter.iteritems():
-                tags[tag_pk] = tags.get(tag_pk, 0) + value
-        for tag in self.tags.exclude(category__in=('book', 'theme', 'set')).order_by():
-            tags[tag.pk] = 1
-        self.set__tag_counter_value(tags)
-        self.save(reset_short_html=False)
-        return tags
-
     def reset_tag_counter(self):
-        self._tag_counter = None
-        self.save(reset_short_html=False)
+        if self.id is None:
+            return
+
+        cache_key = "Book.tag_counter/%d" % self.id
+        cache.delete(cache_key)
         if self.parent:
             self.parent.reset_tag_counter()
 
     @property
     def tag_counter(self):
-        if self._tag_counter is None:
-            return self.refresh_tag_counter()
-        return dict((int(k), v) for k, v in self.get__tag_counter_value().iteritems())
-
-    def refresh_theme_counter(self):
-        tags = {}
-        for fragment in Fragment.tagged.with_any([self.book_tag()]).order_by():
-            for tag in fragment.tags.filter(category='theme').order_by():
-                tags[tag.pk] = tags.get(tag.pk, 0) + 1
-        self.set__theme_counter_value(tags)
-        self.save(reset_short_html=False)
+        if self.id:
+            cache_key = "Book.tag_counter/%d" % self.id
+            tags = cache.get(cache_key)
+        else:
+            tags = None
+
+        if tags is None:
+            tags = {}
+            for child in self.children.all().order_by():
+                for tag_pk, value in child.tag_counter.iteritems():
+                    tags[tag_pk] = tags.get(tag_pk, 0) + value
+            for tag in self.tags.exclude(category__in=('book', 'theme', 'set')).order_by():
+                tags[tag.pk] = 1
+
+            if self.id:
+                cache.set(cache_key, tags, CACHE_FOREVER)
         return tags
 
     def reset_theme_counter(self):
-        self._theme_counter = None
-        self.save(reset_short_html=False)
+        if self.id is None:
+            return
+
+        cache_key = "Book.theme_counter/%d" % self.id
+        cache.delete(cache_key)
         if self.parent:
             self.parent.reset_theme_counter()
 
     @property
     def theme_counter(self):
-        if self._theme_counter is None:
-            return self.refresh_theme_counter()
-        return dict((int(k), v) for k, v in self.get__theme_counter_value().iteritems())
+        if self.id:
+            cache_key = "Book.theme_counter/%d" % self.id
+            tags = cache.get(cache_key)
+        else:
+            tags = None
+
+        if tags is None:
+            tags = {}
+            for fragment in Fragment.tagged.with_any([self.book_tag()]).order_by():
+                for tag in fragment.tags.filter(category='theme').order_by():
+                    tags[tag.pk] = tags.get(tag.pk, 0) + 1
+
+            if self.id:
+                cache.set(cache_key, tags, CACHE_FOREVER)
+        return tags
 
     def pretty_title(self, html_links=False):
         book = self
@@ -689,11 +823,46 @@ class Book(models.Model):
 
         return ', '.join(names)
 
+    @classmethod
+    def tagged_top_level(cls, tags):
+        """ Returns top-level books tagged with `tags'.
+
+        It only returns those books which don't have ancestors which are
+        also tagged with those tags.
+
+        """
+        # get relevant books and their tags
+        objects = cls.tagged.with_all(tags)
+        # eliminate descendants
+        l_tags = Tag.objects.filter(category='book', slug__in=[book.book_tag_slug() for book in objects])
+        descendants_keys = [book.pk for book in cls.tagged.with_any(l_tags)]
+        if descendants_keys:
+            objects = objects.exclude(pk__in=descendants_keys)
+
+        return objects
+
+
+def _has_factory(ftype):
+    has = lambda self: bool(getattr(self, "%s_file" % ftype))
+    has.short_description = t.upper()
+    has.boolean = True
+    has.__name__ = "has_%s_file" % ftype
+    return has
+
+    
+# add the file fields
+for t in Book.file_types:
+    field_name = "%s_file" % t
+    models.FileField(_("%s file" % t.upper()),
+            upload_to=book_upload_path(t),
+            blank=True).contribute_to_class(Book, field_name)
+
+    setattr(Book, "has_%s_file" % t, _has_factory(t))
+
 
 class Fragment(models.Model):
     text = models.TextField()
     short_text = models.TextField(editable=False)
-    _short_html = models.TextField(editable=False)
     anchor = models.CharField(max_length=120)
     book = models.ForeignKey(Book, related_name='fragments')
 
@@ -709,16 +878,29 @@ class Fragment(models.Model):
     def get_absolute_url(self):
         return '%s#m%s' % (reverse('book_text', kwargs={'slug': self.book.slug}), self.anchor)
 
+    def reset_short_html(self):
+        if self.id is None:
+            return
+
+        cache_key = "Fragment.short_html/%d/%s"
+        for lang, langname in settings.LANGUAGES:
+            cache.delete(cache_key % (self.id, lang))
+
     def short_html(self):
-        key = '_short_html_%s' % get_language()
-        short_html = getattr(self, key)
-        if short_html and len(short_html):
+        if self.id:
+            cache_key = "Fragment.short_html/%d/%s" % (self.id, get_language())
+            short_html = cache.get(cache_key)
+        else:
+            short_html = None
+
+        if short_html is not None:
             return mark_safe(short_html)
         else:
-            setattr(self, key, unicode(render_to_string('catalogue/fragment_short.html',
-                {'fragment': self})))
-            self.save()
-            return mark_safe(getattr(self, key))
+            short_html = unicode(render_to_string('catalogue/fragment_short.html',
+                {'fragment': self}))
+            if self.id:
+                cache.set(cache_key, short_html, CACHE_FOREVER)
+            return mark_safe(short_html)
 
 
 class FileRecord(models.Model):
@@ -744,7 +926,8 @@ class FileRecord(models.Model):
 
 def _tags_updated_handler(sender, affected_tags, **kwargs):
     # reset tag global counter
-    Tag.objects.filter(pk__in=[tag.pk for tag in affected_tags]).update(book_count=None)
+    # we want Tag.changed_at updated for API to know the tag was touched
+    Tag.objects.filter(pk__in=[tag.pk for tag in affected_tags]).update(book_count=None, changed_at=datetime.now())
 
     # if book tags changed, reset book tag counter
     if isinstance(sender, Book) and \
@@ -759,23 +942,14 @@ def _tags_updated_handler(sender, affected_tags, **kwargs):
 tags_updated.connect(_tags_updated_handler)
 
 
-def _m2m_changed_handler(sender, instance, action, reverse, pk_set, **kwargs):
-    """ refresh all the short_html stuff on BookMedia delete """
-    if sender == Book.medias.through and reverse and action == 'pre_clear':
-        for book in instance.book_set.all():
-            book.save()
-m2m_changed.connect(_m2m_changed_handler)
-
 def _pre_delete_handler(sender, instance, **kwargs):
-    """ explicitly clear m2m, so that Books can be refreshed """
+    """ refresh Book on BookMedia delete """
     if sender == BookMedia:
-        instance.book_set.clear()
+        instance.book.save()
 pre_delete.connect(_pre_delete_handler)
 
 def _post_save_handler(sender, instance, **kwargs):
     """ refresh all the short_html stuff on BookMedia update """
     if sender == BookMedia:
-        for book in instance.book_set.all():
-            book.save()
+        instance.book.save()
 post_save.connect(_post_save_handler)
-