Issue #23206: Make ``json.dumps(..., ensure_ascii=False)`` as fast as the default...

diff --git a/Lib/json/encoder.py b/Lib/json/encoder.py
index 05138383bccf8749a2fd3d1c7eb0405f5e825cb7..26e9eb2bc69255a400a3c48686a1f821cab562db 100644 (file)
--- a/Lib/json/encoder.py
+++ b/Lib/json/encoder.py
@@ -6,6 +6,10 @@ try:
     from _json import encode_basestring_ascii as c_encode_basestring_ascii
 except ImportError:
     c_encode_basestring_ascii = None
+try:
+    from _json import encode_basestring as c_encode_basestring
+except ImportError:
+    c_encode_basestring = None
 try:
     from _json import make_encoder as c_make_encoder
 except ImportError:
@@ -30,7 +34,7 @@ for i in range(0x20):
 INFINITY = float('inf')
 FLOAT_REPR = repr
 
-def encode_basestring(s):
+def py_encode_basestring(s):
     """Return a JSON representation of a Python string
 
     """
@@ -39,6 +43,9 @@ def encode_basestring(s):
     return '"' + ESCAPE.sub(replace, s) + '"'
 
 
+encode_basestring = (c_encode_basestring or py_encode_basestring)
+
+
 def py_encode_basestring_ascii(s):
     """Return an ASCII-only JSON representation of a Python string
 

diff --git a/Lib/test/test_json/test_encode_basestring_ascii.py b/Lib/test/test_json/test_encode_basestring_ascii.py
index 480afd686e453d7178c6b82e264ec9760faca210..3a4ad1816aeca96a2dabe985a3b0dfa4596d0d34 100644 (file)
--- a/Lib/test/test_json/test_encode_basestring_ascii.py
+++ b/Lib/test/test_json/test_encode_basestring_ascii.py
@@ -11,9 +11,6 @@ CASES = [
     (' s p a c e d ', '" s p a c e d "'),
     ('\U0001d120', '"\\ud834\\udd20"'),
     ('\u03b1\u03a9', '"\\u03b1\\u03a9"'),
-    ('\u03b1\u03a9', '"\\u03b1\\u03a9"'),
-    ('\u03b1\u03a9', '"\\u03b1\\u03a9"'),
-    ('\u03b1\u03a9', '"\\u03b1\\u03a9"'),
     ("`1~!@#$%^&*()_+-={':[,]}|;.</>?", '"`1~!@#$%^&*()_+-={\':[,]}|;.</>?"'),
     ('\x08\x0c\n\r\t', '"\\b\\f\\n\\r\\t"'),
     ('\u0123\u4567\u89ab\ucdef\uabcd\uef4a', '"\\u0123\\u4567\\u89ab\\ucdef\\uabcd\\uef4a"'),

diff --git a/Misc/ACKS b/Misc/ACKS
index 42ff010891467e03feb3fe68e7e40367407a909e..95f12ef05d621d519465e88862294e5e2e77aef3 100644 (file)
--- a/Misc/ACKS
+++ b/Misc/ACKS
@@ -626,6 +626,7 @@ Ali Ikinci
 Aaron Iles
 Lars Immisch
 Bobby Impollonia
+Naoki Inada
 Meador Inge
 Peter Ingebretson
 Tony Ingraldi

diff --git a/Misc/NEWS b/Misc/NEWS
index 9a10abff6a9496aa1f1ae782d37649b5a93a8244..3c71d9e9fc59520879c98ff340184c38ab17d9b4 100644 (file)
--- a/Misc/NEWS
+++ b/Misc/NEWS
@@ -203,6 +203,9 @@ Core and Builtins
 Library
 -------
 
+- Issue #23206: Make ``json.dumps(..., ensure_ascii=False)`` as fast as the
+  default case of ``ensure_ascii=True``.  Patch by Naoki Inada.
+
 - Issue #23185: Add math.inf and math.nan constants.
 
 - Issue #23186: Add ssl.SSLObject.shared_ciphers() and

diff --git a/Modules/_json.c b/Modules/_json.c
index 1580ee6b658d3f91453562e69e9704486aec04a0..94309905e38a2bb835c6efdeb226e5a0a0528385 100644 (file)
--- a/Modules/_json.c
+++ b/Modules/_json.c
@@ -47,7 +47,7 @@ typedef struct _PyEncoderObject {
     PyObject *item_separator;
     PyObject *sort_keys;
     PyObject *skipkeys;
-    int fast_encode;
+    PyCFunction fast_encode;
     int allow_nan;
 } PyEncoderObject;
 
@@ -218,6 +218,97 @@ ascii_escape_unicode(PyObject *pystr)
     return rval;
 }
 
+static PyObject *
+escape_unicode(PyObject *pystr)
+{
+    /* Take a PyUnicode pystr and return a new escaped PyUnicode */
+    Py_ssize_t i;
+    Py_ssize_t input_chars;
+    Py_ssize_t output_size;
+    Py_ssize_t chars;
+    PyObject *rval;
+    void *input;
+    int kind;
+    Py_UCS4 maxchar;
+
+    if (PyUnicode_READY(pystr) == -1)
+        return NULL;
+
+    maxchar = PyUnicode_MAX_CHAR_VALUE(pystr);
+    input_chars = PyUnicode_GET_LENGTH(pystr);
+    input = PyUnicode_DATA(pystr);
+    kind = PyUnicode_KIND(pystr);
+
+    /* Compute the output size */
+    for (i = 0, output_size = 2; i < input_chars; i++) {
+        Py_UCS4 c = PyUnicode_READ(kind, input, i);
+        switch (c) {
+        case '\\': case '"': case '\b': case '\f':
+        case '\n': case '\r': case '\t':
+            output_size += 2;
+            break;
+        default:
+            if (c <= 0x1f)
+                output_size += 6;
+            else
+                output_size++;
+        }
+    }
+
+    rval = PyUnicode_New(output_size, maxchar);
+    if (rval == NULL)
+        return NULL;
+
+    kind = PyUnicode_KIND(rval);
+
+#define ENCODE_OUTPUT do { \
+        chars = 0; \
+        output[chars++] = '"'; \
+        for (i = 0; i < input_chars; i++) { \
+            Py_UCS4 c = PyUnicode_READ(kind, input, i); \
+            switch (c) { \
+            case '\\': output[chars++] = '\\'; output[chars++] = c; break; \
+            case '"':  output[chars++] = '\\'; output[chars++] = c; break; \
+            case '\b': output[chars++] = '\\'; output[chars++] = 'b'; break; \
+            case '\f': output[chars++] = '\\'; output[chars++] = 'f'; break; \
+            case '\n': output[chars++] = '\\'; output[chars++] = 'n'; break; \
+            case '\r': output[chars++] = '\\'; output[chars++] = 'r'; break; \
+            case '\t': output[chars++] = '\\'; output[chars++] = 't'; break; \
+            default: \
+                if (c <= 0x1f) { \
+                    output[chars++] = '\\'; \
+                    output[chars++] = 'u'; \
+                    output[chars++] = '0'; \
+                    output[chars++] = '0'; \
+                    output[chars++] = Py_hexdigits[(c >> 4) & 0xf]; \
+                    output[chars++] = Py_hexdigits[(c     ) & 0xf]; \
+                } else { \
+                    output[chars++] = c; \
+                } \
+            } \
+        } \
+        output[chars++] = '"'; \
+    } while (0)
+
+    if (kind == PyUnicode_1BYTE_KIND) {
+        Py_UCS1 *output = PyUnicode_1BYTE_DATA(rval);
+        ENCODE_OUTPUT;
+    } else if (kind == PyUnicode_2BYTE_KIND) {
+        Py_UCS2 *output = PyUnicode_2BYTE_DATA(rval);
+        ENCODE_OUTPUT;
+    } else {
+        Py_UCS4 *output = PyUnicode_4BYTE_DATA(rval);
+        assert(kind == PyUnicode_4BYTE_KIND);
+        ENCODE_OUTPUT;
+    }
+#undef ENCODE_OUTPUT
+
+#ifdef Py_DEBUG
+    assert(_PyUnicode_CheckConsistency(rval, 1));
+#endif
+    return rval;
+}
+
 static void
 raise_errmsg(char *msg, PyObject *s, Py_ssize_t end)
 {
@@ -530,6 +621,31 @@ py_encode_basestring_ascii(PyObject* self UNUSED, PyObject *pystr)
     return rval;
 }
 
+
+PyDoc_STRVAR(pydoc_encode_basestring,
+    "encode_basestring(string) -> string\n"
+    "\n"
+    "Return a JSON representation of a Python string"
+);
+
+static PyObject *
+py_encode_basestring(PyObject* self UNUSED, PyObject *pystr)
+{
+    PyObject *rval;
+    /* Return a JSON representation of a Python string */
+    /* METH_O */
+    if (PyUnicode_Check(pystr)) {
+        rval = escape_unicode(pystr);
+    }
+    else {
+        PyErr_Format(PyExc_TypeError,
+                     "first argument must be a string, not %.80s",
+                     Py_TYPE(pystr)->tp_name);
+        return NULL;
+    }
+    return rval;
+}
+
 static void
 scanner_dealloc(PyObject *self)
 {
@@ -1223,7 +1339,14 @@ encoder_init(PyObject *self, PyObject *args, PyObject *kwds)
     s->item_separator = item_separator;
     s->sort_keys = sort_keys;
     s->skipkeys = skipkeys;
-    s->fast_encode = (PyCFunction_Check(s->encoder) && PyCFunction_GetFunction(s->encoder) == (PyCFunction)py_encode_basestring_ascii);
+    s->fast_encode = NULL;
+    if (PyCFunction_Check(s->encoder)) {
+        PyCFunction f = PyCFunction_GetFunction(s->encoder);
+        if (f == (PyCFunction)py_encode_basestring_ascii ||
+                f == (PyCFunction)py_encode_basestring) {
+            s->fast_encode = f;
+        }
+    }
     s->allow_nan = PyObject_IsTrue(allow_nan);
 
     Py_INCREF(s->markers);
@@ -1372,7 +1495,7 @@ encoder_encode_string(PyEncoderObject *s, PyObject *obj)
 {
     /* Return the JSON representation of a string */
     if (s->fast_encode)
-        return py_encode_basestring_ascii(NULL, obj);
+        return s->fast_encode(NULL, obj);
     else
         return PyObject_CallFunctionObjArgs(s->encoder, obj, NULL);
 }
@@ -1840,6 +1963,10 @@ static PyMethodDef speedups_methods[] = {
         (PyCFunction)py_encode_basestring_ascii,
         METH_O,
         pydoc_encode_basestring_ascii},
+    {"encode_basestring",
+        (PyCFunction)py_encode_basestring,
+        METH_O,
+        pydoc_encode_basestring},
     {"scanstring",
         (PyCFunction)py_scanstring,
         METH_VARARGS,