epub uploading

pull/30/head
Pavel Yakunin 9 years ago
parent dc5074a865
commit 18e341d650

@ -9,19 +9,33 @@ try:
from wand.image import Image
use_generic_pdf_cover = False
except ImportError, e:
logger.warning('cannot import Image, generating pdf covers for pdf uploads will not work')
logger.warning('cannot import Image, generating pdf covers for pdf uploads will not work: %s', e)
use_generic_pdf_cover = True
try:
from PyPDF2 import PdfFileReader
use_pdf_meta = True
except ImportError, e:
logger.warning('cannot import PyPDF2, extracting pdf metadata will not work')
logger.warning('cannot import PyPDF2, extracting pdf metadata will not work: %s', e)
use_pdf_meta = False
try:
import epub
use_epub_meta = True
except ImportError, e:
logger.warning('cannot import PyPDF2, extracting pdf metadata will not work: %s', e)
use_epub_meta = False
def process(tmp_file_path, original_file_name, original_file_extension):
if (".PDF" == original_file_extension.upper()):
try:
if ".PDF" == original_file_extension.upper():
return pdf_meta(tmp_file_path, original_file_name, original_file_extension)
else: return default_meta(tmp_file_path, original_file_name, original_file_extension)
if ".EPUB" == original_file_extension.upper() and use_pdf_meta == True:
return epub.get_epub_info(tmp_file_path, original_file_name, original_file_extension)
except Exception, e:
logger.warning('cannot parse metadata, using default: %s', e)
return default_meta(tmp_file_path, original_file_name, original_file_extension)

@ -0,0 +1,67 @@
import zipfile
from lxml import etree
import os
import uploader
def extractCover(zip, coverFile, tmp_file_name):
if (coverFile is None):
return None
else:
cf = zip.read("OPS/" + coverFile)
prefix = os.path.splitext(tmp_file_name)[0]
tmp_cover_name = prefix + "." + coverFile
image = open(tmp_cover_name, 'wb')
image.write(cf)
image.close()
return tmp_cover_name
def get_epub_info(tmp_file_path, original_file_name, original_file_extension):
ns = {
'n':'urn:oasis:names:tc:opendocument:xmlns:container',
'pkg':'http://www.idpf.org/2007/opf',
'dc':'http://purl.org/dc/elements/1.1/'
}
zip = zipfile.ZipFile(tmp_file_path)
txt = zip.read('META-INF/container.xml')
tree = etree.fromstring(txt)
cfname = tree.xpath('n:rootfiles/n:rootfile/@full-path',namespaces=ns)[0]
cf = zip.read(cfname)
tree = etree.fromstring(cf)
p = tree.xpath('/pkg:package/pkg:metadata',namespaces=ns)[0]
epub_metadata = {}
for s in ['title', 'description', 'creator']:
tmp = p.xpath('dc:%s/text()'%(s),namespaces=ns)
if (len(tmp) > 0):
epub_metadata[s] = p.xpath('dc:%s/text()'%(s),namespaces=ns)[0]
else:
epub_metadata[s] = "Unknown"
coversection = tree.xpath("/pkg:package/pkg:manifest/pkg:item[@id='cover']/@href",namespaces=ns)
if (len(coversection) > 0):
coverfile = extractCover(zip, coversection[0], tmp_file_path)
else:
coverfile = None
if epub_metadata['title'] is None:
title = original_file_name
else:
title = epub_metadata['title']
return uploader.BookMeta(
file_path = tmp_file_path,
extension = original_file_extension,
title = title,
author = epub_metadata['creator'],
cover = coverfile,
description = epub_metadata['description'],
tags = "",
series = "",
series_id="")
Loading…
Cancel
Save