SF bug [#460467] file objects should be subclassable.

Preliminary support.  What's here works, but needs fine-tuning.

Include/fileobject.h

@@ -9,7 +9,8 @@ extern "C" {
 
 extern DL_IMPORT(PyTypeObject) PyFile_Type;
 
-#define PyFile_Check(op) ((op)->ob_type == &PyFile_Type)
+#define PyFile_Check(op) PyObject_TypeCheck(op, &PyFile_Type)
+#define PyFile_CheckExact(op) ((op)->ob_type == &PyFile_Type)
 
 extern DL_IMPORT(PyObject *) PyFile_FromString(char *, char *);
 extern DL_IMPORT(void) PyFile_SetBufSize(PyObject *, int);

Include/moduleobject.h

@@ -10,6 +10,7 @@ extern "C" {
 extern DL_IMPORT(PyTypeObject) PyModule_Type;
 
 #define PyModule_Check(op) PyObject_TypeCheck(op, &PyModule_Type)
+#define PyModule_CheckExact(op) ((op)->ob_type == &PyModule_Type)
 
 extern DL_IMPORT(PyObject *) PyModule_New(char *);
 extern DL_IMPORT(PyObject *) PyModule_GetDict(PyObject *);

Lib/test/test_descr.py

@@ -1,6 +1,6 @@
 # Test descriptor-related enhancements
 
-from test_support import verify, verbose, TestFailed
+from test_support import verify, verbose, TestFailed, TESTFN
 from copy import deepcopy
 
 def testunop(a, res, expr="len(a)", meth="__len__"):
@@ -1636,6 +1636,53 @@ def inherits():
     verify(u[0:0].__class__ is unicode)
     verify(u[0:0] == u"")
 
+    class CountedInput(file):
+        """Counts lines read by self.readline().
+
+        self.lineno is the 0-based ordinal of the last line read, up to
+        a maximum of one greater than the number of lines in the file.
+
+        self.ateof is true if and only if the final "" line has been read,
+        at which point self.lineno stops incrementing, and further calls
+        to readline() continue to return "".
+        """
+
+        lineno = 0
+        ateof = 0
+        def readline(self):
+            if self.ateof:
+                return ""
+            s = file.readline(self)
+            # Next line works too.
+            # s = super(CountedInput, self).readline()
+            self.lineno += 1
+            if s == "":
+                self.ateof = 1
+            return s
+
+    f = open(TESTFN, 'w')
+    lines = ['a\n', 'b\n', 'c\n']
+    try:
+        f.writelines(lines)
+        f.close()
+        f = CountedInput(TESTFN)
+        for (i, expected) in zip(range(1, 5) + [4], lines + 2 * [""]):
+            got = f.readline()
+            verify(expected == got)
+            verify(f.lineno == i)
+            verify(f.ateof == (i > len(lines)))
+        f.close()
+    finally:
+        try:
+            f.close()
+        except:
+            pass
+        try:
+            import os
+            os.unlink(TESTFN)
+        except:
+            pass
+
 def all():
     lists()
     dicts()

Lib/types.py

@@ -57,12 +57,7 @@ BuiltinFunctionType = type(len)
 BuiltinMethodType = type([].append)     # Same as BuiltinFunctionType
 
 ModuleType = type(sys)
-
+FileType = file
-try:
-    FileType = type(sys.__stdin__)
-except AttributeError:
-    # Not available in restricted mode
-    pass
 XRangeType = type(xrange(0))
 
 try:

Misc/NEWS

@@ -3,6 +3,11 @@ What's New in Python 2.2a4?
 
 Core
 
+- The builtin file type can be subclassed now.  In the usual pattern,
+  "file" is the name of the builtin type, and file() is a new builtin
+  constructor, with the same signature as the builtin open() function.
+  file() is now the preferred way to open a file.
+
 - In 2.2a3, hash() applied to an instance of a subclass of str or unicode
   always returned 0.  This has been repaired.
 

Objects/fileobject.c

@@ -65,37 +65,33 @@ PyFile_Name(PyObject *f)
 		return ((PyFileObject *)f)->f_name;
 }
 
-PyObject *
+
-PyFile_FromFile(FILE *fp, char *name, char *mode, int (*close)(FILE *))
+static PyObject *
+fill_file_fields(PyFileObject *f, FILE *fp, char *name, char *mode,
+		 int (*close)(FILE *))
 {
-	PyFileObject *f = PyObject_NEW(PyFileObject, &PyFile_Type);
+	assert(f != NULL);
-	if (f == NULL)
+	assert(PyFile_Check(f));
-		return NULL;
 	f->f_fp = NULL;
 	f->f_name = PyString_FromString(name);
 	f->f_mode = PyString_FromString(mode);
 	f->f_close = close;
 	f->f_softspace = 0;
-	if (strchr(mode,'b') != NULL)
+	f->f_binary = strchr(mode,'b') != NULL;
-	    f->f_binary = 1;
+	if (f->f_name == NULL || f->f_mode == NULL)
-	else
-	    f->f_binary = 0;
-	if (f->f_name == NULL || f->f_mode == NULL) {
-		Py_DECREF(f);
 		return NULL;
-	}
 	f->f_fp = fp;
 	return (PyObject *) f;
 }
 
-PyObject *
+static PyObject *
-PyFile_FromString(char *name, char *mode)
+open_the_file(PyFileObject *f, char *name, char *mode)
 {
-	extern int fclose(FILE *);
+	assert(f != NULL);
-	PyFileObject *f;
+	assert(PyFile_Check(f));
-	f = (PyFileObject *) PyFile_FromFile((FILE *)NULL, name, mode, fclose);
+	assert(name != NULL);
-	if (f == NULL)
+	assert(mode != NULL);
-		return NULL;
+
 #ifdef HAVE_FOPENRF
 	if (*mode == '*') {
 		FILE *fopenRF();
@@ -118,8 +114,36 @@ PyFile_FromString(char *name, char *mode)
 		}
 #endif
 		PyErr_SetFromErrnoWithFilename(PyExc_IOError, name);
+		f = NULL;
+	}
+	return (PyObject *)f;
+}
+
+PyObject *
+PyFile_FromFile(FILE *fp, char *name, char *mode, int (*close)(FILE *))
+{
+	PyFileObject *f = PyObject_NEW(PyFileObject, &PyFile_Type);
+	if (f != NULL) {
+		if (fill_file_fields(f, fp, name, mode, close) == NULL) {
 			Py_DECREF(f);
-		return NULL;
+			f = NULL;
+		}
+	}
+	return (PyObject *) f;
+}
+
+PyObject *
+PyFile_FromString(char *name, char *mode)
+{
+	extern int fclose(FILE *);
+	PyFileObject *f;
+
+	f = (PyFileObject *)PyFile_FromFile((FILE *)NULL, name, mode, fclose);
+	if (f != NULL) {
+		if (open_the_file(f, name, mode) == NULL) {
+			Py_DECREF(f);
+			f = NULL;
+		}
 	}
 	return (PyObject *)f;
 }
@@ -1293,6 +1317,52 @@ file_getiter(PyObject *f)
 	return PyObject_CallMethod(f, "xreadlines", "");
 }
 
+static PyObject *
+file_new(PyTypeObject *type, PyObject *args, PyObject *kwds)
+{
+	/* XXX As for all XXX_new functions, file_new is called
+	   with kwds=NULL by type_call(), so the kwlist is impotent. */
+	static char *kwlist[] = {"name", "mode", "buffering", 0};
+	char *name = NULL;
+	char *mode = "r";
+	int bufsize = -1;
+	PyObject *f;
+	extern int fclose(FILE *);
+
+	if (!PyArg_ParseTupleAndKeywords(args, kwds, "et|si:file", kwlist,
+					 Py_FileSystemDefaultEncoding, &name,
+					 &mode, &bufsize))
+		return NULL;
+	f = PyType_GenericAlloc(type, 0);
+	if (f != NULL) {
+		PyFileObject *g = (PyFileObject *)f;
+		if (fill_file_fields(g, NULL, name, mode, fclose) == NULL) {
+			Py_DECREF(f);
+			f = NULL;
+		}
+		if (f != NULL && open_the_file(g, name, mode) == NULL) {
+			Py_DECREF(f);
+			f = NULL;
+		}
+		if (f != NULL)
+			PyFile_SetBufSize(f, bufsize);
+	}
+	PyMem_Free(name); /* free the encoded string */
+	return f;
+}
+
+/* XXX Keep this in synch with open_doc in bltinmodule.c. */
+static char file_doc[] =
+"file(name[, mode[, buffering]]) -> file object\n"
+"\n"
+"Open a file.  The mode can be 'r', 'w' or 'a' for reading (default),\n"
+"writing or appending.  The file will be created if it doesn't exist\n"
+"when opened for writing or appending; it will be truncated when\n"
+"opened for writing.  Add a 'b' to the mode for binary files.\n"
+"Add a '+' to the mode to allow simultaneous reading and writing.\n"
+"If the buffering argument is given, 0 means unbuffered, 1 means line\n"
+"buffered, and larger numbers specify the buffer size.";
+
 PyTypeObject PyFile_Type = {
 	PyObject_HEAD_INIT(&PyType_Type)
 	0,
@@ -1314,8 +1384,9 @@ PyTypeObject PyFile_Type = {
 	PyObject_GenericGetAttr,		/* tp_getattro */
 	0,					/* tp_setattro */
 	0,					/* tp_as_buffer */
-	Py_TPFLAGS_DEFAULT,			/* tp_flags */
+	Py_TPFLAGS_DEFAULT |
-	0,					/* tp_doc */
+			Py_TPFLAGS_BASETYPE,	/* tp_flags */
+	file_doc,				/* tp_doc */
 	0,					/* tp_traverse */
 	0,					/* tp_clear */
 	0,					/* tp_richcompare */
@@ -1327,6 +1398,12 @@ PyTypeObject PyFile_Type = {
 	file_getsetlist,			/* tp_getset */
 	0,					/* tp_base */
 	0,					/* tp_dict */
+	0,					/* tp_descr_get */
+	0,					/* tp_descr_set */
+	0,					/* tp_dictoffset */
+	0,					/* tp_init */
+	0,					/* tp_alloc */
+	file_new,				/* tp_new */
 };
 
 /* Interface for the 'soft space' between print items. */

Python/bltinmodule.c

@@ -1192,31 +1192,20 @@ Return the octal representation of an integer or long integer.";
 static PyObject *
 builtin_open(PyObject *self, PyObject *args)
 {
-	char *name = NULL;
+	return PyFile_Type.tp_new(&PyFile_Type, args, NULL);
-	char *mode = "r";
-	int bufsize = -1;
-	PyObject *f;
-
-	if (!PyArg_ParseTuple(args, "et|si:open", Py_FileSystemDefaultEncoding, 
-	                      &name, &mode, &bufsize))
-		return NULL;
-	f = PyFile_FromString(name, mode);
-	PyMem_Free(name); /* free the encoded string */
-	if (f != NULL)
-		PyFile_SetBufSize(f, bufsize);
-	return f;
 }
 
+/* XXX Keep this in synch with file_doc in fileobject.c. */
 static char open_doc[] =
-"open(filename[, mode[, buffering]]) -> file object\n\
+"open(name[, mode[, buffering]]) -> file object\n"
-\n\
+"\n"
-Open a file.  The mode can be 'r', 'w' or 'a' for reading (default),\n\
+"Open a file.  The mode can be 'r', 'w' or 'a' for reading (default),\n"
-writing or appending.  The file will be created if it doesn't exist\n\
+"writing or appending.  The file will be created if it doesn't exist\n"
-when opened for writing or appending; it will be truncated when\n\
+"when opened for writing or appending; it will be truncated when\n"
-opened for writing.  Add a 'b' to the mode for binary files.\n\
+"opened for writing.  Add a 'b' to the mode for binary files.\n"
-Add a '+' to the mode to allow simultaneous reading and writing.\n\
+"Add a '+' to the mode to allow simultaneous reading and writing.\n"
-If the buffering argument is given, 0 means unbuffered, 1 means line\n\
+"If the buffering argument is given, 0 means unbuffered, 1 means line\n"
-buffered, and larger numbers specify the buffer size.";
+"buffered, and larger numbers specify the buffer size.";
 
 
 static PyObject *
@@ -1894,6 +1883,8 @@ _PyBuiltin_Init(void)
 		return NULL;
 	if (PyDict_SetItemString(dict, "type", (PyObject *) &PyType_Type) < 0)
 		return NULL;
+	if (PyDict_SetItemString(dict, "file", (PyObject *) &PyFile_Type) < 0)
+		return NULL;
 #ifdef Py_USING_UNICODE
 	if (PyDict_SetItemString(dict, "unicode",
 				 (PyObject *) &PyUnicode_Type) < 0)