merge with http://bitbucket.org/methane/sphinx-speedup/

sphinx/builders/html.py

@@ -23,7 +23,7 @@ except ImportError:
 
 from docutils import nodes
 from docutils.io import DocTreeInput, StringOutput
-from docutils.core import publish_parts
+from docutils.core import Publisher, publish_parts
 from docutils.utils import new_document
 from docutils.frontend import OptionParser
 from docutils.readers.doctree import Reader as DoctreeReader
@@ -74,6 +74,8 @@ class StandaloneHTMLBuilder(Builder):
     script_files = ['_static/jquery.js', '_static/doctools.js']
 
     def init(self):
+        # cached publisher object for snippets
+        self._publisher = None
         # a hash of all config values that, if changed, cause a full rebuild
         self.config_hash = ''
         self.tags_hash = ''
@@ -181,13 +183,24 @@ class StandaloneHTMLBuilder(Builder):
         """Utility: Render a lone doctree node."""
         doc = new_document('<partial node>')
         doc.append(node)
-        return publish_parts(
-            doc,
-            source_class=DocTreeInput,
-            reader=DoctreeReader(),
-            writer=HTMLWriter(self),
-            settings_overrides={'output_encoding': 'unicode'}
-        )
+
+        if self._publisher is None:
+            self._publisher = Publisher(
+                    source_class = DocTreeInput,
+                    destination_class=StringOutput)
+            self._publisher.set_components('standalone',
+                                           'restructuredtext', 'pseudoxml')
+
+        pub = self._publisher
+
+        pub.reader = DoctreeReader()
+        pub.writer = HTMLWriter(self)
+        pub.process_programmatic_settings(
+            None, {'output_encoding': 'unicode'}, None)
+        pub.set_source(doc, None)
+        pub.set_destination(None, None)
+        pub.publish()
+        return pub.writer.parts
 
     def prepare_writing(self, docnames):
         from sphinx.search import IndexBuilder

sphinx/search.py

@@ -14,8 +14,14 @@ from cStringIO import StringIO
 
 from docutils.nodes import Text, NodeVisitor
 
-from sphinx.util.stemmer import PorterStemmer
 from sphinx.util import jsdump, rpartition
+try:
+    # http://bitbucket.org/methane/porterstemmer/
+    from porterstemmer import Stemmer as CStemmer
+    CSTEMMER = True
+except ImportError:
+    from sphinx.util.stemmer import PorterStemmer
+    CSTEMMER = False
 
 
 word_re = re.compile(r'\w+(?u)')
@@ -62,15 +68,23 @@ class _JavaScriptIndex(object):
 js_index = _JavaScriptIndex()
 
 
-class Stemmer(PorterStemmer):
-    """
-    All those porter stemmer implementations look hideous.
-    make at least the stem method nicer.
-    """
+if CSTEMMER:
+    class Stemmer(CStemmer):
+
+        def stem(self, word):
+            return self(word.lower())
+
+else:
+    class Stemmer(PorterStemmer):
+        """
+        All those porter stemmer implementations look hideous.
+        make at least the stem method nicer.
+        """
+
+        def stem(self, word):
+            word = word.lower()
+            return PorterStemmer.stem(self, word, 0, len(word) - 1)
 
-    def stem(self, word):
-        word = word.lower()
-        return PorterStemmer.stem(self, word, 0, len(word) - 1)
 
 
 class WordCollector(NodeVisitor):
@@ -196,11 +210,11 @@ class IndexBuilder(object):
         visitor = WordCollector(doctree)
         doctree.walk(visitor)
 
-        def add_term(word, prefix='', stem=self._stemmer.stem):
+        def add_term(word, stem=self._stemmer.stem):
             word = stem(word)
             if len(word) < 3 or word in stopwords or word.isdigit():
                 return
-            self._mapping.setdefault(prefix + word, set()).add(filename)
+            self._mapping.setdefault(word, set()).add(filename)
 
         for word in word_re.findall(title):
             add_term(word)

sphinx/util/__init__.py

@@ -416,30 +416,28 @@ def copy_static_entry(source, target, builder, context={}):
 # traverse() is called so many times during a build that it saves
 # on average 20-25% overall build time!
 
-def _all_traverse(self):
+def _all_traverse(self, result):
     """Version of Node.traverse() that doesn't need a condition."""
-    result = []
     result.append(self)
     for child in self.children:
-        result.extend(child._all_traverse())
+        child._all_traverse(result)
     return result
 
-def _fast_traverse(self, cls):
+def _fast_traverse(self, cls, result):
     """Version of Node.traverse() that only supports instance checks."""
-    result = []
     if isinstance(self, cls):
         result.append(self)
     for child in self.children:
-        result.extend(child._fast_traverse(cls))
+        child._fast_traverse(cls, result)
     return result
 
 def _new_traverse(self, condition=None,
                  include_self=1, descend=1, siblings=0, ascend=0):
     if include_self and descend and not siblings and not ascend:
         if condition is None:
-            return self._all_traverse()
+            return self._all_traverse([])
         elif isinstance(condition, (types.ClassType, type)):
-            return self._fast_traverse(condition)
+            return self._fast_traverse(condition, [])
     return self._old_traverse(condition, include_self,
                               descend, siblings, ascend)