Issue 4474: On platforms with sizeof(wchar_t) == 4 and

mdickinson · mdickinson · commit 081dfee4f154 · 2009-03-18T14:47:41.000Z
sizeof(Py_UNICODE) == 2, PyUnicode_FromWideChar now converts
each character outside the BMP to the appropriate surrogate pair.

Thanks Victor Stinner for the patch.
diff --git a/Misc/NEWS b/Misc/NEWS
@@ -12,6 +12,10 @@ What's New in Python 3.1 alpha 2?
 Core and Builtins
 -----------------
 
+- Issue #4474: PyUnicode_FromWideChar now converts characters outside
+  the BMP to surrogate pairs, on systems with sizeof(wchar_t) == 4
+  and sizeof(Py_UNICODE) == 2.
+
 - Issue #5237: Allow auto-numbered fields in str.format(). For
   example: '{} {}'.format(1, 2) == '1 2'.
 
diff --git a/Modules/_testcapimodule.c b/Modules/_testcapimodule.c
@@ -707,6 +707,50 @@ test_Z_code(PyObject *self)
 	Py_RETURN_NONE;
 }
 
+static PyObject *
+test_widechar(PyObject *self)
+{
+#if defined(SIZEOF_WCHAR_T) && (SIZEOF_WCHAR_T == 4)
+	const wchar_t wtext[2] = {(wchar_t)0x10ABCDu};
+	size_t wtextlen = 1;
+#else
+	const wchar_t wtext[3] = {(wchar_t)0xDBEAu, (wchar_t)0xDFCDu};
+	size_t wtextlen = 2;
+#endif
+	PyObject *wide, *utf8;
+
+	wide = PyUnicode_FromWideChar(wtext, wtextlen);
+	if (wide == NULL)
+		return NULL;
+
+	utf8 = PyUnicode_FromString("\xf4\x8a\xaf\x8d");
+	if (utf8 == NULL) {
+		Py_DECREF(wide);
+		return NULL;
+	}
+
+	if (PyUnicode_GET_SIZE(wide) != PyUnicode_GET_SIZE(utf8)) {
+		Py_DECREF(wide);
+		Py_DECREF(utf8);
+		return raiseTestError("test_widechar",
+				      "wide string and utf8 string "
+				      "have different length");
+	}
+	if (PyUnicode_Compare(wide, utf8)) {
+		Py_DECREF(wide);
+		Py_DECREF(utf8);
+		if (PyErr_Occurred())
+			return NULL;
+		return raiseTestError("test_widechar",
+				      "wide string and utf8 string "
+				      "are different");
+	}
+
+	Py_DECREF(wide);
+	Py_DECREF(utf8);
+	Py_RETURN_NONE;
+}
+
 static PyObject *
 test_empty_argparse(PyObject *self)
 {
@@ -1206,6 +1250,7 @@ static PyMethodDef TestMethods[] = {
 	{"test_s_code",		(PyCFunction)test_s_code,	 METH_NOARGS},
 	{"test_u_code",		(PyCFunction)test_u_code,	 METH_NOARGS},
 	{"test_Z_code",		(PyCFunction)test_Z_code,	 METH_NOARGS},
+ 	{"test_widechar",	(PyCFunction)test_widechar,	 METH_NOARGS},
 #ifdef WITH_THREAD
 	{"_test_thread_state",  test_thread_state, 		 METH_VARARGS},
 	{"_pending_threadfunc",	pending_threadfunc,		 METH_VARARGS},
diff --git a/Objects/unicodeobject.c b/Objects/unicodeobject.c
@@ -561,6 +561,66 @@ PyObject *PyUnicode_FromString(const char *u)
 
 #ifdef HAVE_WCHAR_H
 
+#if (Py_UNICODE_SIZE == 2) && defined(SIZEOF_WCHAR_T) && (SIZEOF_WCHAR_T == 4)
+# define CONVERT_WCHAR_TO_SURROGATES
+#endif
+
+#ifdef CONVERT_WCHAR_TO_SURROGATES
+
+/* Here sizeof(wchar_t) is 4 but Py_UNICODE_SIZE == 2, so we need
+   to convert from UTF32 to UTF16. */
+
+PyObject *PyUnicode_FromWideChar(register const wchar_t *w,
+                                 Py_ssize_t size)
+{
+    PyUnicodeObject *unicode;
+    register Py_ssize_t i;
+    Py_ssize_t alloc;
+    const wchar_t *orig_w;
+
+    if (w == NULL) {
+        if (size == 0)
+            return PyUnicode_FromStringAndSize(NULL, 0);
+        PyErr_BadInternalCall();
+        return NULL;
+    }
+
+    if (size == -1) {
+        size = wcslen(w);
+    }
+
+    alloc = size;
+    orig_w = w;
+    for (i = size; i > 0; i--) {
+        if (*w > 0xFFFF)
+            alloc++;
+        w++;
+    }
+    w = orig_w;
+    unicode = _PyUnicode_New(alloc);
+    if (!unicode)
+        return NULL;
+
+    /* Copy the wchar_t data into the new object */
+    {
+        register Py_UNICODE *u;
+        u = PyUnicode_AS_UNICODE(unicode);
+        for (i = size; i > 0; i--) {
+            if (*w > 0xFFFF) {
+                wchar_t ordinal = *w++;
+                ordinal -= 0x10000;
+                *u++ = 0xD800 | (ordinal >> 10);
+                *u++ = 0xDC00 | (ordinal & 0x3FF);
+            }
+            else
+                *u++ = *w++;
+        }
+    }
+    return (PyObject *)unicode;
+}
+
+#else
+
 PyObject *PyUnicode_FromWideChar(register const wchar_t *w,
                                  Py_ssize_t size)
 {
@@ -597,6 +657,10 @@ PyObject *PyUnicode_FromWideChar(register const wchar_t *w,
     return (PyObject *)unicode;
 }
 
+#endif /* CONVERT_WCHAR_TO_SURROGATES */
+
+#undef CONVERT_WCHAR_TO_SURROGATES
+
 static void
 makefmt(char *fmt, int longflag, int size_tflag, int zeropad, int width, int precision, char c)
 {
