Question

While trying to write a tiny, obfuscated type checker, an unacceptable code pattern was discovered. However, it inconsistently fails to work properly. This is the code that was initally written to test it with.

def statictypes(a):
    def b(a, b, c):
        if b in a and not isinstance(c, a[b]): raise TypeError('{} should be {}, not {}'.format(b, a[b], type(c)))
        return c
    return __import__('functools').wraps(a)(lambda *c: b(a.__annotations__, 'return', a(*(b(a.__annotations__, *d) for d in zip(a.__code__.co_varnames, c)))))

@statictypes
def isallinstance(iterable: object, class_or_type_or_tuple: (type, tuple)) -> bool:
    """isallinstance(iterable, class_or_type_or_tuple) -> bool

    Return whether all items in an iterable are instances of a class or of a
    subclass thereof. With a type as second argument, return whether that is
    all items' type. The form using a tuple, isallinstance(x, (A, B, ...)),
    is a shortcut for any(isallinstance(x, y) for y in (A, B, ...)).
    """
    return all(isinstance(item, class_or_type_or_tuple) for item in iterable)

The following shows a conversation with Python’s interpreter and highlights the error that comes up. A TypeError is generated, but not the one that was expected. While generators were fine, now they fail.

>>> isallinstance(range(1000000), int)
True
>>> isallinstance(range(1000000), (int, float))
True
>>> isallinstance(range(1000000), [int, float])
Traceback (most recent call last):
  File "<pyshell#26>", line 1, in <module>
    isallinstance(range(1000000), [int, float])
  File "C:\Users\schappell\Downloads\test.py", line 5, in <lambda>
    return __import__('functools').wraps(a)(lambda *c: b(a.__annotations__, 'return', a(*(b(a.__annotations__, *d) for d in zip(a.__code__.co_varnames, c)))))
TypeError: isallinstance() argument after * must be a sequence, not generator

The statictypes function can be rewritten, and the isallinstance function redefined and wrapped. The simplest solution is to rewrite the generatior in statictypes to be a list comprehension.

def statictypes(a):
    def b(a, b, c):
        if b in a and not isinstance(c, a[b]): raise TypeError('{} should be {}, not {}'.format(b, a[b], type(c)))
        return c
    return __import__('functools').wraps(a)(lambda *c: b(a.__annotations__, 'return', a(*[b(a.__annotations__, *d) for d in zip(a.__code__.co_varnames, c)])))

After that, the isallinstance will starting working as expected once it is recreated from scratch. The TypeError stating what was wrong with the second argument is properly generated as was desired.

>>> isallinstance(range(1000000), int)
True
>>> isallinstance(range(1000000), (int, float))
True
>>> isallinstance(range(1000000), [int, float])
Traceback (most recent call last):
  File "<pyshell#29>", line 1, in <module>
    isallinstance(range(1000000), [int, float])
  File "C:\Users\schappell\Downloads\test.py", line 5, in <lambda>
    return __import__('functools').wraps(a)(lambda *c: b(a.__annotations__, 'return', a(*[b(a.__annotations__, *d) for d in zip(a.__code__.co_varnames, c)])))
  File "C:\Users\schappell\Downloads\test.py", line 5, in <listcomp>
    return __import__('functools').wraps(a)(lambda *c: b(a.__annotations__, 'return', a(*[b(a.__annotations__, *d) for d in zip(a.__code__.co_varnames, c)])))
  File "C:\Users\schappell\Downloads\test.py", line 3, in b
    if b in a and not isinstance(c, a[b]): raise TypeError('{} should be {}, not {}'.format(b, a[b], type(c)))
TypeError: class_or_type_or_tuple should be (<class 'type'>, <class 'tuple'>), not <class 'list'>

Questions:

1. Why does the first function with the generator somtimes work and other times fail?
2. Why is a generator not considered a sequence (since it generates a sequence)?
3. Why is a sequence needed when a generator obviously works some of the time?

Answer 1

1. Because isinstance, like a couple of other screwy standard library functions, does a different thing when you give it a tuple than other sequences. Namely, it works, and checks that the type is any of the ones given.
2. Because it isn’t. See the sequence protocol definition. It’d need to implement __getitem__ to be one.
3. A bug, which still hasn’t been merged, which is telling you that your generator is broken, but with the incorrect error message.

Also, please don’t dirty our lovely language with type checking like this for anything but good reasons :).