CVS patch #477161: New "access" keyword for mmap, from Jay T Miller.

This gives mmap() on Windows the ability to create read-only, write-
through and copy-on-write mmaps.  A new keyword argument is introduced
because the mmap() signatures diverged between Windows and Unix, so
while they (now) both support this functionality, there wasn't a way to
spell it in a common way without introducing a new spelling gimmick.
The old spellings are still accepted, so there isn't a backward-
compatibility issue here.

Lib/test/output/test_mmap

@@ -17,4 +17,17 @@ test_mmap
   Try to seek beyond end of mmap...
   Try to seek to negative position...
   Attempting resize()
+  Creating 10 byte test data file.
+  Opening mmap with access=ACCESS_READ
+  Ensuring that readonly mmap can't be slice assigned.
+  Ensuring that readonly mmap can't be item assigned.
+  Ensuring that readonly mmap can't be write() to.
+  Ensuring that readonly mmap can't be write_byte() to.
+  Ensuring that readonly mmap can't be resized.
+  Opening mmap with access=ACCESS_WRITE
+  Modifying write-through memory map.
+  Opening mmap with access=ACCESS_COPY
+  Modifying copy-on-write memory map.
+  Ensuring copy-on-write maps cannot be resized.
+  Ensuring invalid access parameter raises exception.
  Test passed

Lib/test/test_mmap.py

-from test_support import verify, TESTFN
+from test_support import verify, vereq, TESTFN
 import mmap
 import os, re
 
@@ -25,15 +25,15 @@ def test_both():
 
         print type(m)  # SF bug 128713:  segfaulted on Linux
         print '  Position of foo:', m.find('foo') / float(PAGESIZE), 'pages'
-        verify(m.find('foo') == PAGESIZE)
+        vereq(m.find('foo'), PAGESIZE)
 
         print '  Length of file:', len(m) / float(PAGESIZE), 'pages'
-        verify(len(m) == 2*PAGESIZE)
+        vereq(len(m), 2*PAGESIZE)
 
         print '  Contents of byte 0:', repr(m[0])
-        verify(m[0] == '\0')
+        vereq(m[0], '\0')
         print '  Contents of first 3 bytes:', repr(m[0:3])
-        verify(m[0:3] == '\0\0\0')
+        vereq(m[0:3], '\0\0\0')
 
         # Modify the file's content
         print "\n  Modifying file's content..."
@@ -42,11 +42,11 @@ def test_both():
 
         # Check that the modification worked
         print '  Contents of byte 0:', repr(m[0])
-        verify(m[0] == '3')
+        vereq(m[0], '3')
         print '  Contents of first 3 bytes:', repr(m[0:3])
-        verify(m[0:3] == '3\0\0')
+        vereq(m[0:3], '3\0\0')
         print '  Contents of second page:',  repr(m[PAGESIZE-1 : PAGESIZE + 7])
-        verify(m[PAGESIZE-1 : PAGESIZE + 7] == '\0foobar\0')
+        vereq(m[PAGESIZE-1 : PAGESIZE + 7], '\0foobar\0')
 
         m.flush()
 
@@ -61,19 +61,19 @@ def test_both():
             print '  Regex match on mmap (page start, length of match):',
             print start / float(PAGESIZE), length
 
-            verify(start == PAGESIZE)
-            verify(end == PAGESIZE + 6)
+            vereq(start, PAGESIZE)
+            vereq(end, PAGESIZE + 6)
 
         # test seeking around (try to overflow the seek implementation)
         m.seek(0,0)
         print '  Seek to zeroth byte'
-        verify(m.tell() == 0)
+        vereq(m.tell(), 0)
         m.seek(42,1)
         print '  Seek to 42nd byte'
-        verify(m.tell() == 42)
+        vereq(m.tell(), 42)
         m.seek(0,2)
         print '  Seek to last byte'
-        verify(m.tell() == len(m))
+        vereq(m.tell(), len(m))
 
         print '  Try to seek to negative position...'
         try:
@@ -132,6 +132,118 @@ def test_both():
         except OSError:
             pass
 
+    # Test for "access" keyword parameter
+    try:
+        mapsize = 10
+        print "  Creating", mapsize, "byte test data file."
+        open(TESTFN, "wb").write("a"*mapsize)
+        print "  Opening mmap with access=ACCESS_READ"
+        f = open(TESTFN, "rb")
+        m = mmap.mmap(f.fileno(), mapsize, access=mmap.ACCESS_READ)
+        verify(m[:] == 'a'*mapsize, "Readonly memory map data incorrect.")
+
+        print "  Ensuring that readonly mmap can't be slice assigned."
+        try:
+            m[:] = 'b'*mapsize
+        except TypeError:
+            pass
+        else:
+            verify(0, "Able to write to readonly memory map")
+
+        print "  Ensuring that readonly mmap can't be item assigned."
+        try:
+            m[0] = 'b'
+        except TypeError:
+            pass
+        else:
+            verify(0, "Able to write to readonly memory map")
+
+        print "  Ensuring that readonly mmap can't be write() to."
+        try:
+            m.seek(0,0)
+            m.write('abc')
+        except TypeError:
+            pass
+        else:
+            verify(0, "Able to write to readonly memory map")
+
+        print "  Ensuring that readonly mmap can't be write_byte() to."
+        try:
+            m.seek(0,0)
+            m.write_byte('d')
+        except TypeError:
+            pass
+        else:
+            verify(0, "Able to write to readonly memory map")
+
+        print "  Ensuring that readonly mmap can't be resized."
+        try:
+            m.resize(2*mapsize)
+        except SystemError:   # resize is not universally supported
+            pass
+        except TypeError:
+            pass
+        else:
+            verify(0, "Able to resize readonly memory map")
+        del m, f
+        verify(open(TESTFN, "rb").read() == 'a'*mapsize,
+               "Readonly memory map data file was modified")
+
+        print "  Opening mmap with access=ACCESS_WRITE"
+        f = open(TESTFN, "r+b")
+        m = mmap.mmap(f.fileno(), mapsize, access=mmap.ACCESS_WRITE)
+        print "  Modifying write-through memory map."
+        m[:] = 'c'*mapsize
+        verify(m[:] == 'c'*mapsize,
+               "Write-through memory map memory not updated properly.")
+        m.flush()
+        del m, f
+        verify(open(TESTFN).read() == 'c'*mapsize,
+               "Write-through memory map data file not updated properly.")
+
+        print "  Opening mmap with access=ACCESS_COPY"
+        f = open(TESTFN, "r+b")
+        m = mmap.mmap(f.fileno(), mapsize, access=mmap.ACCESS_COPY)
+        print "  Modifying copy-on-write memory map."
+        m[:] = 'd'*mapsize
+        verify(m[:] == 'd' * mapsize,
+               "Copy-on-write memory map data not written correctly.")
+        m.flush()
+        verify(open(TESTFN, "rb").read() == 'c'*mapsize,
+               "Copy-on-write test data file should not be modified.")
+        try:
+            print "  Ensuring copy-on-write maps cannot be resized."
+            m.resize(2*mapsize)
+        except TypeError:
+            pass
+        else:
+            verify(0, "Copy-on-write mmap resize did not raise exception.")
+        del m, f
+        try:
+            print "  Ensuring invalid access parameter raises exception."
+            f = open(TESTFN, "r+b")
+            m = mmap.mmap(f.fileno(), mapsize, access=4)
+        except ValueError:
+            pass
+        else:
+            verify(0, "Invalid access code should have raised exception.")
+
+        if os.name == "posix":
+            print "  Trying incompatible flags, prot and access parameters."
+            f=open(TESTFN, "r+b")
+            try:
+                m = mmap.mmap(f.fileno(), mapsize, flags=mmap.MAP_PRIVATE,
+                              prot=mmap.PROT_READ, access=mmap.ACCESS_WRITE)
+            except ValueError:
+                pass
+            else:
+                verify(0, "Incompatible parameters should raise ValueError.")
+    finally:
+        try:
+            os.unlink(TESTFN)
+        except OSError:
+            pass
+
     print ' Test passed'
 
 test_both()

Misc/NEWS

@@ -36,6 +36,13 @@ Core and builtins
 
 Extension modules
 
+- mmap has a new keyword argument, "access", allowing a uniform way for
+  both Windows and Unix users to create read-only, write-through and
+  copy-on-write memory mappings.  This was previously possible only on
+  Unix.  A new keyword argument was required to support this in a
+  uniform way because the mmap() signuatures had diverged across
+  platforms.  Thanks to Jay T Miller for repairing this!
+
 - By default, the gc.garbage list now contains only those instances in
   unreachable cycles that have __del__ methods; in 2.1 it contained all
   instances in unreachable cycles.  "Instances" here has been generalized

Modules/mmapmodule.c

@@ -62,6 +62,14 @@ my_getpagesize(void)
 
 static PyObject *mmap_module_error;
 
+typedef enum
+{
+	ACCESS_DEFAULT,
+	ACCESS_READ,
+	ACCESS_WRITE,
+	ACCESS_COPY
+} access_mode;
+
 typedef struct {
 	PyObject_HEAD
 	char *	data;
@@ -77,8 +85,11 @@ typedef struct {
 #ifdef UNIX
         int fd;
 #endif
+
+        access_mode access;
 } mmap_object;
 
+
 static void
 mmap_object_dealloc(mmap_object *m_obj)
 {
@@ -260,6 +271,26 @@ mmap_find_method(mmap_object *self,
 	}
 }
 
+static int 
+is_writeable(mmap_object *self)
+{
+	if (self->access != ACCESS_READ)
+		return 1; 
+	PyErr_Format(PyExc_TypeError, "mmap can't modify a readonly memory map.");
+	return 0;
+}
+
+static int 
+is_resizeable(mmap_object *self)
+{
+	if ((self->access == ACCESS_WRITE) || (self->access == ACCESS_DEFAULT))
+		return 1; 
+	PyErr_Format(PyExc_TypeError, 
+		     "mmap can't resize a readonly or copy-on-write memory map.");
+	return 0;
+}
+
+
 static PyObject *
 mmap_write_method(mmap_object *self,
 		  PyObject *args)
@@ -271,6 +302,9 @@ mmap_write_method(mmap_object *self,
 	if (!PyArg_ParseTuple (args, "s#:write", &data, &length))
 		return(NULL);
 
+	if (!is_writeable(self))
+		return NULL;
+
 	if ((self->pos + length) > self->size) {
 		PyErr_SetString (PyExc_ValueError, "data out of range");
 		return NULL;
@@ -291,6 +325,8 @@ mmap_write_byte_method(mmap_object *self,
 	if (!PyArg_ParseTuple (args, "c:write_byte", &value))
 		return(NULL);
 
+	if (!is_writeable(self))
+		return NULL;
 	*(self->data+self->pos) = value;
 	self->pos += 1;
 	Py_INCREF (Py_None);
@@ -342,7 +378,8 @@ mmap_resize_method(mmap_object *self,
 {
 	unsigned long new_size;
 	CHECK_VALID(NULL);
-	if (!PyArg_ParseTuple (args, "l:resize", &new_size)) {
+	if (!PyArg_ParseTuple (args, "l:resize", &new_size) || 
+	    !is_resizeable(self)) {
 		return NULL;
 #ifdef MS_WIN32
 	} else { 
@@ -386,12 +423,12 @@ mmap_resize_method(mmap_object *self,
 
 #ifdef UNIX
 #ifndef HAVE_MREMAP 
-} else {
+	} else {
 		PyErr_SetString(PyExc_SystemError,
 				"mmap: resizing not available--no mremap()");
 		return NULL;
 #else
-} else {
+	} else {
 		void *newmap;
 
 #ifdef MREMAP_MAYMOVE
@@ -410,7 +447,7 @@ mmap_resize_method(mmap_object *self,
 		return Py_None;
 #endif /* HAVE_MREMAP */
 #endif /* UNIX */
-}
+	}
 }
 
 static PyObject *
@@ -491,7 +528,7 @@ mmap_seek_method(mmap_object *self, PyObject *args)
 		return (Py_None);
 	}
 
-onoutofrange:
+  onoutofrange:
 	PyErr_SetString (PyExc_ValueError, "seek out of range");
 	return NULL;
 }
@@ -501,7 +538,8 @@ mmap_move_method(mmap_object *self, PyObject *args)
 {
 	unsigned long dest, src, count;
 	CHECK_VALID(NULL);
-	if (!PyArg_ParseTuple (args, "iii:move", &dest, &src, &count)) {
+	if (!PyArg_ParseTuple (args, "iii:move", &dest, &src, &count) ||
+	    !is_writeable(self)) {
 		return NULL;
 	} else {
 		/* bounds check the values */
@@ -561,6 +599,8 @@ mmap_buffer_getwritebuf(mmap_object *self, int index, const void **ptr)
 				"Accessing non-existent mmap segment");
 		return -1;
 	}
+	if (!is_writeable(self))
+		return -1;
 	*ptr = self->data;
 	return self->size;
 }
@@ -678,6 +718,8 @@ mmap_ass_slice(mmap_object *self, int ilow, int ihigh, PyObject *v)
 				"mmap slice assignment is wrong size");
 		return -1;
 	}
+	if (!is_writeable(self))
+		return -1;
 	buf = PyString_AsString(v);
 	memcpy(self->data + ilow, buf, ihigh-ilow);
 	return 0;
@@ -703,6 +745,8 @@ mmap_ass_item(mmap_object *self, int i, PyObject *v)
 				"mmap assignment must be single-character string");
 		return -1;
 	}
+	if (!is_writeable(self))
+		return -1;
 	buf = PyString_AsString(v);
 	self->data[i] = buf[0];
 	return 0;
@@ -796,12 +840,12 @@ _GetMapSize(PyObject *o)
 		return -1;
 	}
 
-onnegoverflow:
+  onnegoverflow:
 	PyErr_SetString(PyExc_OverflowError,
 			"memory mapped size must be positive");
 	return -1;
 
-onposoverflow:
+  onposoverflow:
 	PyErr_SetString(PyExc_OverflowError,
 			"memory mapped size is too large (limited by C int)");
 	return -1;
@@ -815,17 +859,43 @@ new_mmap_object(PyObject *self, PyObject *args, PyObject *kwdict)
 	PyObject *map_size_obj = NULL;
 	int map_size;
 	int fd, flags = MAP_SHARED, prot = PROT_WRITE | PROT_READ;
-	char *keywords[] = {"file", "size", "flags", "prot", NULL};
+	access_mode access = ACCESS_DEFAULT;
+	char *keywords[] = {"fileno", "length", 
+			    "flags", "prot", 
+			    "access", NULL};
 
-	if (!PyArg_ParseTupleAndKeywords(args, kwdict, 
-					 "iO|ii", keywords, 
-					 &fd, &map_size_obj, &flags, &prot)
-		)
+	if (!PyArg_ParseTupleAndKeywords(args, kwdict, "iO|iii", keywords, 
+					 &fd, &map_size_obj, &flags, &prot, &access))
 		return NULL;
 	map_size = _GetMapSize(map_size_obj);
 	if (map_size < 0)
 		return NULL;
 
+	if ((access != ACCESS_DEFAULT) && 
+	    ((flags != MAP_SHARED) || ( prot != (PROT_WRITE | PROT_READ))))
+		return PyErr_Format(PyExc_ValueError, 
+				    "mmap can't specify both access and flags, prot.");
+	switch(access) {
+	case ACCESS_READ:
+		flags = MAP_SHARED;
+		prot = PROT_READ;
+		break;
+	case ACCESS_WRITE:
+		flags = MAP_SHARED;
+		prot = PROT_READ | PROT_WRITE;
+		break;
+	case ACCESS_COPY:
+		flags = MAP_PRIVATE;
+		prot = PROT_READ | PROT_WRITE;
+		break;
+	case ACCESS_DEFAULT: 
+		/* use the specified or default values of flags and prot */
+		break;
+	default:
+		return PyErr_Format(PyExc_ValueError, 
+				    "mmap invalid access parameter.");
+	}
+	
 	m_obj = PyObject_New (mmap_object, &mmap_object_type);
 	if (m_obj == NULL) {return NULL;}
 	m_obj->size = (size_t) map_size;
@@ -834,36 +904,56 @@ new_mmap_object(PyObject *self, PyObject *args, PyObject *kwdict)
 	m_obj->data = mmap(NULL, map_size, 
 			   prot, flags,
 			   fd, 0);
-	if (m_obj->data == (char *)-1)
-	{
+	if (m_obj->data == (char *)-1) {
 		Py_DECREF(m_obj);
 		PyErr_SetFromErrno(mmap_module_error);
 		return NULL;
 	}
+	m_obj->access = access;
 	return (PyObject *)m_obj;
 }
 #endif /* UNIX */
 
 #ifdef MS_WIN32
 static PyObject *
-new_mmap_object(PyObject *self, PyObject *args)
+new_mmap_object(PyObject *self, PyObject *args, PyObject *kwdict)
 {
 	mmap_object *m_obj;
 	PyObject *map_size_obj = NULL;
 	int map_size;
 	char *tagname = "";
-
 	DWORD dwErr = 0;
 	int fileno;
 	HANDLE fh = 0;
-
-	if (!PyArg_ParseTuple(args,
-			  "iO|z",
-			  &fileno,
-			  &map_size_obj,
-			  &tagname)
-		)
+	access_mode   access = ACCESS_DEFAULT;
+	DWORD flProtect, dwDesiredAccess;
+	char *keywords[] = { "fileno", "length", 
+			     "tagname", 
+			     "access", NULL };
+
+	if (!PyArg_ParseTupleAndKeywords(args, kwdict, "iO|zi", keywords,
+					 &fileno, &map_size_obj, 
+					 &tagname, &access)) {
 		return NULL;
+	}
+
+	switch(access) {
+	case ACCESS_READ:
+		flProtect = PAGE_READONLY;
+		dwDesiredAccess = FILE_MAP_READ;
+		break;
+	case ACCESS_DEFAULT:  case ACCESS_WRITE:
+		flProtect = PAGE_READWRITE;
+		dwDesiredAccess = FILE_MAP_WRITE;
+		break;
+	case ACCESS_COPY:
+		flProtect = PAGE_WRITECOPY;
+		dwDesiredAccess = FILE_MAP_COPY;
+		break;
+	default:
+		return PyErr_Format(PyExc_ValueError, 
+				    "mmap invalid access parameter.");
+	}
 
 	map_size = _GetMapSize(map_size_obj);
 	if (map_size < 0)
@@ -932,15 +1022,16 @@ new_mmap_object(PyObject *self, PyObject *args)
 	else
 		m_obj->tagname = NULL;
 
+	m_obj->access = access;
 	m_obj->map_handle = CreateFileMapping (m_obj->file_handle,
 					       NULL,
-					       PAGE_READWRITE,
+					       flProtect,
 					       0,
 					       m_obj->size,
 					       m_obj->tagname);
 	if (m_obj->map_handle != NULL) {
 		m_obj->data = (char *) MapViewOfFile (m_obj->map_handle,
-						      FILE_MAP_WRITE,
+						      dwDesiredAccess,
 						      0,
 						      0,
 						      0);
@@ -966,7 +1057,7 @@ static struct PyMethodDef mmap_functions[] = {
 };
 
 DL_EXPORT(void)
-initmmap(void)
+	initmmap(void)
 {
 	PyObject *dict, *module;
 
@@ -1011,5 +1102,11 @@ initmmap(void)
 
 	PyDict_SetItemString (dict, "PAGESIZE",
 			      PyInt_FromLong( (long)my_getpagesize() ) );
-}
 
+	PyDict_SetItemString (dict, "ACCESS_READ",	
+			      PyInt_FromLong(ACCESS_READ));
+	PyDict_SetItemString (dict, "ACCESS_WRITE", 
+			      PyInt_FromLong(ACCESS_WRITE));
+	PyDict_SetItemString (dict, "ACCESS_COPY",	
+			      PyInt_FromLong(ACCESS_COPY));
+}