Replace Calibre path resolve system for some of the modules.

In Calibre there is a module, which registers two functions in global
scope, one for filepath and one for images path. It relies on additional
three variables, registered during application startup. I found it
overcomplicated, and couple of the resource paths has been calculated
using pkg_resource module.

ebook_converter/__init__.py

@@ -4,6 +4,7 @@ __copyright__ = '2008, Kovid Goyal <kovid@kovidgoyal.net>'
 __docformat__ = 'restructuredtext en'
 
 import sys, os, re, time, random, warnings
+import pkg_resources
 from ebook_converter.polyglot.builtins import codepoint_to_chr, unicode_type, range, hasenv, native_string_type
 from math import floor
 from functools import partial
@@ -34,7 +35,8 @@ _mt_inited = False
 def _init_mimetypes():
     global _mt_inited
     import mimetypes
-    mimetypes.init([P('mime.types')])
+    mimetypes.init([pkg_resources.resource_filename('ebook_converter',
+                                                    'data/mime.types')])
     _mt_inited = True
 
 

ebook_converter/ebooks/conversion/plugins/rtf_input.py

@@ -3,6 +3,7 @@ __license__   = 'GPL v3'
 __copyright__ = '2008, Kovid Goyal <kovid at kovidgoyal.net>'
 
 import os, glob, re, textwrap
+import pkg_resources
 
 from ebook_converter.customize.conversion import InputFormatPlugin, OptionRecommendation
 from ebook_converter.polyglot.builtins import iteritems, filter, getcwd, as_bytes
@@ -282,7 +283,9 @@ class RTFInput(InputFormatPlugin):
 
         self.log('Converting XML to HTML...')
         inline_class = InlineClass(self.log)
-        styledoc = safe_xml_fromstring(P('templates/rtf.xsl', data=True), recover=False)
+        with open(pkg_resources.resource_filename('ebook_converter',
+                                                  'data/rtf.xsl')) as fobj:
+            styledoc = safe_xml_fromstring(fobj.read())
         extensions = {('calibre', 'inline-class') : inline_class}
         transform = etree.XSLT(styledoc, extensions=extensions)
         result = transform(doc)

ebook_converter/ebooks/oeb/stylizer.py

@@ -9,6 +9,7 @@ __license__   = 'GPL v3'
 __copyright__ = '2008, Marshall T. Vandegrift <llasram@gmail.com>'
 
 import os, re, logging, copy, unicodedata, numbers
+import pkg_resources
 from operator import itemgetter
 from weakref import WeakKeyDictionary
 from xml.dom import SyntaxErr as CSSSyntaxError
@@ -32,7 +33,8 @@ _html_css_stylesheet = None
 def html_css_stylesheet():
     global _html_css_stylesheet
     if _html_css_stylesheet is None:
-        with open(P('templates/html.css'), 'rb') as f:
+        with open(pkg_resources.resource_filename('ebook_converter',
+                                                  'data/html.css'), 'rb') as f:
             html_css = f.read().decode('utf-8')
         _html_css_stylesheet = parseString(html_css, validate=False)
     return _html_css_stylesheet

ebook_converter/utils/localization.py

@@ -3,7 +3,9 @@ __copyright__ = '2009, Kovid Goyal <kovid@kovidgoyal.net>'
 __docformat__ = 'restructuredtext en'
 
 import os, locale, re, io, sys
+import json
 from gettext import GNUTranslations, NullTranslations
+import pkg_resources
 
 from ebook_converter.polyglot.builtins import is_py3, iteritems, unicode_type
 
@@ -361,12 +363,53 @@ for k in _extra_lang_codes:
 
 def _load_iso639():
     global _iso639
+
+    # NOTE(gryf): msgpacked data was originally added for speed purposes. In
+    # my tests, I cannot see any speed gain either on python2 or python3. It
+    # is even slower (around 4-8 times), than just using code below (which is
+    # excerpt form Calibre transform code which is executed during Calibre
+    # build).
     if _iso639 is None:
-        ip = P('localization/iso639.calibre_msgpack', allow_user_override=False, data=True)
+        src = pkg_resources.resource_filename('ebook_converter',
-        from ebook_converter.utils.serialize import msgpack_loads
+                                              'data/iso_639-3.json')
-        _iso639 = msgpack_loads(ip)
+
-        if 'by_3' not in _iso639:
+        with open(src, 'rb') as f:
-            _iso639['by_3'] = _iso639['by_3t']
+            root = json.load(f)
+
+        entries = root['639-3']
+        by_2 = {}
+        by_3 = {}
+        m2to3 = {}
+        m3to2 = {}
+        nm = {}
+        codes2, codes3 = set(), set()
+        for x in entries:
+            two = x.get('alpha_2')
+            threeb = x.get('alpha_3')
+            if threeb is None:
+                continue
+            name = x.get('inverted_name') or x.get('name')
+            if not name or name[0] in '!~=/\'"':
+                continue
+
+            if two is not None:
+                by_2[two] = name
+                codes2.add(two)
+                m2to3[two] = threeb
+                m3to2[threeb] = two
+            codes3.add(threeb)
+            by_3[threeb] = name
+            base_name = name.lower()
+            nm[base_name] = threeb
+
+        _iso639 = {'by_2': by_2,
+                   'by_3': by_3,
+                   'codes2': codes2,
+                   'codes3': codes3,
+                   '2to3': m2to3,
+                   '3to2': m3to2,
+                   'name_map': nm}
+
     return _iso639