Issue #26719: More efficient formatting of ints and floats in json.

This commit is contained in:
Serhiy Storchaka 2016-04-10 14:41:19 +03:00
parent fc43511867
commit e0805cf10e
2 changed files with 14 additions and 62 deletions

View File

@ -32,7 +32,6 @@ for i in range(0x20):
#ESCAPE_DCT.setdefault(chr(i), '\\u%04x' % (i,)) #ESCAPE_DCT.setdefault(chr(i), '\\u%04x' % (i,))
INFINITY = float('inf') INFINITY = float('inf')
FLOAT_REPR = repr
def py_encode_basestring(s): def py_encode_basestring(s):
"""Return a JSON representation of a Python string """Return a JSON representation of a Python string
@ -221,7 +220,7 @@ class JSONEncoder(object):
_encoder = encode_basestring _encoder = encode_basestring
def floatstr(o, allow_nan=self.allow_nan, def floatstr(o, allow_nan=self.allow_nan,
_repr=FLOAT_REPR, _inf=INFINITY, _neginf=-INFINITY): _repr=float.__repr__, _inf=INFINITY, _neginf=-INFINITY):
# Check for specials. Note that this type of test is processor # Check for specials. Note that this type of test is processor
# and/or platform-specific, so do tests which don't depend on the # and/or platform-specific, so do tests which don't depend on the
# internals. # internals.
@ -268,6 +267,7 @@ def _make_iterencode(markers, _default, _encoder, _indent, _floatstr,
list=list, list=list,
str=str, str=str,
tuple=tuple, tuple=tuple,
_intstr=int.__str__,
): ):
if _indent is not None and not isinstance(_indent, str): if _indent is not None and not isinstance(_indent, str):
@ -309,10 +309,10 @@ def _make_iterencode(markers, _default, _encoder, _indent, _floatstr,
# Subclasses of int/float may override __str__, but we still # Subclasses of int/float may override __str__, but we still
# want to encode them as integers/floats in JSON. One example # want to encode them as integers/floats in JSON. One example
# within the standard library is IntEnum. # within the standard library is IntEnum.
yield buf + str(int(value)) yield buf + _intstr(value)
elif isinstance(value, float): elif isinstance(value, float):
# see comment above for int # see comment above for int
yield buf + _floatstr(float(value)) yield buf + _floatstr(value)
else: else:
yield buf yield buf
if isinstance(value, (list, tuple)): if isinstance(value, (list, tuple)):
@ -359,7 +359,7 @@ def _make_iterencode(markers, _default, _encoder, _indent, _floatstr,
# also allow them. Many encoders seem to do something like this. # also allow them. Many encoders seem to do something like this.
elif isinstance(key, float): elif isinstance(key, float):
# see comment for int/float in _make_iterencode # see comment for int/float in _make_iterencode
key = _floatstr(float(key)) key = _floatstr(key)
elif key is True: elif key is True:
key = 'true' key = 'true'
elif key is False: elif key is False:
@ -368,7 +368,7 @@ def _make_iterencode(markers, _default, _encoder, _indent, _floatstr,
key = 'null' key = 'null'
elif isinstance(key, int): elif isinstance(key, int):
# see comment for int/float in _make_iterencode # see comment for int/float in _make_iterencode
key = str(int(key)) key = _intstr(key)
elif _skipkeys: elif _skipkeys:
continue continue
else: else:
@ -389,10 +389,10 @@ def _make_iterencode(markers, _default, _encoder, _indent, _floatstr,
yield 'false' yield 'false'
elif isinstance(value, int): elif isinstance(value, int):
# see comment for int/float in _make_iterencode # see comment for int/float in _make_iterencode
yield str(int(value)) yield _intstr(value)
elif isinstance(value, float): elif isinstance(value, float):
# see comment for int/float in _make_iterencode # see comment for int/float in _make_iterencode
yield _floatstr(float(value)) yield _floatstr(value)
else: else:
if isinstance(value, (list, tuple)): if isinstance(value, (list, tuple)):
chunks = _iterencode_list(value, _current_indent_level) chunks = _iterencode_list(value, _current_indent_level)
@ -419,10 +419,10 @@ def _make_iterencode(markers, _default, _encoder, _indent, _floatstr,
yield 'false' yield 'false'
elif isinstance(o, int): elif isinstance(o, int):
# see comment for int/float in _make_iterencode # see comment for int/float in _make_iterencode
yield str(int(o)) yield _intstr(o)
elif isinstance(o, float): elif isinstance(o, float):
# see comment for int/float in _make_iterencode # see comment for int/float in _make_iterencode
yield _floatstr(float(o)) yield _floatstr(o)
elif isinstance(o, (list, tuple)): elif isinstance(o, (list, tuple)):
yield from _iterencode_list(o, _current_indent_level) yield from _iterencode_list(o, _current_indent_level)
elif isinstance(o, dict): elif isinstance(o, dict):

View File

@ -116,8 +116,6 @@ raise_errmsg(char *msg, PyObject *s, Py_ssize_t end);
static PyObject * static PyObject *
encoder_encode_string(PyEncoderObject *s, PyObject *obj); encoder_encode_string(PyEncoderObject *s, PyObject *obj);
static PyObject * static PyObject *
encoder_encode_long(PyEncoderObject* s UNUSED, PyObject *obj);
static PyObject *
encoder_encode_float(PyEncoderObject *s, PyObject *obj); encoder_encode_float(PyEncoderObject *s, PyObject *obj);
#define S_CHAR(c) (c >= ' ' && c <= '~' && c != '\\' && c != '"') #define S_CHAR(c) (c >= ' ' && c <= '~' && c != '\\' && c != '"')
@ -1444,39 +1442,10 @@ _encoded_const(PyObject *obj)
} }
} }
static PyObject *
encoder_encode_long(PyEncoderObject* s UNUSED, PyObject *obj)
{
/* Return the JSON representation of a PyLong and PyLong subclasses.
Calls int() on PyLong subclasses in case the str() was changed.
Added specifically to deal with IntEnum. See Issue18264. */
PyObject *encoded, *longobj;
if (PyLong_CheckExact(obj)) {
encoded = PyObject_Str(obj);
}
else {
longobj = PyNumber_Long(obj);
if (longobj == NULL) {
PyErr_SetString(
PyExc_ValueError,
"Unable to coerce int subclass to int"
);
return NULL;
}
encoded = PyObject_Str(longobj);
Py_DECREF(longobj);
}
return encoded;
}
static PyObject * static PyObject *
encoder_encode_float(PyEncoderObject *s, PyObject *obj) encoder_encode_float(PyEncoderObject *s, PyObject *obj)
{ {
/* Return the JSON representation of a PyFloat. /* Return the JSON representation of a PyFloat. */
Modified to call float() on float subclasses in case the subclass
changes the repr. See Issue18264. */
PyObject *encoded, *floatobj;
double i = PyFloat_AS_DOUBLE(obj); double i = PyFloat_AS_DOUBLE(obj);
if (!Py_IS_FINITE(i)) { if (!Py_IS_FINITE(i)) {
if (!s->allow_nan) { if (!s->allow_nan) {
@ -1496,24 +1465,7 @@ encoder_encode_float(PyEncoderObject *s, PyObject *obj)
return PyUnicode_FromString("NaN"); return PyUnicode_FromString("NaN");
} }
} }
/* coerce float subclasses to float (primarily for Enum) */ return PyFloat_Type.tp_repr(obj);
if (PyFloat_CheckExact(obj)) {
/* Use a better float format here? */
encoded = PyObject_Repr(obj);
}
else {
floatobj = PyNumber_Float(obj);
if (floatobj == NULL) {
PyErr_SetString(
PyExc_ValueError,
"Unable to coerce float subclass to float"
);
return NULL;
}
encoded = PyObject_Repr(floatobj);
Py_DECREF(floatobj);
}
return encoded;
} }
static PyObject * static PyObject *
@ -1557,7 +1509,7 @@ encoder_listencode_obj(PyEncoderObject *s, _PyAccu *acc,
return _steal_accumulate(acc, encoded); return _steal_accumulate(acc, encoded);
} }
else if (PyLong_Check(obj)) { else if (PyLong_Check(obj)) {
PyObject *encoded = encoder_encode_long(s, obj); PyObject *encoded = PyLong_Type.tp_str(obj);
if (encoded == NULL) if (encoded == NULL)
return -1; return -1;
return _steal_accumulate(acc, encoded); return _steal_accumulate(acc, encoded);
@ -1722,7 +1674,7 @@ encoder_listencode_dict(PyEncoderObject *s, _PyAccu *acc,
goto bail; goto bail;
} }
else if (PyLong_Check(key)) { else if (PyLong_Check(key)) {
kstr = encoder_encode_long(s, key); kstr = PyLong_Type.tp_str(key);
if (kstr == NULL) { if (kstr == NULL) {
goto bail; goto bail;
} }