autocat3/CoverPages.py

105 lines
3.0 KiB
Python
Raw Permalink Normal View History

2019-03-28 13:45:03 +00:00
#!/usr/bin/env python
# -*- mode: python; indent-tabs-mode: nil; -*- coding: utf-8 -*-
"""
CoverPages.py
Copyright 2009-2010 by Marcello Perathoner
Distributable under the GNU General Public License Version 3 or newer.
Serve cover images of most popular and latest ebooks.
"""
from __future__ import unicode_literals
import cherrypy
import six
from libgutenberg import GutenbergGlobals as gg
import BaseSearcher
class CoverPages (object):
""" Output a gallery of cover pages. """
orders = { 'latest': 'release_date',
'popular': 'downloads' }
@staticmethod
def serve (rows, size):
""" Output a gallery of coverpages. """
cherrypy.response.headers['Content-Type'] = 'text/html; charset=utf-8'
cherrypy.response.headers['Content-Language'] = 'en'
s = """<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN"
"http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
<html xmlns="http://www.w3.org/1999/xhtml" lang="en" xml:lang="en" xml:base="http://www.gutenberg.org">
<head>
<title>Cover Flow</title>
<!--<style>
2019-03-28 13:45:03 +00:00
.cover-thumb {
display: inline-block;
background-position: center;
background-repeat: no-repeat;
}
.cover-thumb-small {
width: 76px;
height: 110px;
}
.cover-thumb-medium {
width: 210px;
height: 310px;
}
</style>-->
2019-03-28 13:45:03 +00:00
</head>
<body><div>"""
for row in rows:
url = '/' + row.filename
href = '/ebooks/%d' % row.pk
title = gg.xmlspecialchars (row.title)
title = title.replace ('"', '&quot;')
s += """<a href="{href}"
title="{title}"
class="cover-thumb cover-thumb-{size}" target="_top"
style="background-image: url({url})"> </a>\n""".format (
2019-03-28 13:45:03 +00:00
url = url, href = href, title = title, size = size)
2019-03-28 13:45:03 +00:00
return (s + '</div></body></html>\n').encode ('utf-8')
def index (self, count, size, order, **kwargs):
""" Internal help function. """
try:
count = int (count)
if count < 1:
raise ValueError ('count < 0')
if size not in ('small', 'medium'):
raise ValueError ('bogus size')
order = 'books.%s' % self.orders[order]
rows = BaseSearcher.SQLSearcher.execute (
"""SELECT files.filename, books.pk, books.title FROM files, books
WHERE files.fk_books = books.pk
AND files.diskstatus = 0
AND files.fk_filetypes = %%(size)s
ORDER BY %s DESC
OFFSET 1 LIMIT %%(count)s -- %s""" % (order, cherrypy.request.remote.ip),
{ 'count': count,
'size': 'cover.%s' % size,
})
if rows:
return self.serve (rows, size)
except (ValueError, KeyError) as what:
raise cherrypy.HTTPError (400, 'Bad Request. %s' % six.text_type (what))
except IOError:
pass
raise cherrypy.HTTPError (500, 'Internal Server Error.')