builtin_zip(): Take a good guess at how big the result list will be,

and allocate it in one gulp.

This isn't a bugfix, it's just a minor optimization that may or may not
pay off.

Lib/test/test_iter.py

@@ -467,6 +467,34 @@ class TestCase(unittest.TestCase):
             except OSError:
                 pass
 
+        self.assertEqual(zip(xrange(5)), [(i,) for i in range(5)])
+
+        # Classes that lie about their lengths.
+        class NoGuessLen5:
+            def __getitem__(self, i):
+                if i >= 5:
+                    raise IndexError
+                return i
+
+        class Guess3Len5(NoGuessLen5):
+            def __len__(self):
+                return 3
+
+        class Guess30Len5(NoGuessLen5):
+            def __len__(self):
+                return 30
+
+        self.assertEqual(len(Guess3Len5()), 3)
+        self.assertEqual(len(Guess30Len5()), 30)
+        self.assertEqual(zip(NoGuessLen5()), zip(range(5)))
+        self.assertEqual(zip(Guess3Len5()), zip(range(5)))
+        self.assertEqual(zip(Guess30Len5()), zip(range(5)))
+
+        expected = [(i, i) for i in range(5)]
+        for x in NoGuessLen5(), Guess3Len5(), Guess30Len5():
+            for y in NoGuessLen5(), Guess3Len5(), Guess30Len5():
+                self.assertEqual(zip(x, y), expected)
+
     # Test reduces()'s use of iterators.
     def test_builtin_reduce(self):
         from operator import add

Python/bltinmodule.c

@@ -1704,9 +1704,10 @@ static PyObject*
 builtin_zip(PyObject *self, PyObject *args)
 {
 	PyObject *ret;
-	int itemsize = PySequence_Length(args);
+	const int itemsize = PySequence_Length(args);
 	int i;
 	PyObject *itlist;  /* tuple of iterators */
+	int len;	   /* guess at result length */
 
 	if (itemsize < 1) {
 		PyErr_SetString(PyExc_TypeError,
@@ -1716,8 +1717,21 @@ builtin_zip(PyObject *self, PyObject *args)
 	/* args must be a tuple */
 	assert(PyTuple_Check(args));
 
+	/* Guess at result length:  the shortest of the input lengths. */
+	len = -1;	/* unknown */
+	for (i = 0; i < itemsize; ++i) {
+		PyObject *item = PyTuple_GET_ITEM(args, i);
+		int thislen = PySequence_Length(item);
+		if (thislen < 0)
+			PyErr_Clear();
+		else if (len < 0 || thislen < len)
+			len = thislen;
+	}
+
 	/* allocate result list */
-	if ((ret = PyList_New(0)) == NULL)
+	if (len < 0)
+		len = 10;	/* arbitrary */
+	if ((ret = PyList_New(len)) == NULL)
 		return NULL;
 
 	/* obtain iterators */
@@ -1738,14 +1752,14 @@ builtin_zip(PyObject *self, PyObject *args)
 	}
 
 	/* build result into ret list */
-	for (;;) {
+	for (i = 0; ; ++i) {
-		int status;
+		int j;
 		PyObject *next = PyTuple_New(itemsize);
 		if (!next)
 			goto Fail_ret_itlist;
 
-		for (i = 0; i < itemsize; i++) {
+		for (j = 0; j < itemsize; j++) {
-			PyObject *it = PyTuple_GET_ITEM(itlist, i);
+			PyObject *it = PyTuple_GET_ITEM(itlist, j);
 			PyObject *item = PyIter_Next(it);
 			if (!item) {
 				if (PyErr_Occurred()) {
@@ -1754,16 +1768,29 @@ builtin_zip(PyObject *self, PyObject *args)
 				}
 				Py_DECREF(next);
 				Py_DECREF(itlist);
-				return ret;
+				goto Done;
 			}
-			PyTuple_SET_ITEM(next, i, item);
+			PyTuple_SET_ITEM(next, j, item);
 		}
 
-		status = PyList_Append(ret, next);
+		if (i < len)
+			PyList_SET_ITEM(ret, i, next);
+		else {
+			int status = PyList_Append(ret, next);
 			Py_DECREF(next);
+			++len;
 			if (status < 0)
 				goto Fail_ret_itlist;
 		}
+	}
+
+Done:
+	if (ret != NULL && i < len) {
+		/* The list is too big. */
+		if (PyList_SetSlice(ret, i, len, NULL) < 0)
+			return NULL;
+	}
+	return ret;
 
 Fail_ret_itlist:
 	Py_DECREF(itlist);