Edit Book: Fix various problems when editing EPUB files with non-ascii filenames on OS X. Fixes #1317883 [calibre should care of 'decomposed UTF-8' filenames on Darwin platform](https://bugs.launchpad.net/calibre/+bug/1317883)

This commit is contained in:
Kovid Goyal 2014-06-22 11:38:30 +05:30
parent cad2a50df9
commit 78be882aff
2 changed files with 18 additions and 19 deletions

View File

@ -17,9 +17,7 @@ from lxml import etree
from cssutils import replaceUrls, getUrls from cssutils import replaceUrls, getUrls
from calibre import CurrentDir from calibre import CurrentDir
from calibre.constants import isosx from calibre.customize.ui import (plugin_for_input_format, plugin_for_output_format)
from calibre.customize.ui import (plugin_for_input_format,
plugin_for_output_format)
from calibre.ebooks.chardet import xml_to_unicode from calibre.ebooks.chardet import xml_to_unicode
from calibre.ebooks.conversion.plugins.epub_input import ( from calibre.ebooks.conversion.plugins.epub_input import (
ADOBE_OBFUSCATION, IDPF_OBFUSCATION, decrypt_font_data) ADOBE_OBFUSCATION, IDPF_OBFUSCATION, decrypt_font_data)
@ -126,16 +124,14 @@ class Container(object): # {{{
for f in filenames: for f in filenames:
path = join(dirpath, f) path = join(dirpath, f)
name = self.abspath_to_name(path) name = self.abspath_to_name(path)
if isosx: # OS X silently changes all file names to NFD form. The EPUB
# OS X silently changes all file names to NFD form. The # spec requires all text including filenames to be in NFC form.
# EPUB spec requires all text including filenames to be in # The proper fix is to implement a VFS that maps between
# NFC form. The proper fix is to implement a VFS that maps # canonical names and their file system representation, however,
# between canonical names and their filesystem # I dont have the time for that now. Note that the container
# representation, however, I dont have the time for that # ensures that all text files are normalized to NFC when
# now, so this will at least fix the problem for books that # decoding them anyway, so there should be no mismatch between
# properly use the NFC form. Books that use the NFD form # names in the text and NFC canonical file names.
# will be broken by this, but that's the price you pay for
# using OS X.
name = unicodedata.normalize('NFC', name) name = unicodedata.normalize('NFC', name)
self.name_path_map[name] = path self.name_path_map[name] = path
self.mime_map[name] = guess_type(path) self.mime_map[name] = guess_type(path)
@ -373,7 +369,7 @@ class Container(object): # {{{
abspath_to_name() for that.''' abspath_to_name() for that.'''
return relpath(path, base or self.root) return relpath(path, base or self.root)
def decode(self, data): def decode(self, data, normalize_to_nfc=True):
"""Automatically decode :param:`data` into a `unicode` object.""" """Automatically decode :param:`data` into a `unicode` object."""
def fix_data(d): def fix_data(d):
return d.replace('\r\n', '\n').replace('\r', '\n') return d.replace('\r\n', '\n').replace('\r', '\n')
@ -402,6 +398,8 @@ class Container(object): # {{{
except UnicodeDecodeError: except UnicodeDecodeError:
pass pass
data, self.used_encoding = xml_to_unicode(data) data, self.used_encoding = xml_to_unicode(data)
if normalize_to_nfc:
data = unicodedata.normalize('NFC', data)
return fix_data(data) return fix_data(data)
def ok_to_be_unmanifested(self, name): def ok_to_be_unmanifested(self, name):
@ -422,6 +420,7 @@ class Container(object): # {{{
def parse_xml(self, data): def parse_xml(self, data):
data, self.used_encoding = xml_to_unicode( data, self.used_encoding = xml_to_unicode(
data, strip_encoding_pats=True, assume_utf8=True, resolve_entities=True) data, strip_encoding_pats=True, assume_utf8=True, resolve_entities=True)
data = unicodedata.normalize('NFC', data)
return etree.fromstring(data, parser=RECOVER_PARSER) return etree.fromstring(data, parser=RECOVER_PARSER)
def parse_xhtml(self, data, fname='<string>', force_html5_parse=False): def parse_xhtml(self, data, fname='<string>', force_html5_parse=False):
@ -447,11 +446,11 @@ class Container(object): # {{{
data = self.parse_css(data, self.relpath(path)) data = self.parse_css(data, self.relpath(path))
return data return data
def raw_data(self, name, decode=True): def raw_data(self, name, decode=True, normalize_to_nfc=True):
ans = self.open(name).read() ans = self.open(name).read()
mime = self.mime_map.get(name, guess_type(name)) mime = self.mime_map.get(name, guess_type(name))
if decode and (mime in OEB_STYLES or mime in OEB_DOCS or mime == 'text/plain' or mime[-4:] in {'+xml', '/xml'}): if decode and (mime in OEB_STYLES or mime in OEB_DOCS or mime == 'text/plain' or mime[-4:] in {'+xml', '/xml'}):
ans = self.decode(ans) ans = self.decode(ans, normalize_to_nfc=normalize_to_nfc)
return ans return ans
def parse_css(self, data, fname='<string>', is_declaration=False): def parse_css(self, data, fname='<string>', is_declaration=False):

View File

@ -7,7 +7,7 @@ __license__ = 'GPL v3'
__copyright__ = '2012, Kovid Goyal <kovid@kovidgoyal.net>' __copyright__ = '2012, Kovid Goyal <kovid@kovidgoyal.net>'
__docformat__ = 'restructuredtext en' __docformat__ = 'restructuredtext en'
import sys, os, shlex, subprocess, shutil import sys, os, shlex, subprocess, shutil, unicodedata
from calibre import prints, as_unicode, walk from calibre import prints, as_unicode, walk
from calibre.constants import iswindows, __appname__ from calibre.constants import iswindows, __appname__
@ -67,7 +67,7 @@ def zip_rebuilder(tdir, path):
if fn in exclude_files: if fn in exclude_files:
continue continue
absfn = os.path.join(root, fn) absfn = os.path.join(root, fn)
zfn = os.path.relpath(absfn, tdir).replace(os.sep, '/') zfn = unicodedata.normalize('NFC', os.path.relpath(absfn, tdir).replace(os.sep, '/'))
zf.write(absfn, zfn) zf.write(absfn, zfn)
def get_tools(fmt): def get_tools(fmt):