Add an lxml-optimized treewalker and fix caching of etree modules (do not cache at the __init__ level, let the etree submodule do its own caching based on the implementation)

tbroyer · tbroyer · commit cda0b54be81f · 2007-07-02T23:54:20.000Z
--HG--
extra : convert_revision : svn%3Aacbfec75-9323-0410-a652-858a13e371e0/trunk%40843
diff --git a/src/html5lib/treewalkers/__init__.py b/src/html5lib/treewalkers/__init__.py
@@ -20,15 +20,16 @@ def getTreeWalker(treeType, implementation=None, **kwargs):
                               more pythonic idioms.
                 "dom" - The xml.dom.minidom DOM implementation
                 "pulldom" - The xml.dom.pulldom event stream
-                "etree" - A generic builder for tree implementations exposing an
+                "etree" - A generic walker for tree implementations exposing an
                           elementtree-like interface (known to work with
                           ElementTree, cElementTree and lxml.etree).
+                "lxml" - Optimized walker for lxml.etree
                 "beautifulsoup" - Beautiful soup (if installed)
                 "genshi" - a Genshi stream
 
     implementation - (Currently applies to the "etree" tree type only). A module
                       implementing the tree type e.g. xml.etree.ElementTree or
-                      lxml.etree."""
+                      cElementTree."""
 
     treeType = treeType.lower()
     if treeType not in treeWalkerCache:
@@ -41,7 +42,11 @@ def getTreeWalker(treeType, implementation=None, **kwargs):
         elif treeType == "beautifulsoup":
             import soup
             treeWalkerCache[treeType] = soup.TreeWalker
+        elif treeType == "lxml":
+            import lxmletree
+            treeWalkerCache[treeType] = lxmletree.TreeWalker
         elif treeType == "etree":
             import etree
-            treeWalkerCache[treeType] = etree.getETreeModule(implementation, **kwargs).TreeWalker
+            # XXX: NEVER cache here, caching is done in the etree submodule
+            return etree.getETreeModule(implementation, **kwargs).TreeWalker
     return treeWalkerCache.get(treeType)
diff --git a/src/html5lib/treewalkers/lxmletree.py b/src/html5lib/treewalkers/lxmletree.py
@@ -0,0 +1,66 @@
+from lxml import etree
+
+from gettext import gettext
+_ = gettext
+
+import _base
+
+from html5lib.constants import voidElements
+
+class TreeWalker(_base.NonRecursiveTreeWalker):
+    def getNodeDetails(self, node):
+        if isinstance(node, tuple): # Text node
+            node, key = node
+            assert key in ("text", "tail"), _("Text nodes are text or tail, found %s") % key
+            return _base.TEXT, getattr(node, key)
+
+        if not(hasattr(node, "tag")):
+            node = node.getroot()
+
+        if node.tag in ("<DOCUMENT_ROOT>", "<DOCUMENT_FRAGMENT>"):
+            return (_base.DOCUMENT,)
+
+        elif node.tag == "<!DOCTYPE>":
+            return _base.DOCTYPE, node.text
+
+        elif node.tag == etree.Comment:
+            return _base.COMMENT, node.text
+
+        else:
+            #This is assumed to be an ordinary element
+            return _base.ELEMENT, node.tag, node.attrib.items(), bool(node) or node.text
+
+    def getFirstChild(self, node):
+        assert not isinstance(node, tuple), _("Text nodes have no children")
+
+        assert bool(node) or node.text, "Node has no children"
+        if node.text:
+            return (node, "text")
+        else:
+            return node[0]
+
+    def getNextSibling(self, node):
+        if isinstance(node, tuple): # Text node
+            node, key = node
+            assert key in ("text", "tail"), _("Text nodes are text or tail, found %s") % key
+            if key == "text":
+                # XXX: we cannot use a "bool(node) and node[0] or None" construct here
+                # because node[0] might evaluate to False if it has no child element
+                if bool(node):
+                    return node[0]
+                else:
+                    return None
+            else: # tail
+                return node.getnext()
+
+        return node.tail and (node, "tail") or node.getnext()
+
+    def getParentNode(self, node):
+        if isinstance(node, tuple): # Text node
+            node, key = node
+            assert key in ("text", "tail"), _("Text nodes are text or tail, found %s") % key
+            if key == "text":
+                return node
+            # else: fallback to "normal" processing
+
+        return node.getparent()
diff --git a/tests/test_treewalkers.py b/tests/test_treewalkers.py
@@ -80,9 +80,12 @@ def PullDOMAdapter(node):
 
 try:
     import lxml.etree as ElementTree
-    treeTypes['lxml'] = \
+    treeTypes['lxml_as_etree'] = \
         {"builder": treebuilders.getTreeBuilder("etree", ElementTree),
          "walker":  treewalkers.getTreeWalker("etree", ElementTree)}
+    treeTypes['lxml_native'] = \
+        {"builder": treebuilders.getTreeBuilder("etree", ElementTree),
+         "walker":  treewalkers.getTreeWalker("lxml")}
 except ImportError:
     pass
 
