[Python-checkins] r67601 - in python/trunk: Lib/test/test_sys.py Misc/NEWS Objects/stringobject.c

[mark.dickinson]

Author: mark.dickinson
Date: Fri Dec  5 22:55:28 2008
New Revision: 67601

Log:
Issue #4445: save 3 bytes (on average, on a typical machine) per
string allocation.


Modified:
   python/trunk/Lib/test/test_sys.py
   python/trunk/Misc/NEWS
   python/trunk/Objects/stringobject.c

Modified: python/trunk/Lib/test/test_sys.py
==============================================================================
--- python/trunk/Lib/test/test_sys.py	(original)
+++ python/trunk/Lib/test/test_sys.py	Fri Dec  5 22:55:28 2008
@@ -631,8 +631,8 @@
         # slice
         check(slice(1), size(h + '3P'))
         # str
-        check('', size(vh + 'lic'))
-        check('abc', size(vh + 'lic') + 3*self.c)
+        check('', struct.calcsize(vh + 'li') + 1)
+        check('abc', struct.calcsize(vh + 'li') + 1 + 3*self.c)
         # super
         check(super(int), size(h + '3P'))
         # tuple

Modified: python/trunk/Misc/NEWS
==============================================================================
--- python/trunk/Misc/NEWS	(original)
+++ python/trunk/Misc/NEWS	Fri Dec  5 22:55:28 2008
@@ -12,6 +12,11 @@
 Core and Builtins
 -----------------
 
+- Issue #4445: Replace "sizeof(PyStringObject)" with
+  "offsetof(PyStringObject, ob_sval) + 1" when allocating memory for
+  str instances.  On a typical machine this saves 3 bytes of memory
+  (on average) per string allocation.
+
 - Issue #3996: On Windows, the PyOS_CheckStack function would cause the
   interpreter to abort ("Fatal Python error: Could not reset the stack!")
   instead of throwing a MemoryError.

Modified: python/trunk/Objects/stringobject.c
==============================================================================
--- python/trunk/Objects/stringobject.c	(original)
+++ python/trunk/Objects/stringobject.c	Fri Dec  5 22:55:28 2008
@@ -4,6 +4,7 @@
 
 #include "Python.h"
 #include <ctype.h>
+#include <stddef.h>
 
 #ifdef COUNT_ALLOCS
 int null_strings, one_strings;
@@ -22,6 +23,14 @@
 */
 static PyObject *interned;
 
+/* PyStringObject_SIZE gives the basic size of a string; any memory allocation
+   for a string of length n should request PyStringObject_SIZE + n bytes.
+
+   Using PyStringObject_SIZE instead of sizeof(PyStringObject) saves
+   3 bytes per string allocation on a typical system.
+*/
+#define PyStringObject_SIZE (offsetof(PyStringObject, ob_sval) + 1)
+
 /*
    For both PyString_FromString() and PyString_FromStringAndSize(), the
    parameter `size' denotes number of characters to allocate, not counting any
@@ -74,13 +83,13 @@
 		return (PyObject *)op;
 	}
 
-	if (size > PY_SSIZE_T_MAX - sizeof(PyStringObject)) {
+	if (size > PY_SSIZE_T_MAX - PyStringObject_SIZE) {
 		PyErr_SetString(PyExc_OverflowError, "string is too large");
 		return NULL;
 	}
 
 	/* Inline PyObject_NewVar */
-	op = (PyStringObject *)PyObject_MALLOC(sizeof(PyStringObject) + size);
+	op = (PyStringObject *)PyObject_MALLOC(PyStringObject_SIZE + size);
 	if (op == NULL)
 		return PyErr_NoMemory();
 	PyObject_INIT_VAR(op, &PyString_Type, size);
@@ -114,7 +123,7 @@
 
 	assert(str != NULL);
 	size = strlen(str);
-	if (size > PY_SSIZE_T_MAX - sizeof(PyStringObject)) {
+	if (size > PY_SSIZE_T_MAX - PyStringObject_SIZE) {
 		PyErr_SetString(PyExc_OverflowError,
 			"string is too long for a Python string");
 		return NULL;
@@ -135,7 +144,7 @@
 	}
 
 	/* Inline PyObject_NewVar */
-	op = (PyStringObject *)PyObject_MALLOC(sizeof(PyStringObject) + size);
+	op = (PyStringObject *)PyObject_MALLOC(PyStringObject_SIZE + size);
 	if (op == NULL)
 		return PyErr_NoMemory();
 	PyObject_INIT_VAR(op, &PyString_Type, size);
@@ -992,14 +1001,14 @@
 				"strings are too large to concat");
 		return NULL;
 	}
-	  
+
 	/* Inline PyObject_NewVar */
-	if (size > PY_SSIZE_T_MAX - sizeof(PyStringObject)) {
+	if (size > PY_SSIZE_T_MAX - PyStringObject_SIZE) {
 		PyErr_SetString(PyExc_OverflowError,
 				"strings are too large to concat");
 		return NULL;
 	}
-	op = (PyStringObject *)PyObject_MALLOC(sizeof(PyStringObject) + size);
+	op = (PyStringObject *)PyObject_MALLOC(PyStringObject_SIZE + size);
 	if (op == NULL)
 		return PyErr_NoMemory();
 	PyObject_INIT_VAR(op, &PyString_Type, size);
@@ -1036,13 +1045,12 @@
 		return (PyObject *)a;
 	}
 	nbytes = (size_t)size;
-	if (nbytes + sizeof(PyStringObject) <= nbytes) {
+	if (nbytes + PyStringObject_SIZE <= nbytes) {
 		PyErr_SetString(PyExc_OverflowError,
 			"repeated string is too long");
 		return NULL;
 	}
-	op = (PyStringObject *)
-		PyObject_MALLOC(sizeof(PyStringObject) + nbytes);
+	op = (PyStringObject *)PyObject_MALLOC(PyStringObject_SIZE + nbytes);
 	if (op == NULL)
 		return PyErr_NoMemory();
 	PyObject_INIT_VAR(op, &PyString_Type, size);
@@ -3940,7 +3948,7 @@
 string_sizeof(PyStringObject *v)
 {
 	Py_ssize_t res;
-	res = sizeof(PyStringObject) + v->ob_size * v->ob_type->tp_itemsize;
+	res = PyStringObject_SIZE + v->ob_size * v->ob_type->tp_itemsize;
 	return PyInt_FromSsize_t(res);
 }
 
@@ -4179,7 +4187,7 @@
 PyTypeObject PyString_Type = {
 	PyVarObject_HEAD_INIT(&PyType_Type, 0)
 	"str",
-	sizeof(PyStringObject),
+	PyStringObject_SIZE,
 	sizeof(char),
  	string_dealloc, 			/* tp_dealloc */
 	(printfunc)string_print, 		/* tp_print */
@@ -4275,7 +4283,7 @@
 	_Py_DEC_REFTOTAL;
 	_Py_ForgetReference(v);
 	*pv = (PyObject *)
-		PyObject_REALLOC((char *)v, sizeof(PyStringObject) + newsize);
+		PyObject_REALLOC((char *)v, PyStringObject_SIZE + newsize);
 	if (*pv == NULL) {
 		PyObject_Del(v);
 		PyErr_NoMemory();