# HG changeset patch
# User Fredrik Lundh <fredrik@effbot.org>
# Date 1257469909 -3600
# Node ID 3bfa8cf10f8811a23a8dca44c33284b3e1fdeb38
# Parent  c20be1d34e7eb5482f2e22041cabd750e60f8e52
Support arbitrary default encodings.  Use optimized code paths for
ascii, iso-8859-1, latin-1, us-ascii, and utf-8.

diff -r c20be1d34e7eb5482f2e22041cabd750e60f8e52 -r 3bfa8cf10f8811a23a8dca44c33284b3e1fdeb38 cjson.c
--- a/cjson.c	Fri Nov 06 00:37:07 2009 +0100
+++ b/cjson.c	Fri Nov 06 02:11:49 2009 +0100
@@ -21,13 +21,15 @@
     int  all_unicode; // make all output strings unicode if true
 } JSONData;
 
+
 typedef struct JSONEncode {
-    int  utf8; // true if byte strings are utf-8
+    enum { LATIN1, UTF8, OTHER } encoding;
 } JSONEncode;
 
 static PyObject* encode_object(JSONEncode *encode, PyObject *object);
 static PyObject* encode_string_latin1(PyObject *object);
 static PyObject* encode_string_utf8(PyObject *object);
+static PyObject* encode_string_other(PyObject *object);
 static PyObject* encode_unicode(PyObject *object);
 static PyObject* encode_tuple(JSONEncode *encode, PyObject *object);
 static PyObject* encode_list(JSONEncode *encode, PyObject *object);
@@ -175,6 +177,8 @@
 
     len = ptr - jsondata->ptr - 1;
 
+    /* TODO(effbot): DecodeEscape doesn't handle escaped / correctly,
+       so should be replaced with custom code. */
     if (has_unicode || jsondata->all_unicode)
         object = PyUnicode_DecodeUnicodeEscape(jsondata->ptr+1, len, NULL);
     else if (string_escape)
@@ -676,7 +680,7 @@
     }
 }
 
-/* Same, but used when defaultencoding = utf-8 */
+/* Same, but for UTF-8 */
 
 static PyObject*
 encode_string_utf8(PyObject *string)
@@ -758,6 +762,30 @@
     }
 }
 
+/* Same, but for other default encodings. */
+
+static PyObject*
+encode_string_other(PyObject *string)
+{
+    PyObject* unicode_string;
+    PyObject* result;
+
+    unicode_string = PyUnicode_Decode(
+        PyString_AS_STRING(string), PyString_GET_SIZE(string),
+        NULL, /* = use default encoding */
+        "replace");
+
+    if (!unicode_string)
+        return NULL;
+
+    result = encode_unicode(unicode_string);
+
+    Py_DECREF(unicode_string);
+
+    return result;
+}
+
+
 /*
  * This function is an almost verbatim copy of unicodeescape_string() from
  * Python's unicodeobject.c with the following differences:
@@ -1165,10 +1193,14 @@
     } else if (object == Py_None) {
         return PyString_FromString("null");
     } else if (PyString_Check(object)) {
-        if (encode->utf8)
+        switch (encode->encoding) {
+        case LATIN1:
+            return encode_string_latin1(object);
+        case UTF8:
             return encode_string_utf8(object);
-        else
-            return encode_string_latin1(object);
+        default:
+            return encode_string_other(object);
+        }
     } else if (PyUnicode_Check(object)) {
         return encode_unicode(object);
     } else if (PyInt_Check(object)) {
@@ -1214,8 +1246,27 @@
 {
     JSONEncode encode;
     const char *encoding = PyUnicode_GetDefaultEncoding();
-
-    encode.utf8 = encoding[0] == 'u' && (strcmp(encoding, "utf-8") == 0);
+    encode.encoding = OTHER;
+    switch (*encoding) {
+    case 'a':
+        if (strcmp(encoding, "ascii") == 0)
+            encode.encoding = LATIN1; /* ascii is a subset */
+        break;
+    case 'i':
+        if (strcmp(encoding, "iso-8859-1") == 0)
+            encode.encoding = LATIN1;
+        break;
+    case 'l':
+        if (strcmp(encoding, "latin-1") == 0)
+            encode.encoding = LATIN1;
+        break;
+    case 'u':
+        if (strcmp(encoding, "utf-8") == 0)
+            encode.encoding = UTF8;
+        else if (strcmp(encoding, "us-ascii") == 0)
+            encode.encoding = LATIN1;
+        break;
+    }
     return encode_object(&encode, object);
 }
 
diff -r c20be1d34e7eb5482f2e22041cabd750e60f8e52 -r 3bfa8cf10f8811a23a8dca44c33284b3e1fdeb38 jsontest.py
--- a/jsontest.py	Fri Nov 06 00:37:07 2009 +0100
+++ b/jsontest.py	Fri Nov 06 02:11:49 2009 +0100
@@ -300,14 +300,24 @@
         campus_utf8 = campus.encode("utf-8")
         campus_raw = campus_utf8.decode("iso-8859-1")
 
+        island = u'\xe5 i \xe5a \xe4 e \xf6'
+
         sys.setdefaultencoding("ascii")
         self.assertEqual(cjson.encode(campus_utf8), cjson.encode(campus_raw))
 
         sys.setdefaultencoding("iso-8859-1")
         self.assertEqual(cjson.encode(campus_utf8), cjson.encode(campus_raw))
+        self.assertEqual(cjson.encode(island.encode("iso-8859-1")),
+                         cjson.encode(island))
+
+        sys.setdefaultencoding("cp850")
+        self.assertEqual(cjson.encode(island.encode("cp850")),
+                         cjson.encode(island))
 
         sys.setdefaultencoding("utf-8")
         self.assertEqual(cjson.encode(campus_utf8), cjson.encode(campus))
+        self.assertEqual(cjson.encode(island.encode("utf-8")),
+                         cjson.encode(island))
 
         # full BMP roundtrip
         for i in range(0xd800) + range(0xe000, 0xffff):