Merge pull request #13039 from lazka/speed-up-iter-segments

tacaswell · web-flow · commit 87085bd9448c · 2018-12-24T11:59:35.000-05:00
PERF: Speed up Path.iter_segments()
diff --git a/doc/api/next_api_changes/2018-12-23-path-code-types.rst b/doc/api/next_api_changes/2018-12-23-path-code-types.rst
@@ -0,0 +1,7 @@
+Path code types like ``Path.MOVETO`` are now ``np.uint8`` instead of ``int``
+````````````````````````````````````````````````````````````````````````````
+
+``Path.STOP``, ``Path.MOVETO``, ``Path.LINETO``, ``Path.CURVE3``,
+``Path.CURVE4`` and ``Path.CLOSEPOLY`` are now of the type ``Path.code_type``
+(``np.uint8`` by default) instead of plain ``int``. This makes their type
+match the array value type of the ``Path.codes`` array.
diff --git a/lib/matplotlib/path.py b/lib/matplotlib/path.py
@@ -74,13 +74,15 @@ class Path(object):
 
     """
 
+    code_type = np.uint8
+
     # Path codes
-    STOP = 0         # 1 vertex
-    MOVETO = 1       # 1 vertex
-    LINETO = 2       # 1 vertex
-    CURVE3 = 3       # 2 vertices
-    CURVE4 = 4       # 3 vertices
-    CLOSEPOLY = 79   # 1 vertex
+    STOP = code_type(0)         # 1 vertex
+    MOVETO = code_type(1)       # 1 vertex
+    LINETO = code_type(2)       # 1 vertex
+    CURVE3 = code_type(3)       # 2 vertices
+    CURVE4 = code_type(4)       # 3 vertices
+    CLOSEPOLY = code_type(79)   # 1 vertex
 
     #: A dictionary mapping Path codes to the number of vertices that the
     #: code expects.
@@ -91,8 +93,6 @@ class Path(object):
                              CURVE4: 3,
                              CLOSEPOLY: 1}
 
-    code_type = np.uint8
-
     def __init__(self, vertices, codes=None, _interpolation_steps=1,
                  closed=False, readonly=False):
         """
@@ -399,24 +399,22 @@ def iter_segments(self, transform=None, remove_nans=True, clip=None,
                                snap=snap, stroke_width=stroke_width,
                                simplify=simplify, curves=curves,
                                sketch=sketch)
-        vertices = cleaned.vertices
-        codes = cleaned.codes
-        len_vertices = vertices.shape[0]
 
         # Cache these object lookups for performance in the loop.
         NUM_VERTICES_FOR_CODE = self.NUM_VERTICES_FOR_CODE
         STOP = self.STOP
 
-        i = 0
-        while i < len_vertices:
-            code = codes[i]
+        vertices = iter(cleaned.vertices)
+        codes = iter(cleaned.codes)
+        for curr_vertices, code in zip(vertices, codes):
             if code == STOP:
-                return
-            else:
-                num_vertices = NUM_VERTICES_FOR_CODE[code]
-                curr_vertices = vertices[i:i+num_vertices].flatten()
-                yield curr_vertices, code
-                i += num_vertices
+                break
+            extra_vertices = NUM_VERTICES_FOR_CODE[code] - 1
+            if extra_vertices:
+                for i in range(extra_vertices):
+                    next(codes)
+                    curr_vertices = np.append(curr_vertices, next(vertices))
+            yield curr_vertices, code
 
     def cleaned(self, transform=None, remove_nans=False, clip=None,
                 quantize=False, simplify=False, curves=False,
