Fix #1135 (EPUB coversion error)

src/calibre/ebooks/epub/from_any.py

@@ -13,7 +13,7 @@ from contextlib import nested
 from calibre import extract, walk
 from calibre.ebooks import DRMError
 from calibre.ebooks.epub import config as common_config
-from calibre.ebooks.epub.from_html import convert as html2epub
+from calibre.ebooks.epub.from_html import convert as html2epub, find_html_index
 from calibre.ptempfile import TemporaryDirectory
 from calibre.ebooks.metadata import MetaInformation
 from calibre.ebooks.metadata.opf2 import OPFCreator
@@ -103,18 +103,7 @@ def unarchive(path, tdir):
                 if ext in ['txt', 'rtf'] and os.stat(f).st_size < 2048:
                     continue
                 return f, ext
-    html_pat = re.compile(r'\.(x){0,1}htm(l){0,1}$', re.IGNORECASE)
-    html_files = [f for f in files if html_pat.search(f) is not None]
-    if not html_files:
-        raise ValueError(_('Could not find an ebook inside the archive'))
-    html_files = [(f, os.stat(f).st_size) for f in html_files]
-    html_files.sort(cmp = lambda x, y: cmp(x[1], y[1]))
-    html_files = [f[0] for f in html_files]
-    for q in ('toc', 'index'):
-        for f in html_files:
-            if os.path.splitext(f)[0].lower() == q:
-                return f, os.path.splitext(f)[1].lower()[1:]
-    return html_files[-1], os.path.splitext(html_files[-1])[1].lower()[1:]
+    return find_html_index(files)
 
 def any2epub(opts, path, notification=None):
     ext = os.path.splitext(path)[1]

src/calibre/ebooks/epub/from_html.py

@@ -32,7 +32,7 @@ Conversion of HTML/OPF files follows several stages:
     * The EPUB container is created.
 '''
 
-import os, sys, cStringIO, logging
+import os, sys, cStringIO, logging, re
 
 from lxml.etree import XPath
 try:
@@ -51,7 +51,25 @@ from calibre.ebooks.epub import initialize_container, PROFILES
 from calibre.ebooks.epub.split import split
 from calibre.ebooks.epub.fonts import Rationalizer
 from calibre.constants import preferred_encoding
+from calibre import walk
 
+def find_html_index(files):
+    '''
+    Given a list of files, find the most likely root HTML file in the
+    list.
+    '''
+    html_pat = re.compile(r'\.(x){0,1}htm(l){0,1}$', re.IGNORECASE)
+    html_files = [f for f in files if html_pat.search(f) is not None]
+    if not html_files:
+        raise ValueError(_('Could not find an ebook inside the archive'))
+    html_files = [(f, os.stat(f).st_size) for f in html_files]
+    html_files.sort(cmp = lambda x, y: cmp(x[1], y[1]))
+    html_files = [f[0] for f in html_files]
+    for q in ('toc', 'index'):
+        for f in html_files:
+            if os.path.splitext(f)[0].lower() == q:
+                return f, os.path.splitext(f)[1].lower()[1:]
+    return html_files[-1], os.path.splitext(html_files[-1])[1].lower()[1:]
 
 class HTMLProcessor(Processor, Rationalizer):
     
@@ -203,6 +221,10 @@ def convert(htmlfile, opts, notification=None):
     if htmlfile.lower().endswith('.opf'):
         opf = OPF(htmlfile, os.path.dirname(os.path.abspath(htmlfile)))
         filelist = opf_traverse(opf, verbose=opts.verbose, encoding=opts.encoding)
+        if not filelist:
+            # Bad OPF look for a HTML file instead
+            htmlfile = find_html_index(walk(os.path.dirname(htmlfile)))[0]
+            filelist = get_filelist(htmlfile, opts)[1]
         mi = MetaInformation(opf)
     else:
         opf, filelist = get_filelist(htmlfile, opts)

src/calibre/ebooks/html.py

@@ -304,8 +304,14 @@ def opf_traverse(opf_reader, verbose=0, encoding=None):
                 flat[i] = path
                 for item in opf_reader.itermanifest():
                     item.set('href', item.get('href').replace('&', '%26'))
-    flat = [HTMLFile(path, 0, encoding, verbose) for path in flat]
-    return [f for f in flat if not f.is_binary]
+    ans = []
+    for path in flat:
+        if os.path.exists(path):
+            ans.append(HTMLFile(path, 0, encoding, verbose))
+        else:
+            print 'WARNING: OPF spine item %s does not exist'%path
+    ans = [f for f in ans if not f.is_binary]
+    return ans