Commits

Antonio Cuni committed 4d172e0 Merge

hg merge default

Comments (0)

Files changed (8)

lib-python/2.7/json/__init__.py

 
 __author__ = 'Bob Ippolito <bob@redivi.com>'
 
+try:
+    # PyPy speedup, the interface is different than CPython's _json
+    import _pypyjson
+except ImportError:
+    _pypyjson = None
+
 from .decoder import JSONDecoder
 from .encoder import JSONEncoder
 
 _default_decoder = JSONDecoder(encoding=None, object_hook=None,
                                object_pairs_hook=None)
 
-
 def load(fp, encoding=None, cls=None, object_hook=None, parse_float=None,
         parse_int=None, parse_constant=None, object_pairs_hook=None, **kw):
     """Deserialize ``fp`` (a ``.read()``-supporting file-like object containing
     if (cls is None and encoding is None and object_hook is None and
             parse_int is None and parse_float is None and
             parse_constant is None and object_pairs_hook is None and not kw):
-        return _default_decoder.decode(s)
+        if _pypyjson and not isinstance(s, unicode):
+            return _pypyjson.loads(s)
+        else:
+            return _default_decoder.decode(s)
     if cls is None:
         cls = JSONDecoder
     if object_hook is not None:

pypy/config/pypyoption.py

      "thread", "itertools", "pyexpat", "_ssl", "cpyext", "array",
      "binascii", "_multiprocessing", '_warnings',
      "_collections", "_multibytecodec", "micronumpy", "_ffi",
-     "_continuation", "_cffi_backend", "_csv", "cppyy"]
+     "_continuation", "_cffi_backend", "_csv", "cppyy", "_pypyjson"]
 ))
 
 translation_modules = default_modules.copy()

pypy/doc/config/objspace.usemodules._pypyjson.txt

+RPython speedups for the stdlib json module

pypy/module/_pypyjson/__init__.py

+from pypy.interpreter.mixedmodule import MixedModule
+
+class Module(MixedModule):
+    """fast json implementation"""
+
+    appleveldefs = {}
+
+    interpleveldefs = {
+        'loads' : 'interp_decoder.loads',
+        }

pypy/module/_pypyjson/interp_decoder.py

+import sys
+import math
+from rpython.rlib.rstring import StringBuilder
+from rpython.rlib.objectmodel import specialize
+from rpython.rlib import rfloat
+from rpython.rtyper.lltypesystem import lltype, rffi
+from pypy.interpreter.error import OperationError, operationerrfmt
+from pypy.interpreter.gateway import unwrap_spec
+from pypy.interpreter import unicodehelper
+from rpython.rtyper.annlowlevel import llstr, hlunicode
+
+OVF_DIGITS = len(str(sys.maxint))
+
+def is_whitespace(ch):
+    return ch == ' ' or ch == '\t' or ch == '\r' or ch == '\n'
+
+# precomputing negative powers of 10 is MUCH faster than using e.g. math.pow
+# at runtime
+NEG_POW_10 = [10.0**-i for i in range(16)]
+def neg_pow_10(x, exp):
+    if exp >= len(NEG_POW_10):
+        return 0.0
+    return x * NEG_POW_10[exp]
+
+def strslice2unicode_latin1(s, start, end):
+    """
+    Convert s[start:end] to unicode. s is supposed to be an RPython string
+    encoded in latin-1, which means that the numeric value of each char is the
+    same as the corresponding unicode code point.
+
+    Internally it's implemented at the level of low-level helpers, to avoid
+    the extra copy we would need if we take the actual slice first.
+    
+    No bound checking is done, use carefully.
+    """
+    from rpython.rtyper.annlowlevel import llstr, hlunicode
+    from rpython.rtyper.lltypesystem.rstr import malloc, UNICODE
+    from rpython.rtyper.lltypesystem.lltype import cast_primitive, UniChar
+    length = end-start
+    ll_s = llstr(s)
+    ll_res = malloc(UNICODE, length)
+    ll_res.hash = 0
+    for i in range(length):
+        ch = ll_s.chars[start+i]
+        ll_res.chars[i] = cast_primitive(UniChar, ch)
+    return hlunicode(ll_res)
+
+TYPE_UNKNOWN = 0
+TYPE_STRING = 1
+class JSONDecoder(object):
+    def __init__(self, space, s):
+        self.space = space
+        self.s = s
+        # we put our string in a raw buffer so:
+        # 1) we automatically get the '\0' sentinel at the end of the string,
+        #    which means that we never have to check for the "end of string"
+        # 2) we can pass the buffer directly to strtod
+        self.ll_chars = rffi.str2charp(s)
+        self.end_ptr = lltype.malloc(rffi.CCHARPP.TO, 1, flavor='raw')
+        self.pos = 0
+        self.last_type = TYPE_UNKNOWN
+
+    def close(self):
+        rffi.free_charp(self.ll_chars)
+        lltype.free(self.end_ptr, flavor='raw')
+
+    def getslice(self, start, end):
+        assert start >= 0
+        assert end >= 0
+        return self.s[start:end]
+
+    def skip_whitespace(self, i):
+        while True:
+            ch = self.ll_chars[i]
+            if is_whitespace(ch):
+                i+=1
+            else:
+                break
+        return i
+
+    @specialize.arg(1)
+    def _raise(self, msg, *args):
+        raise operationerrfmt(self.space.w_ValueError, msg, *args)
+
+    def decode_any(self, i):
+        i = self.skip_whitespace(i)
+        ch = self.ll_chars[i]
+        if ch == '"':
+            return self.decode_string(i+1)
+        elif ch == '[':
+            return self.decode_array(i+1)
+        elif ch == '{':
+            return self.decode_object(i+1)
+        elif ch == 'n':
+            return self.decode_null(i+1)
+        elif ch == 't':
+            return self.decode_true(i+1)
+        elif ch == 'f':
+            return self.decode_false(i+1)
+        elif ch == 'I':
+            return self.decode_infinity(i+1)
+        elif ch == 'N':
+            return self.decode_nan(i+1)
+        elif ch == '-':
+            if self.ll_chars[i+1] == 'I':
+                return self.decode_infinity(i+2, sign=-1)
+            return self.decode_numeric(i)
+        elif ch.isdigit():
+            return self.decode_numeric(i)
+        else:
+            self._raise("No JSON object could be decoded: unexpected '%s' at char %d",
+                        ch, self.pos)
+
+    def decode_null(self, i):
+        if (self.ll_chars[i]   == 'u' and
+            self.ll_chars[i+1] == 'l' and
+            self.ll_chars[i+2] == 'l'):
+            self.pos = i+3
+            return self.space.w_None
+        self._raise("Error when decoding null at char %d", i)
+
+    def decode_true(self, i):
+        if (self.ll_chars[i]   == 'r' and
+            self.ll_chars[i+1] == 'u' and
+            self.ll_chars[i+2] == 'e'):
+            self.pos = i+3
+            return self.space.w_True
+        self._raise("Error when decoding true at char %d", i)
+
+    def decode_false(self, i):
+        if (self.ll_chars[i]   == 'a' and
+            self.ll_chars[i+1] == 'l' and
+            self.ll_chars[i+2] == 's' and
+            self.ll_chars[i+3] == 'e'):
+            self.pos = i+4
+            return self.space.w_False
+        self._raise("Error when decoding false at char %d", i)
+
+    def decode_infinity(self, i, sign=1):
+        if (self.ll_chars[i]   == 'n' and
+            self.ll_chars[i+1] == 'f' and
+            self.ll_chars[i+2] == 'i' and
+            self.ll_chars[i+3] == 'n' and
+            self.ll_chars[i+4] == 'i' and
+            self.ll_chars[i+5] == 't' and
+            self.ll_chars[i+6] == 'y'):
+            self.pos = i+7
+            return self.space.wrap(rfloat.INFINITY * sign)
+        self._raise("Error when decoding Infinity at char %d", i)
+
+    def decode_nan(self, i):
+        if (self.ll_chars[i]   == 'a' and
+            self.ll_chars[i+1] == 'N'):
+            self.pos = i+2
+            return self.space.wrap(rfloat.NAN)
+        self._raise("Error when decoding NaN at char %d", i)
+
+    def decode_numeric(self, i):
+        start = i
+        i, ovf_maybe, intval = self.parse_integer(i)
+        #
+        # check for the optional fractional part
+        ch = self.ll_chars[i]
+        if ch == '.':
+            if not self.ll_chars[i+1].isdigit():
+                self._raise("Expected digit at char %d", i+1)
+            return self.decode_float(start)
+        elif ch == 'e' or ch == 'E':
+            return self.decode_float(start)
+        elif ovf_maybe:
+            return self.decode_int_slow(start)
+
+        self.pos = i
+        return self.space.wrap(intval)
+
+    def decode_float(self, i):
+        from rpython.rlib import rdtoa
+        start = rffi.ptradd(self.ll_chars, i)
+        floatval = rdtoa.dg_strtod(start, self.end_ptr)
+        diff = rffi.cast(rffi.LONG, self.end_ptr[0]) - rffi.cast(rffi.LONG, start)
+        self.pos = i + diff
+        return self.space.wrap(floatval)
+
+    def decode_int_slow(self, i):
+        start = i
+        if self.ll_chars[i] == '-':
+            i += 1
+        while self.ll_chars[i].isdigit():
+            i += 1
+        s = self.getslice(start, i)
+        self.pos = i
+        return self.space.call_function(self.space.w_int, self.space.wrap(s))
+
+    def parse_integer(self, i):
+        "Parse a decimal number with an optional minus sign"
+        sign = 1
+        # parse the sign
+        if self.ll_chars[i] == '-':
+            sign = -1
+            i += 1
+        elif self.ll_chars[i] == '+':
+            i += 1
+        #
+        if self.ll_chars[i] == '0':
+            i += 1
+            return i, False, 0
+
+        intval = 0
+        start = i
+        while True:
+            ch = self.ll_chars[i]
+            if ch.isdigit():
+                intval = intval*10 + ord(ch)-ord('0')
+                i += 1
+            else:
+                break
+        count = i - start
+        if count == 0:
+            self._raise("Expected digit at char %d", i)
+        # if the number has more digits than OVF_DIGITS, it might have
+        # overflowed
+        ovf_maybe = (count >= OVF_DIGITS)
+        return i, ovf_maybe, sign * intval
+    parse_integer._always_inline_ = True
+
+    def decode_array(self, i):
+        w_list = self.space.newlist([])
+        start = i
+        count = 0
+        i = self.skip_whitespace(start)
+        if self.ll_chars[i] == ']':
+            self.pos = i+1
+            return w_list
+        #
+        while True:
+            w_item = self.decode_any(i)
+            i = self.pos
+            self.space.call_method(w_list, 'append', w_item)
+            i = self.skip_whitespace(i)
+            ch = self.ll_chars[i]
+            i += 1
+            if ch == ']':
+                self.pos = i
+                return w_list
+            elif ch == ',':
+                pass
+            elif ch == '\0':
+                self._raise("Unterminated array starting at char %d", start)
+            else:
+                self._raise("Unexpected '%s' when decoding array (char %d)",
+                            ch, self.pos)
+
+    def decode_object(self, i):
+        start = i
+        w_dict = self.space.newdict()
+        #
+        i = self.skip_whitespace(i)
+        if self.ll_chars[i] == '}':
+            self.pos = i+1
+            return w_dict
+        #
+        while True:
+            # parse a key: value
+            self.last_type = TYPE_UNKNOWN
+            w_name = self.decode_any(i)
+            if self.last_type != TYPE_STRING:
+                self._raise("Key name must be string for object starting at char %d", start)
+            i = self.skip_whitespace(self.pos)
+            ch = self.ll_chars[i]
+            if ch != ':':
+                self._raise("No ':' found at char %d", i)
+            i += 1
+            i = self.skip_whitespace(i)
+            #
+            w_value = self.decode_any(i)
+            self.space.setitem(w_dict, w_name, w_value)
+            i = self.skip_whitespace(self.pos)
+            ch = self.ll_chars[i]
+            i += 1
+            if ch == '}':
+                self.pos = i
+                return w_dict
+            elif ch == ',':
+                pass
+            elif ch == '\0':
+                self._raise("Unterminated object starting at char %d", start)
+            else:
+                self._raise("Unexpected '%s' when decoding object (char %d)",
+                            ch, self.pos)
+
+
+    def decode_string(self, i):
+        start = i
+        bits = 0
+        while True:
+            # this loop is a fast path for strings which do not contain escape
+            # characters
+            ch = self.ll_chars[i]
+            i += 1
+            bits |= ord(ch)
+            if ch == '"':
+                if bits & 0x80:
+                    # the 8th bit is set, it's an utf8 strnig
+                    content_utf8 = self.getslice(start, i-1)
+                    content_unicode = unicodehelper.decode_utf8(self.space, content_utf8)
+                else:
+                    # ascii only, fast path (ascii is a strict subset of
+                    # latin1, and we already checked that all the chars are <
+                    # 128)
+                    content_unicode = strslice2unicode_latin1(self.s, start, i-1)
+                self.last_type = TYPE_STRING
+                self.pos = i
+                return self.space.wrap(content_unicode)
+            elif ch == '\\':
+                content_so_far = self.getslice(start, i-1)
+                self.pos = i-1
+                return self.decode_string_escaped(start, content_so_far)
+            elif ch == '\0':
+                self._raise("Unterminated string starting at char %d", start)
+
+
+    def decode_string_escaped(self, start, content_so_far):
+        builder = StringBuilder(len(content_so_far)*2) # just an estimate
+        builder.append(content_so_far)
+        i = self.pos
+        while True:
+            ch = self.ll_chars[i]
+            i += 1
+            if ch == '"':
+                content_utf8 = builder.build()
+                content_unicode = unicodehelper.decode_utf8(self.space, content_utf8)
+                self.last_type = TYPE_STRING
+                self.pos = i
+                return self.space.wrap(content_unicode)
+            elif ch == '\\':
+                i = self.decode_escape_sequence(i, builder)
+            elif ch == '\0':
+                self._raise("Unterminated string starting at char %d", start)
+            else:
+                builder.append_multiple_char(ch, 1) # we should implement append_char
+
+    def decode_escape_sequence(self, i, builder):
+        ch = self.ll_chars[i]
+        i += 1
+        put = builder.append_multiple_char
+        if ch == '\\':  put('\\', 1)
+        elif ch == '"': put('"' , 1)
+        elif ch == '/': put('/' , 1)
+        elif ch == 'b': put('\b', 1)
+        elif ch == 'f': put('\f', 1)
+        elif ch == 'n': put('\n', 1)
+        elif ch == 'r': put('\r', 1)
+        elif ch == 't': put('\t', 1)
+        elif ch == 'u':
+            return self.decode_escape_sequence_unicode(i, builder)
+        else:
+            self._raise("Invalid \\escape: %s (char %d)", ch, self.pos-1)
+        return i
+
+    def decode_escape_sequence_unicode(self, i, builder):
+        # at this point we are just after the 'u' of the \u1234 sequence.
+        start = i
+        i += 4
+        hexdigits = self.getslice(start, i)
+        try:
+            val = int(hexdigits, 16)
+            if val & 0xfc00 == 0xd800:
+                # surrogate pair
+                val = self.decode_surrogate_pair(i, val)
+                i += 6
+        except ValueError:
+            self._raise("Invalid \uXXXX escape (char %d)", i-1)
+            return # help the annotator to know that we'll never go beyond
+                   # this point
+        #
+        uchr = unichr(val)
+        utf8_ch = unicodehelper.encode_utf8(self.space, uchr)
+        builder.append(utf8_ch)
+        return i
+
+    def decode_surrogate_pair(self, i, highsurr):
+        if self.ll_chars[i] != '\\' or self.ll_chars[i+1] != 'u':
+            self._raise("Unpaired high surrogate at char %d", i)
+        i += 2
+        hexdigits = self.getslice(i, i+4)
+        lowsurr = int(hexdigits, 16) # the possible ValueError is caugth by the caller
+        return 0x10000 + (((highsurr - 0xd800) << 10) | (lowsurr - 0xdc00))
+
+def loads(space, w_s):
+    if space.isinstance_w(w_s, space.w_unicode):
+        raise OperationError(space.w_TypeError,
+                             space.wrap("Expected utf8-encoded str, got unicode"))
+    s = space.str_w(w_s)
+    decoder = JSONDecoder(space, s)
+    try:
+        w_res = decoder.decode_any(0)
+        i = decoder.skip_whitespace(decoder.pos)
+        if i < len(s):
+            start = i
+            end = len(s) - 1
+            raise operationerrfmt(space.w_ValueError, "Extra data: char %d - %d", start, end)
+        return w_res
+    finally:
+        decoder.close()

pypy/module/_pypyjson/targetjson.py

+import sys
+import py
+ROOT = py.path.local(__file__).dirpath('..', '..', '..')
+sys.path.insert(0, str(ROOT))
+
+import time
+from rpython.rlib.streamio import open_file_as_stream
+from pypy.interpreter.error import OperationError
+from pypy.module._pypyjson.interp_decoder import loads
+
+
+
+## MSG = open('msg.json').read()
+
+class W_Root(object):
+    pass
+
+class W_Dict(W_Root):
+    def __init__(self):
+        self.dictval = {}
+
+class W_Unicode(W_Root):
+    def __init__(self, x):
+        self.unival = x
+
+class W_String(W_Root):
+    def __init__(self, x):
+        self.strval = x
+
+class W_Int(W_Root):
+    def __init__(self, x):
+        self.intval = x
+
+class W_Float(W_Root):
+    def __init__(self, x):
+        self.floatval = x
+
+class W_List(W_Root):
+    def __init__(self):
+        self.listval = []
+
+class W_Singleton(W_Root):
+    def __init__(self, name):
+        self.name = name
+
+class FakeSpace(object):
+
+    w_None = W_Singleton('None')
+    w_True = W_Singleton('True')
+    w_False = W_Singleton('False')
+    w_ValueError = W_Singleton('ValueError')
+    w_UnicodeDecodeError = W_Singleton('UnicodeDecodeError')
+    w_unicode = W_Unicode
+    w_int = W_Int
+    w_float = W_Float
+
+    def newtuple(self, items):
+        return None
+
+    def newdict(self):
+        return W_Dict()
+
+    def newlist(self, items):
+        return W_List()
+
+    def isinstance_w(self, w_x, w_type):
+        return isinstance(w_x, w_type)
+
+    def str_w(self, w_x):
+        assert isinstance(w_x, W_String)
+        return w_x.strval
+
+    def call_method(self, obj, name, arg):
+        assert name == 'append'
+        assert isinstance(obj, W_List)
+        obj.listval.append(arg)
+    call_method._dont_inline_ = True
+
+    def call_function(self, w_func, *args_w):
+        return self.w_None # XXX
+
+    def setitem(self, d, key, value):
+        assert isinstance(d, W_Dict)
+        assert isinstance(key, W_Unicode)
+        d.dictval[key.unival] = value
+
+    def wrapunicode(self, x):
+        return W_Unicode(x)
+
+    def wrapint(self, x):
+        return W_Int(x)
+
+    def wrapfloat(self, x):
+        return W_Float(x)
+    
+    def wrap(self, x):
+        if isinstance(x, int):
+            return W_Int(x)
+        elif isinstance(x, float):
+            return W_Float(x)
+        ## elif isinstance(x, str):
+        ##     assert False
+        else:
+            return W_Unicode(unicode(x))
+    wrap._annspecialcase_ = "specialize:argtype(1)"
+
+
+fakespace = FakeSpace()
+
+def myloads(msg):
+    return loads(fakespace, W_String(msg))
+    
+
+def bench(title, N, fn, arg):
+    a = time.clock()
+    for i in range(N):
+        res = fn(arg)
+    b = time.clock()
+    print title, (b-a) / N * 1000
+
+def entry_point(argv):
+    if len(argv) != 3:
+        print 'Usage: %s FILE n' % argv[0]
+        return 1
+    filename = argv[1]
+    N = int(argv[2])
+    f = open_file_as_stream(filename)
+    msg = f.readall()
+    
+    try:
+        bench('loads     ', N, myloads,  msg)
+    except OperationError, e:
+        print 'Error', e._compute_value(fakespace)
+        
+    return 0
+
+# _____ Define and setup target ___
+
+def target(*args):
+    return entry_point, None
+
+if __name__ == '__main__':
+    entry_point(sys.argv)

pypy/module/_pypyjson/test/test__pypyjson.py

+# -*- encoding: utf-8 -*-
+import py
+from pypy.module._pypyjson.interp_decoder import JSONDecoder
+
+def test_skip_whitespace():
+    s = '   hello   '
+    dec = JSONDecoder('fake space', s)
+    assert dec.pos == 0
+    assert dec.skip_whitespace(0) == 3
+    assert dec.skip_whitespace(3) == 3
+    assert dec.skip_whitespace(8) == len(s)
+    dec.close()
+
+    
+
+class AppTest(object):
+    spaceconfig = {"objspace.usemodules._pypyjson": True}
+
+    def test_raise_on_unicode(self):
+        import _pypyjson
+        raises(TypeError, _pypyjson.loads, u"42")
+
+
+    def test_decode_constants(self):
+        import _pypyjson
+        assert _pypyjson.loads('null') is None
+        raises(ValueError, _pypyjson.loads, 'nul')
+        raises(ValueError, _pypyjson.loads, 'nu')
+        raises(ValueError, _pypyjson.loads, 'n')
+        raises(ValueError, _pypyjson.loads, 'nuXX')
+        #
+        assert _pypyjson.loads('true') is True
+        raises(ValueError, _pypyjson.loads, 'tru')
+        raises(ValueError, _pypyjson.loads, 'tr')
+        raises(ValueError, _pypyjson.loads, 't')
+        raises(ValueError, _pypyjson.loads, 'trXX')
+        #
+        assert _pypyjson.loads('false') is False
+        raises(ValueError, _pypyjson.loads, 'fals')
+        raises(ValueError, _pypyjson.loads, 'fal')
+        raises(ValueError, _pypyjson.loads, 'fa')
+        raises(ValueError, _pypyjson.loads, 'f')
+        raises(ValueError, _pypyjson.loads, 'falXX')
+        
+
+    def test_decode_string(self):
+        import _pypyjson
+        res = _pypyjson.loads('"hello"')
+        assert res == u'hello'
+        assert type(res) is unicode
+
+    def test_decode_string_utf8(self):
+        import _pypyjson
+        s = u'àèìòù'
+        res = _pypyjson.loads('"%s"' % s.encode('utf-8'))
+        assert res == s
+
+    def test_skip_whitespace(self):
+        import _pypyjson
+        s = '   "hello"   '
+        assert _pypyjson.loads(s) == u'hello'
+        s = '   "hello"   extra'
+        raises(ValueError, "_pypyjson.loads(s)")
+
+    def test_unterminated_string(self):
+        import _pypyjson
+        s = '"hello' # missing the trailing "
+        raises(ValueError, "_pypyjson.loads(s)")
+
+    def test_escape_sequence(self):
+        import _pypyjson
+        assert _pypyjson.loads(r'"\\"') == u'\\'
+        assert _pypyjson.loads(r'"\""') == u'"'
+        assert _pypyjson.loads(r'"\/"') == u'/'       
+        assert _pypyjson.loads(r'"\b"') == u'\b'
+        assert _pypyjson.loads(r'"\f"') == u'\f'
+        assert _pypyjson.loads(r'"\n"') == u'\n'
+        assert _pypyjson.loads(r'"\r"') == u'\r'
+        assert _pypyjson.loads(r'"\t"') == u'\t'
+
+    def test_escape_sequence_in_the_middle(self):
+        import _pypyjson
+        s = r'"hello\nworld"'
+        assert _pypyjson.loads(s) == "hello\nworld"
+
+    def test_unterminated_string_after_escape_sequence(self):
+        import _pypyjson
+        s = r'"hello\nworld' # missing the trailing "
+        raises(ValueError, "_pypyjson.loads(s)")
+        
+    def test_escape_sequence_unicode(self):
+        import _pypyjson
+        s = r'"\u1234"'
+        assert _pypyjson.loads(s) == u'\u1234'
+
+    def test_invalid_utf_8(self):
+        import _pypyjson
+        s = '"\xe0"' # this is an invalid UTF8 sequence inside a string
+        raises(UnicodeDecodeError, "_pypyjson.loads(s)")
+
+    def test_decode_numeric(self):
+        import sys
+        import _pypyjson
+        def check(s, val):
+            res = _pypyjson.loads(s)
+            assert type(res) is type(val)
+            assert res == val
+        #
+        check('42', 42)
+        check('-42', -42)
+        check('42.123', 42.123)
+        check('42E0', 42.0)
+        check('42E3', 42000.0)
+        check('42E-1', 4.2)
+        check('42E+1', 420.0)
+        check('42.123E3', 42123.0)
+        check('0', 0)
+        check('-0', 0)
+        check('0.123', 0.123)
+        check('0E3', 0.0)
+        check('5E0001', 50.0)
+        check(str(1 << 32), 1 << 32)
+        check(str(1 << 64), 1 << 64)
+        #
+        x = str(sys.maxint+1) + '.123'
+        check(x, float(x))
+        x = str(sys.maxint+1) + 'E1'
+        check(x, float(x))
+        x = str(sys.maxint+1) + 'E-1'
+        check(x, float(x))
+        #
+        check('1E400', float('inf'))
+        ## # these are non-standard but supported by CPython json
+        check('Infinity', float('inf'))
+        check('-Infinity', float('-inf'))
+
+    def test_nan(self):
+        import math
+        import _pypyjson
+        res = _pypyjson.loads('NaN')
+        assert math.isnan(res)
+
+    def test_decode_numeric_invalid(self):
+        import _pypyjson
+        def error(s):
+            raises(ValueError, _pypyjson.loads, s)
+        #
+        error('  42   abc')
+        error('.123')
+        error('+123')
+        error('12.')
+        error('12.-3')
+        error('12E')
+        error('12E-')
+        error('0123') # numbers can't start with 0
+
+    def test_decode_object(self):
+        import _pypyjson
+        assert _pypyjson.loads('{}') == {}
+        assert _pypyjson.loads('{  }') == {}
+        #
+        s = '{"hello": "world", "aaa": "bbb"}'
+        assert _pypyjson.loads(s) == {'hello': 'world',
+                                      'aaa': 'bbb'}
+        raises(ValueError, _pypyjson.loads, '{"key"')
+        raises(ValueError, _pypyjson.loads, '{"key": 42')
+
+    def test_decode_object_nonstring_key(self):
+        import _pypyjson
+        raises(ValueError, "_pypyjson.loads('{42: 43}')")
+        
+    def test_decode_array(self):
+        import _pypyjson
+        assert _pypyjson.loads('[]') == []
+        assert _pypyjson.loads('[  ]') == []
+        assert _pypyjson.loads('[1]') == [1]
+        assert _pypyjson.loads('[1, 2]') == [1, 2]
+        raises(ValueError, "_pypyjson.loads('[1: 2]')")
+        raises(ValueError, "_pypyjson.loads('[1, 2')")
+        raises(ValueError, """_pypyjson.loads('["extra comma",]')""")
+        
+    def test_unicode_surrogate_pair(self):
+        import _pypyjson
+        expected = u'z\U0001d120x'
+        res = _pypyjson.loads('"z\\ud834\\udd20x"')
+        assert res == expected
+
+

pypy/tool/gdb_pypy.py

 
     def invoke(self, arg, from_tty):
         # some magic code to automatically reload the python file while developing
-        ## from pypy.tool import gdb_pypy
-        ## reload(gdb_pypy)
-        ## gdb_pypy.RPyType.prog2typeids = self.prog2typeids # persist the cache
-        ## self.__class__ = gdb_pypy.RPyType
+        from pypy.tool import gdb_pypy
+        reload(gdb_pypy)
+        gdb_pypy.RPyType.prog2typeids = self.prog2typeids # persist the cache
+        self.__class__ = gdb_pypy.RPyType
         print self.do_invoke(arg, from_tty)
 
     def do_invoke(self, arg, from_tty):
-        obj = self.gdb.parse_and_eval(arg)
-        hdr = lookup(obj, '_gcheader')
-        tid = hdr['h_tid']
-        offset = tid & 0xFFFFFFFF # 64bit only
-        offset = int(offset) # convert from gdb.Value to python int
+        try:
+            offset = int(arg)
+        except ValueError:
+            obj = self.gdb.parse_and_eval(arg)
+            hdr = lookup(obj, '_gcheader')
+            tid = hdr['h_tid']
+            offset = tid & 0xFFFFFFFF # 64bit only
+            offset = int(offset) # convert from gdb.Value to python int
+
         typeids = self.get_typeids()
         if offset in typeids:
             return typeids[offset]