HTML Input: Handle @import directives in linked css files. Fixes #5135 (style import error during HTML conversion with ebook-convert.exe)

src/calibre/ebooks/conversion/preprocess.py

@@ -107,9 +107,21 @@ class CSSPreProcessor(object):
 
     PAGE_PAT   = re.compile(r'@page[^{]*?{[^}]*?}')
 
-    def __call__(self, data):
+    def __call__(self, data, add_namespace=False):
+        from calibre.ebooks.oeb.base import XHTML_CSS_NAMESPACE
         data = self.PAGE_PAT.sub('', data)
-        return data
+        if not add_namespace:
+            return data
+        ans, namespaced = [], False
+        for line in data.splitlines():
+            ll = line.lstrip()
+            if not (namespaced or ll.startswith('@import') or
+                        ll.startswith('@charset')):
+                ans.append(XHTML_CSS_NAMESPACE.strip())
+                namespaced = True
+            ans.append(line)
+
+        return u'\n'.join(ans)
 
 class HTMLPreProcessor(object):
 

src/calibre/ebooks/html/input.py

@@ -312,6 +312,7 @@ class HTMLInput(InputFormatPlugin):
             xpath
         from calibre import guess_type
         import cssutils
+        self.OEB_STYLES = OEB_STYLES
         oeb = create_oebbook(log, None, opts, self,
                 encoding=opts.input_encoding, populate=False)
         self.oeb = oeb
@@ -376,7 +377,7 @@ class HTMLInput(InputFormatPlugin):
             rewrite_links(item.data, partial(self.resource_adder, base=dpath))
 
         for item in oeb.manifest.values():
-            if item.media_type in OEB_STYLES:
+            if item.media_type in self.OEB_STYLES:
                 dpath = None
                 for path, href in self.added_resources.items():
                     if href == item.href:
@@ -414,25 +415,30 @@ class HTMLInput(InputFormatPlugin):
         oeb.container = DirContainer(os.getcwdu(), oeb.log)
         return oeb
 
-
-    def resource_adder(self, link_, base=None):
+    def link_to_local_path(self, link_, base=None):
         if not isinstance(link_, unicode):
             try:
                 link_ = link_.decode('utf-8', 'error')
             except:
                 self.log.warn('Failed to decode link %r. Ignoring'%link_)
-                return link_
+                return None, None
         try:
-            l = Link(link_, base if base else os.path.getcwdu())
+            l = Link(link_, base if base else os.getcwdu())
         except:
             self.log.exception('Failed to process link: %r'%link_)
-            return link_
+            return None, None
         if l.path is None:
             # Not a local resource
-            return link_
+            return None, None
         link = l.path.replace('/', os.sep).strip()
         frag = l.fragment
         if not link:
+            return None, None
+        return link, frag
+
+    def resource_adder(self, link_, base=None):
+        link, frag = self.link_to_local_path(link_, base=base)
+        if link is None:
             return link_
         try:
             if base and not os.path.isabs(link):
@@ -460,6 +466,9 @@ class HTMLInput(InputFormatPlugin):
 
             item = self.oeb.manifest.add(id, href, media_type)
             item.html_input_href = bhref
+            if guessed in self.OEB_STYLES:
+                item.override_css_fetch = partial(
+                        self.css_import_handler, os.path.dirname(link))
             item.data
             self.added_resources[link] = href
 
@@ -468,7 +477,17 @@ class HTMLInput(InputFormatPlugin):
             nlink = '#'.join((nlink, frag))
         return nlink
 
-
+    def css_import_handler(self, base, href):
+        link, frag = self.link_to_local_path(href, base=base)
+        if link is None or not os.access(link, os.R_OK) or os.path.isdir(link):
+            return (None, None)
+        try:
+            raw = open(link, 'rb').read().decode('utf-8', 'replace')
+            raw = self.oeb.css_preprocessor(raw, add_namespace=True)
+        except:
+            self.log.exception('Failed to read CSS file: %r'%link)
+            return (None, None)
+        return (None, raw)
 
 
 

src/calibre/ebooks/oeb/base.py

@@ -17,6 +17,7 @@ from urlparse import urljoin
 
 from lxml import etree, html
 from cssutils import CSSParser
+from cssutils.css import CSSRule
 
 import calibre
 from calibre.constants import filesystem_encoding
@@ -762,6 +763,7 @@ class Manifest(object):
             self.href = self.path = urlnormalize(href)
             self.media_type = media_type
             self.fallback = fallback
+            self.override_css_fetch = None
             self.spine_position = None
             self.linear = True
             if loader is None and data is None:
@@ -982,15 +984,40 @@ class Manifest(object):
 
 
         def _parse_css(self, data):
+
+            def get_style_rules_from_import(import_rule):
+                ans = []
+                if not import_rule.styleSheet:
+                    return ans
+                rules = import_rule.styleSheet.cssRules
+                for rule in rules:
+                    if rule.type == CSSRule.IMPORT_RULE:
+                        ans.extend(get_style_rules_from_import(rule))
+                    elif rule.type in (CSSRule.FONT_FACE_RULE,
+                            CSSRule.STYLE_RULE):
+                        ans.append(rule)
+                return ans
+
             self.oeb.log.debug('Parsing', self.href, '...')
             data = self.oeb.decode(data)
-            data = self.oeb.css_preprocessor(data)
-            data = XHTML_CSS_NAMESPACE + data
+            data = self.oeb.css_preprocessor(data, add_namespace=True)
             parser = CSSParser(loglevel=logging.WARNING,
-                               fetcher=self._fetch_css,
+                               fetcher=self.override_css_fetch or self._fetch_css,
                                log=_css_logger)
             data = parser.parseString(data, href=self.href)
             data.namespaces['h'] = XHTML_NS
+            import_rules = list(data.cssRules.rulesOfType(CSSRule.IMPORT_RULE))
+            rules_to_append = []
+            insert_index = None
+            for r in data.cssRules.rulesOfType(CSSRule.STYLE_RULE):
+                insert_index = data.cssRules.index(r)
+                break
+            for rule in import_rules:
+                rules_to_append.extend(get_style_rules_from_import(rule))
+            for r in reversed(rules_to_append):
+                data.insertRule(r, index=insert_index)
+            for rule in import_rules:
+                data.deleteRule(rule)
             return data
 
         def _fetch_css(self, path):