349 lines
11 KiB
Python
349 lines
11 KiB
Python
import copy
|
|
import unittest
|
|
from test.test_support import run_unittest, TestFailed, check_warnings
|
|
|
|
|
|
# Fake a number that implements numeric methods through __coerce__
|
|
class CoerceNumber:
|
|
def __init__(self, arg):
|
|
self.arg = arg
|
|
|
|
def __repr__(self):
|
|
return '<CoerceNumber %s>' % repr(self.arg)
|
|
|
|
def __coerce__(self, other):
|
|
if isinstance(other, CoerceNumber):
|
|
return self.arg, other.arg
|
|
else:
|
|
return (self.arg, other)
|
|
|
|
# New-style class version of CoerceNumber
|
|
class CoerceTo(object):
|
|
def __init__(self, arg):
|
|
self.arg = arg
|
|
def __coerce__(self, other):
|
|
if isinstance(other, CoerceTo):
|
|
return self.arg, other.arg
|
|
else:
|
|
return self.arg, other
|
|
|
|
|
|
# Fake a number that implements numeric ops through methods.
|
|
class MethodNumber:
|
|
def __init__(self,arg):
|
|
self.arg = arg
|
|
|
|
def __repr__(self):
|
|
return '<MethodNumber %s>' % repr(self.arg)
|
|
|
|
def __add__(self,other):
|
|
return self.arg + other
|
|
|
|
def __radd__(self,other):
|
|
return other + self.arg
|
|
|
|
def __sub__(self,other):
|
|
return self.arg - other
|
|
|
|
def __rsub__(self,other):
|
|
return other - self.arg
|
|
|
|
def __mul__(self,other):
|
|
return self.arg * other
|
|
|
|
def __rmul__(self,other):
|
|
return other * self.arg
|
|
|
|
def __div__(self,other):
|
|
return self.arg / other
|
|
|
|
def __rdiv__(self,other):
|
|
return other / self.arg
|
|
|
|
def __truediv__(self,other):
|
|
return self.arg / other
|
|
|
|
def __rtruediv__(self,other):
|
|
return other / self.arg
|
|
|
|
def __floordiv__(self,other):
|
|
return self.arg // other
|
|
|
|
def __rfloordiv__(self,other):
|
|
return other // self.arg
|
|
|
|
def __pow__(self,other):
|
|
return self.arg ** other
|
|
|
|
def __rpow__(self,other):
|
|
return other ** self.arg
|
|
|
|
def __mod__(self,other):
|
|
return self.arg % other
|
|
|
|
def __rmod__(self,other):
|
|
return other % self.arg
|
|
|
|
def __cmp__(self, other):
|
|
return cmp(self.arg, other)
|
|
|
|
|
|
candidates = [2, 2L, 4.0, 2+0j, [1], (2,), None,
|
|
MethodNumber(2), CoerceNumber(2)]
|
|
|
|
infix_binops = [ '+', '-', '*', '**', '%', '//', '/' ]
|
|
|
|
TE = TypeError
|
|
# b = both normal and augmented give same result list
|
|
# s = single result lists for normal and augmented
|
|
# e = equals other results
|
|
# result lists: ['+', '-', '*', '**', '%', '//', ('classic /', 'new /')]
|
|
# ^^^^^^^^^^^^^^^^^^^^^^
|
|
# 2-tuple if results differ
|
|
# else only one value
|
|
infix_results = {
|
|
# 2
|
|
(0,0): ('b', [4, 0, 4, 4, 0, 1, (1, 1.0)]),
|
|
(0,1): ('e', (0,0)),
|
|
(0,2): ('b', [6.0, -2.0, 8.0, 16.0, 2.0, 0.0, 0.5]),
|
|
(0,3): ('b', [4+0j, 0+0j, 4+0j, 4+0j, 0+0j, 1+0j, 1+0j]),
|
|
(0,4): ('b', [TE, TE, [1, 1], TE, TE, TE, TE]),
|
|
(0,5): ('b', [TE, TE, (2, 2), TE, TE, TE, TE]),
|
|
(0,6): ('b', [TE, TE, TE, TE, TE, TE, TE]),
|
|
(0,7): ('e', (0,0)),
|
|
(0,8): ('e', (0,0)),
|
|
|
|
# 2L
|
|
(1,0): ('e', (0,0)),
|
|
(1,1): ('e', (0,1)),
|
|
(1,2): ('e', (0,2)),
|
|
(1,3): ('e', (0,3)),
|
|
(1,4): ('e', (0,4)),
|
|
(1,5): ('e', (0,5)),
|
|
(1,6): ('e', (0,6)),
|
|
(1,7): ('e', (0,7)),
|
|
(1,8): ('e', (0,8)),
|
|
|
|
# 4.0
|
|
(2,0): ('b', [6.0, 2.0, 8.0, 16.0, 0.0, 2.0, 2.0]),
|
|
(2,1): ('e', (2,0)),
|
|
(2,2): ('b', [8.0, 0.0, 16.0, 256.0, 0.0, 1.0, 1.0]),
|
|
(2,3): ('b', [6+0j, 2+0j, 8+0j, 16+0j, 0+0j, 2+0j, 2+0j]),
|
|
(2,4): ('b', [TE, TE, TE, TE, TE, TE, TE]),
|
|
(2,5): ('e', (2,4)),
|
|
(2,6): ('e', (2,4)),
|
|
(2,7): ('e', (2,0)),
|
|
(2,8): ('e', (2,0)),
|
|
|
|
# (2+0j)
|
|
(3,0): ('b', [4+0j, 0+0j, 4+0j, 4+0j, 0+0j, 1+0j, 1+0j]),
|
|
(3,1): ('e', (3,0)),
|
|
(3,2): ('b', [6+0j, -2+0j, 8+0j, 16+0j, 2+0j, 0+0j, 0.5+0j]),
|
|
(3,3): ('b', [4+0j, 0+0j, 4+0j, 4+0j, 0+0j, 1+0j, 1+0j]),
|
|
(3,4): ('b', [TE, TE, TE, TE, TE, TE, TE]),
|
|
(3,5): ('e', (3,4)),
|
|
(3,6): ('e', (3,4)),
|
|
(3,7): ('e', (3,0)),
|
|
(3,8): ('e', (3,0)),
|
|
|
|
# [1]
|
|
(4,0): ('b', [TE, TE, [1, 1], TE, TE, TE, TE]),
|
|
(4,1): ('e', (4,0)),
|
|
(4,2): ('b', [TE, TE, TE, TE, TE, TE, TE]),
|
|
(4,3): ('b', [TE, TE, TE, TE, TE, TE, TE]),
|
|
(4,4): ('b', [[1, 1], TE, TE, TE, TE, TE, TE]),
|
|
(4,5): ('s', [TE, TE, TE, TE, TE, TE, TE], [[1, 2], TE, TE, TE, TE, TE, TE]),
|
|
(4,6): ('b', [TE, TE, TE, TE, TE, TE, TE]),
|
|
(4,7): ('e', (4,0)),
|
|
(4,8): ('e', (4,0)),
|
|
|
|
# (2,)
|
|
(5,0): ('b', [TE, TE, (2, 2), TE, TE, TE, TE]),
|
|
(5,1): ('e', (5,0)),
|
|
(5,2): ('b', [TE, TE, TE, TE, TE, TE, TE]),
|
|
(5,3): ('e', (5,2)),
|
|
(5,4): ('e', (5,2)),
|
|
(5,5): ('b', [(2, 2), TE, TE, TE, TE, TE, TE]),
|
|
(5,6): ('b', [TE, TE, TE, TE, TE, TE, TE]),
|
|
(5,7): ('e', (5,0)),
|
|
(5,8): ('e', (5,0)),
|
|
|
|
# None
|
|
(6,0): ('b', [TE, TE, TE, TE, TE, TE, TE]),
|
|
(6,1): ('e', (6,0)),
|
|
(6,2): ('e', (6,0)),
|
|
(6,3): ('e', (6,0)),
|
|
(6,4): ('e', (6,0)),
|
|
(6,5): ('e', (6,0)),
|
|
(6,6): ('e', (6,0)),
|
|
(6,7): ('e', (6,0)),
|
|
(6,8): ('e', (6,0)),
|
|
|
|
# MethodNumber(2)
|
|
(7,0): ('e', (0,0)),
|
|
(7,1): ('e', (0,1)),
|
|
(7,2): ('e', (0,2)),
|
|
(7,3): ('e', (0,3)),
|
|
(7,4): ('e', (0,4)),
|
|
(7,5): ('e', (0,5)),
|
|
(7,6): ('e', (0,6)),
|
|
(7,7): ('e', (0,7)),
|
|
(7,8): ('e', (0,8)),
|
|
|
|
# CoerceNumber(2)
|
|
(8,0): ('e', (0,0)),
|
|
(8,1): ('e', (0,1)),
|
|
(8,2): ('e', (0,2)),
|
|
(8,3): ('e', (0,3)),
|
|
(8,4): ('e', (0,4)),
|
|
(8,5): ('e', (0,5)),
|
|
(8,6): ('e', (0,6)),
|
|
(8,7): ('e', (0,7)),
|
|
(8,8): ('e', (0,8)),
|
|
}
|
|
|
|
def process_infix_results():
|
|
for key in sorted(infix_results):
|
|
val = infix_results[key]
|
|
if val[0] == 'e':
|
|
infix_results[key] = infix_results[val[1]]
|
|
else:
|
|
if val[0] == 's':
|
|
res = (val[1], val[2])
|
|
elif val[0] == 'b':
|
|
res = (val[1], val[1])
|
|
for i in range(1):
|
|
if isinstance(res[i][6], tuple):
|
|
if 1/2 == 0:
|
|
# testing with classic (floor) division
|
|
res[i][6] = res[i][6][0]
|
|
else:
|
|
# testing with -Qnew
|
|
res[i][6] = res[i][6][1]
|
|
infix_results[key] = res
|
|
|
|
|
|
with check_warnings(("classic (int|long) division", DeprecationWarning),
|
|
quiet=True):
|
|
process_infix_results()
|
|
# now infix_results has two lists of results for every pairing.
|
|
|
|
prefix_binops = [ 'divmod' ]
|
|
prefix_results = [
|
|
[(1,0), (1L,0L), (0.0,2.0), ((1+0j),0j), TE, TE, TE, TE, (1,0)],
|
|
[(1L,0L), (1L,0L), (0.0,2.0), ((1+0j),0j), TE, TE, TE, TE, (1L,0L)],
|
|
[(2.0,0.0), (2.0,0.0), (1.0,0.0), ((2+0j),0j), TE, TE, TE, TE, (2.0,0.0)],
|
|
[((1+0j),0j), ((1+0j),0j), (0j,(2+0j)), ((1+0j),0j), TE, TE, TE, TE, ((1+0j),0j)],
|
|
[TE, TE, TE, TE, TE, TE, TE, TE, TE],
|
|
[TE, TE, TE, TE, TE, TE, TE, TE, TE],
|
|
[TE, TE, TE, TE, TE, TE, TE, TE, TE],
|
|
[TE, TE, TE, TE, TE, TE, TE, TE, TE],
|
|
[(1,0), (1L,0L), (0.0,2.0), ((1+0j),0j), TE, TE, TE, TE, (1,0)]
|
|
]
|
|
|
|
def format_float(value):
|
|
if abs(value) < 0.01:
|
|
return '0.0'
|
|
else:
|
|
return '%.1f' % value
|
|
|
|
# avoid testing platform fp quirks
|
|
def format_result(value):
|
|
if isinstance(value, complex):
|
|
return '(%s + %sj)' % (format_float(value.real),
|
|
format_float(value.imag))
|
|
elif isinstance(value, float):
|
|
return format_float(value)
|
|
return str(value)
|
|
|
|
class CoercionTest(unittest.TestCase):
|
|
def test_infix_binops(self):
|
|
for ia, a in enumerate(candidates):
|
|
for ib, b in enumerate(candidates):
|
|
results = infix_results[(ia, ib)]
|
|
for op, res, ires in zip(infix_binops, results[0], results[1]):
|
|
if res is TE:
|
|
self.assertRaises(TypeError, eval,
|
|
'a %s b' % op, {'a': a, 'b': b})
|
|
else:
|
|
self.assertEquals(format_result(res),
|
|
format_result(eval('a %s b' % op)),
|
|
'%s %s %s == %s failed' % (a, op, b, res))
|
|
try:
|
|
z = copy.copy(a)
|
|
except copy.Error:
|
|
z = a # assume it has no inplace ops
|
|
if ires is TE:
|
|
try:
|
|
exec 'z %s= b' % op
|
|
except TypeError:
|
|
pass
|
|
else:
|
|
self.fail("TypeError not raised")
|
|
else:
|
|
exec('z %s= b' % op)
|
|
self.assertEquals(ires, z)
|
|
|
|
def test_prefix_binops(self):
|
|
for ia, a in enumerate(candidates):
|
|
for ib, b in enumerate(candidates):
|
|
for op in prefix_binops:
|
|
res = prefix_results[ia][ib]
|
|
if res is TE:
|
|
self.assertRaises(TypeError, eval,
|
|
'%s(a, b)' % op, {'a': a, 'b': b})
|
|
else:
|
|
self.assertEquals(format_result(res),
|
|
format_result(eval('%s(a, b)' % op)),
|
|
'%s(%s, %s) == %s failed' % (op, a, b, res))
|
|
|
|
def test_cmptypes(self):
|
|
# Built-in tp_compare slots expect their arguments to have the
|
|
# same type, but a user-defined __coerce__ doesn't have to obey.
|
|
# SF #980352
|
|
evil_coercer = CoerceTo(42)
|
|
# Make sure these don't crash any more
|
|
self.assertNotEquals(cmp(u'fish', evil_coercer), 0)
|
|
self.assertNotEquals(cmp(slice(1), evil_coercer), 0)
|
|
# ...but that this still works
|
|
class WackyComparer(object):
|
|
def __cmp__(slf, other):
|
|
self.assert_(other == 42, 'expected evil_coercer, got %r' % other)
|
|
return 0
|
|
__hash__ = None # Invalid cmp makes this unhashable
|
|
self.assertEquals(cmp(WackyComparer(), evil_coercer), 0)
|
|
# ...and classic classes too, since that code path is a little different
|
|
class ClassicWackyComparer:
|
|
def __cmp__(slf, other):
|
|
self.assert_(other == 42, 'expected evil_coercer, got %r' % other)
|
|
return 0
|
|
self.assertEquals(cmp(ClassicWackyComparer(), evil_coercer), 0)
|
|
|
|
def test_infinite_rec_classic_classes(self):
|
|
# if __coerce__() returns its arguments reversed it causes an infinite
|
|
# recursion for classic classes.
|
|
class Tester:
|
|
def __coerce__(self, other):
|
|
return other, self
|
|
|
|
exc = TestFailed("__coerce__() returning its arguments reverse "
|
|
"should raise RuntimeError")
|
|
try:
|
|
Tester() + 1
|
|
except (RuntimeError, TypeError):
|
|
return
|
|
except:
|
|
raise exc
|
|
else:
|
|
raise exc
|
|
|
|
def test_main():
|
|
with check_warnings(("complex divmod.., // and % are deprecated",
|
|
DeprecationWarning),
|
|
("classic (int|long) division", DeprecationWarning),
|
|
quiet=True):
|
|
run_unittest(CoercionTest)
|
|
|
|
if __name__ == "__main__":
|
|
test_main()
|