cpython/Lib/test/test_compile.py

import unittest
import sys
import _ast
import types
from test import support

class TestSpecifics(unittest.TestCase):

    def compile_single(self, source):
        compile(source, "<single>", "single")

    def assertInvalidSingle(self, source):
        self.assertRaises(SyntaxError, self.compile_single, source)

    def test_no_ending_newline(self):
        compile("hi", "<test>", "exec")
        compile("hi\r", "<test>", "exec")

    def test_empty(self):
        compile("", "<test>", "exec")

    def test_other_newlines(self):
        compile("\r\n", "<test>", "exec")
        compile("\r", "<test>", "exec")
        compile("hi\r\nstuff\r\ndef f():\n    pass\r", "<test>", "exec")
        compile("this_is\rreally_old_mac\rdef f():\n    pass", "<test>", "exec")

    def test_debug_assignment(self):
        # catch assignments to __debug__
        self.assertRaises(SyntaxError, compile, '__debug__ = 1', '?', 'single')
        import builtins
        prev = builtins.__debug__
        setattr(builtins, '__debug__', 'sure')
        setattr(builtins, '__debug__', prev)

    def test_argument_handling(self):
        # detect duplicate positional and keyword arguments
        self.assertRaises(SyntaxError, eval, 'lambda a,a:0')
        self.assertRaises(SyntaxError, eval, 'lambda a,a=1:0')
        self.assertRaises(SyntaxError, eval, 'lambda a=1,a=1:0')
        self.assertRaises(SyntaxError, exec, 'def f(a, a): pass')
        self.assertRaises(SyntaxError, exec, 'def f(a = 0, a = 1): pass')
        self.assertRaises(SyntaxError, exec, 'def f(a): global a; a = 1')

    def test_syntax_error(self):
        self.assertRaises(SyntaxError, compile, "1+*3", "filename", "exec")

    def test_none_keyword_arg(self):
        self.assertRaises(SyntaxError, compile, "f(None=1)", "<string>", "exec")

    def test_duplicate_global_local(self):
        self.assertRaises(SyntaxError, exec, 'def f(a): global a; a = 1')

    def test_exec_with_general_mapping_for_locals(self):

        class M:
            "Test mapping interface versus possible calls from eval()."
            def __getitem__(self, key):
                if key == 'a':
                    return 12
                raise KeyError
            def __setitem__(self, key, value):
                self.results = (key, value)
            def keys(self):
                return list('xyz')

        m = M()
        g = globals()
        exec('z = a', g, m)
        self.assertEqual(m.results, ('z', 12))
        try:
            exec('z = b', g, m)
        except NameError:
            pass
        else:
            self.fail('Did not detect a KeyError')
        exec('z = dir()', g, m)
        self.assertEqual(m.results, ('z', list('xyz')))
        exec('z = globals()', g, m)
        self.assertEqual(m.results, ('z', g))
        exec('z = locals()', g, m)
        self.assertEqual(m.results, ('z', m))
        self.assertRaises(TypeError, exec, 'z = b', m)

        class A:
            "Non-mapping"
            pass
        m = A()
        self.assertRaises(TypeError, exec, 'z = a', g, m)

        # Verify that dict subclasses work as well
        class D(dict):
            def __getitem__(self, key):
                if key == 'a':
                    return 12
                return dict.__getitem__(self, key)
        d = D()
        exec('z = a', g, d)
        self.assertEqual(d['z'], 12)

    def test_extended_arg(self):
        longexpr = 'x = x or ' + '-x' * 2500
        g = {}
        code = '''
def f(x):
    %s
    %s
    %s
    %s
    %s
    %s
    %s
    %s
    %s
    %s
    # the expressions above have no effect, x == argument
    while x:
        x -= 1
        # EXTENDED_ARG/JUMP_ABSOLUTE here
    return x
''' % ((longexpr,)*10)
        exec(code, g)
        self.assertEqual(g['f'](5), 0)

    def test_argument_order(self):
        self.assertRaises(SyntaxError, exec, 'def f(a=1, b): pass')

    def test_float_literals(self):
        # testing bad float literals
        self.assertRaises(SyntaxError, eval, "2e")
        self.assertRaises(SyntaxError, eval, "2.0e+")
        self.assertRaises(SyntaxError, eval, "1e-")
        self.assertRaises(SyntaxError, eval, "3-4e/21")

    def test_indentation(self):
        # testing compile() of indented block w/o trailing newline"
        s = """
if 1:
    if 2:
        pass"""
        compile(s, "<string>", "exec")

    # This test is probably specific to CPython and may not generalize
    # to other implementations.  We are trying to ensure that when
    # the first line of code starts after 256, correct line numbers
    # in tracebacks are still produced.
    def test_leading_newlines(self):
        s256 = "".join(["\n"] * 256 + ["spam"])
        co = compile(s256, 'fn', 'exec')
        self.assertEqual(co.co_firstlineno, 257)
        self.assertEqual(co.co_lnotab, bytes())

    def test_literals_with_leading_zeroes(self):
        for arg in ["077787", "0xj", "0x.", "0e",  "090000000000000",
                    "080000000000000", "000000000000009", "000000000000008",
                    "0b42", "0BADCAFE", "0o123456789", "0b1.1", "0o4.2",
                    "0b101j2", "0o153j2", "0b100e1", "0o777e1", "0777",
                    "000777", "000000000000007"]:
            self.assertRaises(SyntaxError, eval, arg)

        self.assertEqual(eval("0xff"), 255)
        self.assertEqual(eval("0777."), 777)
        self.assertEqual(eval("0777.0"), 777)
        self.assertEqual(eval("000000000000000000000000000000000000000000000000000777e0"), 777)
        self.assertEqual(eval("0777e1"), 7770)
        self.assertEqual(eval("0e0"), 0)
        self.assertEqual(eval("0000e-012"), 0)
        self.assertEqual(eval("09.5"), 9.5)
        self.assertEqual(eval("0777j"), 777j)
        self.assertEqual(eval("000"), 0)
        self.assertEqual(eval("00j"), 0j)
        self.assertEqual(eval("00.0"), 0)
        self.assertEqual(eval("0e3"), 0)
        self.assertEqual(eval("090000000000000."), 90000000000000.)
        self.assertEqual(eval("090000000000000.0000000000000000000000"), 90000000000000.)
        self.assertEqual(eval("090000000000000e0"), 90000000000000.)
        self.assertEqual(eval("090000000000000e-0"), 90000000000000.)
        self.assertEqual(eval("090000000000000j"), 90000000000000j)
        self.assertEqual(eval("000000000000008."), 8.)
        self.assertEqual(eval("000000000000009."), 9.)
        self.assertEqual(eval("0b101010"), 42)
        self.assertEqual(eval("-0b000000000010"), -2)
        self.assertEqual(eval("0o777"), 511)
        self.assertEqual(eval("-0o0000010"), -8)

    def test_unary_minus(self):
        # Verify treatment of unary minus on negative numbers SF bug #660455
        if sys.maxsize == 2147483647:
            # 32-bit machine
            all_one_bits = '0xffffffff'
            self.assertEqual(eval(all_one_bits), 4294967295)
            self.assertEqual(eval("-" + all_one_bits), -4294967295)
        elif sys.maxsize == 9223372036854775807:
            # 64-bit machine
            all_one_bits = '0xffffffffffffffff'
            self.assertEqual(eval(all_one_bits), 18446744073709551615)
            self.assertEqual(eval("-" + all_one_bits), -18446744073709551615)
        else:
            self.fail("How many bits *does* this machine have???")
        # Verify treatment of constant folding on -(sys.maxsize+1)
        # i.e. -2147483648 on 32 bit platforms.  Should return int, not long.
        self.assertIsInstance(eval("%s" % (-sys.maxsize - 1)), int)
        self.assertIsInstance(eval("%s" % (-sys.maxsize - 2)), int)

    if sys.maxsize == 9223372036854775807:
        def test_32_63_bit_values(self):
            a = +4294967296  # 1 << 32
            b = -4294967296  # 1 << 32
            c = +281474976710656  # 1 << 48
            d = -281474976710656  # 1 << 48
            e = +4611686018427387904  # 1 << 62
            f = -4611686018427387904  # 1 << 62
            g = +9223372036854775807  # 1 << 63 - 1
            h = -9223372036854775807  # 1 << 63 - 1

            for variable in self.test_32_63_bit_values.__code__.co_consts:
                if variable is not None:
                    self.assertIsInstance(variable, int)

    def test_sequence_unpacking_error(self):
        # Verify sequence packing/unpacking with "or".  SF bug #757818
        i,j = (1, -1) or (-1, 1)
        self.assertEqual(i, 1)
        self.assertEqual(j, -1)

    def test_none_assignment(self):
        stmts = [
            'None = 0',
            'None += 0',
            '__builtins__.None = 0',
            'def None(): pass',
            'class None: pass',
            '(a, None) = 0, 0',
            'for None in range(10): pass',
            'def f(None): pass',
            'import None',
            'import x as None',
            'from x import None',
            'from x import y as None'
        ]
        for stmt in stmts:
            stmt += "\n"
            self.assertRaises(SyntaxError, compile, stmt, 'tmp', 'single')
            self.assertRaises(SyntaxError, compile, stmt, 'tmp', 'exec')

    def test_import(self):
        succeed = [
            'import sys',
            'import os, sys',
            'import os as bar',
            'import os.path as bar',
            'from __future__ import nested_scopes, generators',
            'from __future__ import (nested_scopes,\ngenerators)',
            'from __future__ import (nested_scopes,\ngenerators,)',
            'from sys import stdin, stderr, stdout',
            'from sys import (stdin, stderr,\nstdout)',
            'from sys import (stdin, stderr,\nstdout,)',
            'from sys import (stdin\n, stderr, stdout)',
            'from sys import (stdin\n, stderr, stdout,)',
            'from sys import stdin as si, stdout as so, stderr as se',
            'from sys import (stdin as si, stdout as so, stderr as se)',
            'from sys import (stdin as si, stdout as so, stderr as se,)',
            ]
        fail = [
            'import (os, sys)',
            'import (os), (sys)',
            'import ((os), (sys))',
            'import (sys',
            'import sys)',
            'import (os,)',
            'import os As bar',
            'import os.path a bar',
            'from sys import stdin As stdout',
            'from sys import stdin a stdout',
            'from (sys) import stdin',
            'from __future__ import (nested_scopes',
            'from __future__ import nested_scopes)',
            'from __future__ import nested_scopes,\ngenerators',
            'from sys import (stdin',
            'from sys import stdin)',
            'from sys import stdin, stdout,\nstderr',
            'from sys import stdin si',
            'from sys import stdin,'
            'from sys import (*)',
            'from sys import (stdin,, stdout, stderr)',
            'from sys import (stdin, stdout),',
            ]
        for stmt in succeed:
            compile(stmt, 'tmp', 'exec')
        for stmt in fail:
            self.assertRaises(SyntaxError, compile, stmt, 'tmp', 'exec')

    def test_for_distinct_code_objects(self):
        # SF bug 1048870
        def f():
            f1 = lambda x=1: x
            f2 = lambda x=2: x
            return f1, f2
        f1, f2 = f()
        self.assertNotEqual(id(f1.__code__), id(f2.__code__))

    def test_lambda_doc(self):
        l = lambda: "foo"
        self.assertIsNone(l.__doc__)

##     def test_unicode_encoding(self):
##         code = "# -*- coding: utf-8 -*-\npass\n"
##         self.assertRaises(SyntaxError, compile, code, "tmp", "exec")

    def test_subscripts(self):
        # SF bug 1448804
        # Class to make testing subscript results easy
        class str_map(object):
            def __init__(self):
                self.data = {}
            def __getitem__(self, key):
                return self.data[str(key)]
            def __setitem__(self, key, value):
                self.data[str(key)] = value
            def __delitem__(self, key):
                del self.data[str(key)]
            def __contains__(self, key):
                return str(key) in self.data
        d = str_map()
        # Index
        d[1] = 1
        self.assertEqual(d[1], 1)
        d[1] += 1
        self.assertEqual(d[1], 2)
        del d[1]
        self.assertNotIn(1, d)
        # Tuple of indices
        d[1, 1] = 1
        self.assertEqual(d[1, 1], 1)
        d[1, 1] += 1
        self.assertEqual(d[1, 1], 2)
        del d[1, 1]
        self.assertNotIn((1, 1), d)
        # Simple slice
        d[1:2] = 1
        self.assertEqual(d[1:2], 1)
        d[1:2] += 1
        self.assertEqual(d[1:2], 2)
        del d[1:2]
        self.assertNotIn(slice(1, 2), d)
        # Tuple of simple slices
        d[1:2, 1:2] = 1
        self.assertEqual(d[1:2, 1:2], 1)
        d[1:2, 1:2] += 1
        self.assertEqual(d[1:2, 1:2], 2)
        del d[1:2, 1:2]
        self.assertNotIn((slice(1, 2), slice(1, 2)), d)
        # Extended slice
        d[1:2:3] = 1
        self.assertEqual(d[1:2:3], 1)
        d[1:2:3] += 1
        self.assertEqual(d[1:2:3], 2)
        del d[1:2:3]
        self.assertNotIn(slice(1, 2, 3), d)
        # Tuple of extended slices
        d[1:2:3, 1:2:3] = 1
        self.assertEqual(d[1:2:3, 1:2:3], 1)
        d[1:2:3, 1:2:3] += 1
        self.assertEqual(d[1:2:3, 1:2:3], 2)
        del d[1:2:3, 1:2:3]
        self.assertNotIn((slice(1, 2, 3), slice(1, 2, 3)), d)
        # Ellipsis
        d[...] = 1
        self.assertEqual(d[...], 1)
        d[...] += 1
        self.assertEqual(d[...], 2)
        del d[...]
        self.assertNotIn(Ellipsis, d)
        # Tuple of Ellipses
        d[..., ...] = 1
        self.assertEqual(d[..., ...], 1)
        d[..., ...] += 1
        self.assertEqual(d[..., ...], 2)
        del d[..., ...]
        self.assertNotIn((Ellipsis, Ellipsis), d)

    def test_annotation_limit(self):
        # 16 bits are available for # of annotations, but only 8 bits are
        # available for the parameter count, hence 255
        # is the max. Ensure the result of too many annotations is a
        # SyntaxError.
        s = "def f(%s): pass"
        s %= ', '.join('a%d:%d' % (i,i) for i in range(256))
        self.assertRaises(SyntaxError, compile, s, '?', 'exec')
        # Test that the max # of annotations compiles.
        s = "def f(%s): pass"
        s %= ', '.join('a%d:%d' % (i,i) for i in range(255))
        compile(s, '?', 'exec')

    def test_mangling(self):
        class A:
            def f():
                __mangled = 1
                __not_mangled__ = 2
                import __mangled_mod
                import __package__.module

        self.assertIn("_A__mangled", A.f.__code__.co_varnames)
        self.assertIn("__not_mangled__", A.f.__code__.co_varnames)
        self.assertIn("_A__mangled_mod", A.f.__code__.co_varnames)
        self.assertIn("__package__", A.f.__code__.co_varnames)

    def test_compile_ast(self):
        fname = __file__
        if fname.lower().endswith(('pyc', 'pyo')):
            fname = fname[:-1]
        with open(fname, 'r') as f:
            fcontents = f.read()
        sample_code = [
            ['<assign>', 'x = 5'],
            ['<ifblock>', """if True:\n    pass\n"""],
            ['<forblock>', """for n in [1, 2, 3]:\n    print(n)\n"""],
            ['<deffunc>', """def foo():\n    pass\nfoo()\n"""],
            [fname, fcontents],
        ]

        for fname, code in sample_code:
            co1 = compile(code, '%s1' % fname, 'exec')
            ast = compile(code, '%s2' % fname, 'exec', _ast.PyCF_ONLY_AST)
            self.assertTrue(type(ast) == _ast.Module)
            co2 = compile(ast, '%s3' % fname, 'exec')
            self.assertEqual(co1, co2)
            # the code object's filename comes from the second compilation step
            self.assertEqual(co2.co_filename, '%s3' % fname)

        # raise exception when node type doesn't match with compile mode
        co1 = compile('print(1)', '<string>', 'exec', _ast.PyCF_ONLY_AST)
        self.assertRaises(TypeError, compile, co1, '<ast>', 'eval')

        # raise exception when node type is no start node
        self.assertRaises(TypeError, compile, _ast.If(), '<ast>', 'exec')

        # raise exception when node has invalid children
        ast = _ast.Module()
        ast.body = [_ast.BoolOp()]
        self.assertRaises(TypeError, compile, ast, '<ast>', 'exec')

    @support.cpython_only
    def test_same_filename_used(self):
        s = """def f(): pass\ndef g(): pass"""
        c = compile(s, "myfile", "exec")
        for obj in c.co_consts:
            if isinstance(obj, types.CodeType):
                self.assertIs(obj.co_filename, c.co_filename)

    def test_single_statement(self):
        self.compile_single("1 + 2")
        self.compile_single("\n1 + 2")
        self.compile_single("1 + 2\n")
        self.compile_single("1 + 2\n\n")
        self.compile_single("1 + 2\t\t\n")
        self.compile_single("1 + 2\t\t\n        ")
        self.compile_single("1 + 2 # one plus two")
        self.compile_single("1; 2")
        self.compile_single("import sys; sys")
        self.compile_single("def f():\n   pass")
        self.compile_single("while False:\n   pass")
        self.compile_single("if x:\n   f(x)")
        self.compile_single("if x:\n   f(x)\nelse:\n   g(x)")
        self.compile_single("class T:\n   pass")

    def test_bad_single_statement(self):
        self.assertInvalidSingle('1\n2')
        self.assertInvalidSingle('def f(): pass')
        self.assertInvalidSingle('a = 13\nb = 187')
        self.assertInvalidSingle('del x\ndel y')
        self.assertInvalidSingle('f()\ng()')
        self.assertInvalidSingle('f()\n# blah\nblah()')
        self.assertInvalidSingle('f()\nxy # blah\nblah()')
        self.assertInvalidSingle('x = 5 # comment\nx = 6\n')

    @support.cpython_only
    def test_compiler_recursion_limit(self):
        # Expected limit is sys.getrecursionlimit() * the scaling factor
        # in symtable.c (currently 3)
        # We expect to fail *at* that limit, because we use up some of
        # the stack depth limit in the test suite code
        # So we check the expected limit and 75% of that
        # XXX (ncoghlan): duplicating the scaling factor here is a little
        # ugly. Perhaps it should be exposed somewhere...
        fail_depth = sys.getrecursionlimit() * 3
        success_depth = int(fail_depth * 0.75)

        def check_limit(prefix, repeated):
            expect_ok = prefix + repeated * success_depth
            self.compile_single(expect_ok)
            broken = prefix + repeated * fail_depth
            details = "Compiling ({!r} + {!r} * {})".format(
                         prefix, repeated, fail_depth)
            with self.assertRaises(RuntimeError, msg=details):
                self.compile_single(broken)

        check_limit("a", "()")
        check_limit("a", ".b")
        check_limit("a", "[0]")
        check_limit("a", "*a")


def test_main():
    support.run_unittest(TestSpecifics)

if __name__ == "__main__":
    test_main()