DOCX: Performance improvement

src/calibre/ebooks/docx/names.py

@@ -83,11 +83,10 @@ def get(x, attr, default=None):
     return x.attrib.get(expand(attr), default)
 
 def ancestor(elem, name):
-    tag = expand(name)
-    while elem is not None:
-        elem = elem.getparent()
-        if getattr(elem, 'tag', None) == tag:
-            return elem
+    try:
+        return XPath('ancestor::%s[1]' % name)(elem)[0]
+    except IndexError:
+        return None
 
 def generate_anchor(name, existing):
     x = y = 'id_' + re.sub(r'[^0-9a-zA-Z_]', '', ascii_text(name)).lstrip('_')

src/calibre/ebooks/docx/to_html.py

@@ -16,7 +16,7 @@ from lxml.html.builder import (
 from calibre.ebooks.docx.container import DOCX, fromstring
 from calibre.ebooks.docx.names import (
     XPath, is_tag, XML, STYLES, NUMBERING, FONTS, get, generate_anchor,
-    descendants, ancestor, FOOTNOTES, ENDNOTES, children)
+    descendants, FOOTNOTES, ENDNOTES, children)
 from calibre.ebooks.docx.styles import Styles, inherit, PageProperties
 from calibre.ebooks.docx.numbering import Numbering
 from calibre.ebooks.docx.fonts import Fonts
@@ -308,6 +308,7 @@ class Convert(object):
 
         current_anchor = None
         current_hyperlink = None
+        hl_xpath = XPath('ancestor::w:hyperlink[1]')
 
         for x in descendants(p, 'w:r', 'w:bookmarkStart', 'w:hyperlink'):
             if x.tag.endswith('}r'):
@@ -316,10 +317,10 @@ class Convert(object):
                     (dest if len(dest) == 0 else span).set('id', current_anchor)
                     current_anchor = None
                 if current_hyperlink is not None:
-                    hl = ancestor(x, 'w:hyperlink')
-                    if hl is not None:
+                    try:
+                        hl = hl_xpath(x)[0]
                         self.link_map[hl].append(span)
-                    else:
+                    except IndexError:
                         current_hyperlink = None
                 dest.append(span)
                 self.layers[p].append(x)