openmedialibrary/oml/media/epub.py
2016-01-06 18:40:23 +05:30

149 lines
5.5 KiB
Python

# -*- coding: utf-8 -*-
# vi:si:et:sw=4:sts=4:ts=4
import os
import xml.etree.ElementTree as ET
import zipfile
from io import BytesIO
import re
from urllib.parse import unquote
from PIL import Image
import stdnum.isbn
from ox import strip_tags, decode_html
from utils import normalize_isbn, find_isbns, get_language
import logging
logger = logging.getLogger(__name__)
def cover(path):
logger.debug('cover %s', path)
data = None
try:
z = zipfile.ZipFile(path)
except zipfile.BadZipFile:
logger.debug('invalid epub file %s', path)
return data
def use(filename):
logger.debug('using %s', filename)
return z.read(filename)
for f in z.filelist:
if 'cover' in f.filename.lower() and f.filename.split('.')[-1] in ('jpg', 'jpeg', 'png'):
return use(f.filename)
files = [f.filename for f in z.filelist]
opf = [f for f in files if f.endswith('opf')]
if opf:
#logger.debug('opf: %s', z.read(opf[0]).decode())
info = ET.fromstring(z.read(opf[0]))
metadata = info.findall('{http://www.idpf.org/2007/opf}metadata')
if metadata:
metadata = metadata[0]
manifest = info.findall('{http://www.idpf.org/2007/opf}manifest')
if manifest:
manifest = manifest[0]
if metadata and manifest:
for e in metadata.getchildren():
if e.tag == '{http://www.idpf.org/2007/opf}meta' and e.attrib.get('name') == 'cover':
cover_id = e.attrib['content']
for e in manifest.getchildren():
if e.attrib['id'] == cover_id:
filename = unquote(e.attrib['href'])
filename = os.path.normpath(os.path.join(os.path.dirname(opf[0]), filename))
if filename in files:
return use(filename)
if manifest:
images = [e for e in manifest.getchildren() if 'image' in e.attrib['media-type']]
if images:
image_data = []
for e in images:
filename = unquote(e.attrib['href'])
filename = os.path.normpath(os.path.join(os.path.dirname(opf[0]), filename))
if filename in files:
image_data.append((filename, z.read(filename)))
if image_data:
image_data.sort(key=lambda i: len(i[1]))
data = image_data[-1][1]
logger.debug('using %s', image_data[-1][0])
return data
for e in manifest.getchildren():
if 'html' in e.attrib['media-type']:
filename = unquote(e.attrib['href'])
filename = os.path.normpath(os.path.join(os.path.dirname(opf[0]), filename))
html = z.read(filename).decode('utf-8', 'ignore')
img = re.compile('<img.*?src="(.*?)"').findall(html)
#svg image
img += re.compile('<image.*?href="(.*?)"').findall(html)
if img:
img = unquote(img[0])
img = os.path.normpath(os.path.join(os.path.dirname(filename), img))
if img in files:
return use(img)
# fallback return black cover
img = Image.new('RGB', (80, 128))
o = BytesIO()
img.save(o, format='jpeg')
data = o.getvalue()
o.close()
return data
def info(epub):
data = {}
try:
z = zipfile.ZipFile(epub)
except zipfile.BadZipFile:
logger.debug('invalid epub file %s', epub)
return data
opf = [f.filename for f in z.filelist if f.filename.endswith('opf')]
if opf:
info = ET.fromstring(z.read(opf[0]))
metadata = info.findall('{http://www.idpf.org/2007/opf}metadata')
if metadata:
metadata = metadata[0]
for e in metadata.getchildren():
if e.text and e.text.strip() and e.text not in ('unknown', 'none'):
key = e.tag.split('}')[-1]
key = {
'creator': 'author',
}.get(key, key)
value = e.text.strip()
if key == 'identifier':
value = normalize_isbn(value)
if stdnum.isbn.is_valid(value):
data['isbn'] = [value]
elif key == 'author':
data[key] = value.split(', ')
else:
data[key] = value
if 'description' in data:
data['description'] = strip_tags(decode_html(data['description']))
text = extract_text(epub)
data['textsize'] = len(text)
if not 'isbn' in data:
isbn = extract_isbn(text)
if isbn:
data['isbn'] = [isbn]
if 'date' in data and 'T' in data['date']:
data['date'] = data['date'].split('T')[0]
if 'language' in data and isinstance(data['language'], str):
data['language'] = get_language(data['language'])
return data
def extract_text(path):
data = ''
z = zipfile.ZipFile(path)
for f in z.filelist:
if '/._' in f.filename or f.filename.startswith('._'):
continue
if f.filename.endswith('html'):
data += z.read(f.filename).decode('utf-8', 'ignore')
return data
def extract_isbn(data):
isbns = find_isbns(data)
if isbns:
return isbns[0]