- syntaxtree.Node.select: performance slightly improved

DHParser/syntaxtree.py

@@ -70,12 +70,6 @@ ZOMBIE_TAG = "__ZOMBIE__"
 #######################################################################
 
 
-ChildrenType = Tuple['Node', ...]
-NoChildren = cast(ChildrenType, ())  # type: ChildrenType
-StrictResultType = Union[ChildrenType, StringView, str]
-ResultType = Union[ChildrenType, 'Node', StringView, str, None]
-
-
 def flatten_sxpr(sxpr: str, threshold: int = -1) -> str:
     """
     Returns S-expression ``sxpr`` as a one-liner without unnecessary
@@ -114,6 +108,11 @@ def flatten_xml(xml: str) -> str:
     return re.sub(r'\s+(?=<[\w:])', '', re.sub(r'(?P<closing_tag></:?\w+>)\s+', tag_only, xml))
 
 
+ChildrenType = Tuple['Node', ...]
+NoChildren = cast(ChildrenType, ())  # type: ChildrenType
+StrictResultType = Union[ChildrenType, StringView, str]
+ResultType = Union[ChildrenType, 'Node', StringView, str, None]
+
 RX_AMP = re.compile(r'&(?!\w+;)')
 
 
@@ -365,6 +364,7 @@ class Node:  # (collections.abc.Sized): Base class omitted for cython-compatibil
                 self.children = NoChildren
                 self._result = result  # cast(StrictResultType, result)
 
+
     def _content(self) -> List[str]:
         """
         Returns string content as list of string fragments
@@ -378,6 +378,7 @@ class Node:  # (collections.abc.Sized): Base class omitted for cython-compatibil
         self._result = str(self._result)
         return [self._result]
 
+
     @property
     def content(self) -> str:
         """
@@ -697,8 +698,12 @@ class Node:  # (collections.abc.Sized): Base class omitted for cython-compatibil
             yield self
         child_iterator = reversed(self.children) if reverse else self.children
         for child in child_iterator:
-            for node in child.select(match_function, True, reverse):
-                yield node
+            if match_function(child):
+                yield child
+            yield from child.select(match_function, False, reverse)
+        # The above variant is slightly faster
+        # for child in child_iterator:
+        #     yield from child.select(match_function, True, reverse)
 
 
     def select_by_tag(self, tag_names: Union[str, AbstractSet[str]],

examples/LaTeX/tst_LaTeX_docs.py

@@ -55,13 +55,6 @@ def fail_on_error(src, result):
         sys.exit(1)
 
 
-def count_nodes(tree, condition=lambda n: True):
-    N = 0
-    for nd in tree.select(condition, include_root=True):
-        N += 1
-    return N
-
-
 def tst_func():
     with DHParser.log.logging(LOGGING):
         files = os.listdir('testdata')
@@ -73,7 +66,7 @@ def tst_func():
 
                 print('\n\nParsing document: "%s"' % file)
                 result = parser(doc)
-                print("Number of CST-nodes: " + str(count_nodes(result)))
+                print("Number of CST-nodes: " + str(result.tree_size()))
                 # print("Number of empty nodes: " + str(count_nodes(result,
                 #                                                 lambda n: not bool(n.result))))
                 if DHParser.log.is_logging():
@@ -87,7 +80,7 @@ def tst_func():
                 fail_on_error(doc, result)
                 transformer(result)
                 fail_on_error(doc, result)
-                print("Number of AST-nodes: " + str(count_nodes(result)))
+                print("Number of AST-nodes: " + str(result.tree_size()))
                 if DHParser.log.is_logging():
                     print('Saving AST')
                     with open('LOGS/' + file[:-4] + '.ast', 'w', encoding='utf-8') as f: