Welcome, guest | Sign In | My Account | Store | Cart

This is a handy little decorator that lets you annotate function definitions with argument type requirements. These type requirements are automatically checked by the system at function invocation time. The decorator frees you from writing type-checking boilerplate code by hand.

Python, 42 lines
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
def require(arg_name, *allowed_types):
    def make_wrapper(f):
        if hasattr(f, "wrapped_args"):
            wrapped_args = getattr(f, "wrapped_args")
        else:
            code = f.func_code
            wrapped_args = list(code.co_varnames[:code.co_argcount])

        try:
            arg_index = wrapped_args.index(arg_name)
        except ValueError:
            raise NameError, arg_name

        def wrapper(*args, **kwargs):
            if len(args) > arg_index:
                arg = args[arg_index]
                if not isinstance(arg, allowed_types):
                    type_list = " or ".join(str(allowed_type) for allowed_type in allowed_types)
                    raise TypeError, "Expected '%s' to be %s; was %s." % (arg_name, type_list, type(arg))
            else:
                if arg_name in kwargs:
                    arg = kwargs[arg_name]
                    if not isinstance(arg, allowed_types):
                        type_list = " or ".join(str(allowed_type) for allowed_type in allowed_types)
                        raise TypeError, "Expected '%s' to be %s; was %s." % (arg_name, type_list, type(arg))

            return f(*args, **kwargs)

        wrapper.wrapped_args = wrapped_args
        return wrapper

    return make_wrapper

@require("x", int, float)
@require("y", float)
def foo(x, y):
    return x+y

print foo(1, 2.5)      # Prints 3.5.
print foo(2.0, 2.5)    # Prints 4.5.
print foo("asdf", 2.5) # Raises TypeError exception.
print foo(1, 2)        # Raises TypeError exception.

A small restriction on this decorator is that all uses of @require must come first in the stack of decorators applied to a function. In other words,

<pre> @some_decorator @require("x", int) @require("y", int) def f(x, y): pass </pre> is allowed, whereas

<pre> @require("x", int) @require("y", int) @some_decorator def f(x, y): pass </pre> is not allowed. The reason is that the require decorator uses the wrapped_args function attribute to propagate the argument list of the originally decorated function down the decorator stack, and other decorators do not participate in the propagation.

The checking provided by this decorator is fairly basic. I have plans to extend it with more sophisticated contract-checking functionality, but it is usable for type-checking even in its current state.

5 comments

Per Vognsen (author) 18 years, 5 months ago  # | flag

Type-checking considered harmful? Before anyone jumps on me: in posting this, I am not implicitly advocating that explicit type checks are usually a good idea in Python. I know the potentially detrimental effect it can have on polymorphism, and duck typing in particular. In fact, I almost never use these kinds of up-front checks. The recipe is mostly intended as a demonstration of an implementation technique that can be used for more general contract-checking. And if you need to write boilerplate type-checking code for whatever reason (I don't judge!), the require decorator is a useful shortcut.

Josiah Carlson 18 years, 5 months ago  # | flag

There are previously existing recipes which do this.

Brian Hawthorne 18 years ago  # | flag

more compact. nice recipe. might be nice to enhance it so that all the type specs could be specified in a single @require call, like this:

@require(x=(int,float), y=float)
def foo(x,y): pass

seems like it shouldn't be too hard, something along the lines of:

def require(**typemap):
    for arg_name, allowed_types in typemap:
        if type(allowed_types) == types.TypeType:
            allowed_types = (allowed_types,)
        # ... proceed as before
<pre>

</pre>

Jon Jay Obermark 14 years, 6 months ago  # | flag

Here is a simpler, though less flexible, candidate, using named parameters:

def types(**_params_):
    def check_types(_func_, _params_ = _params_):
        def modified(*args, **kw):
            arg_names = _func_.func_code.co_varnames
            kw.update(zip(arg_names, args))
            for name, type in _params_.iteritems():
                param = kw[name]
                assert param is None or isinstance(param, type), "Parameter '%s' should be type '%s'" % (name, type.__name__)
            return _func_(**kw)
        return modified
    return check_types

you say:

@types(foo = str, bar = bool)
def update(foo, bar = False):
    ....
Dan Moore 14 years, 5 months ago  # | flag

Even though type checking isn't supposed to be Pythonic, I think this could be made more so. For example, using your decorator, the following is not allowed:

@require("x", int)
def foo(x):
    return x

print foo("2")

What I think should happen here is that the string "2" is converted to an int. To be properly Pythonic, I don't think the instruction "require x to be an int" means require it to be an instance of int, but require it to be an object that can be used as an int. I've modified your decorator slightly so that in the above case, it will call int() on the specified argument and pass this converted object to the foo function. Obviously a string like "hello" can't be converted to an int and in that case, a normal ValueError will be thrown.

One area in which this is less flexible though is that you can't check that "x" is an int OR a float. In that case, there's no way of knowing if you want the object to be converted to an int or a float. Although since an int will be converted to a float, then you could just check for a float and still pass an int.

I've only tested this with things like ints and floats, but if you've got your own classes with __init__ functions then I see no reason why it shouldn't work for that as well although obviously it can only take one parameter,

def require(arg_name, allowed_type):
    """ Checks that arg_name is an instance of allowed type. If not, attempts to convert the respective argument to the allowed type """

    def make_wrapper(f):
        if hasattr(f, "wrapped_args"):
            wrapped_args = getattr(f, "wrapped_args")
        else:
            code = f.func_code
            wrapped_args = list(code.co_varnames[:code.co_argcount])

        try:
            arg_index = wrapped_args.index(arg_name)
        except ValueError:
            raise NameError, arg_name

        def wrapper(*args, **kwargs):
            args=list(args)

            if len(args) > arg_index:
                arg = args[arg_index]
                if not isinstance(arg, allowed_type):
                    args[arg_index]=allowed_type(args[arg_index])
            else:
                if arg_name in kwargs:
                    arg = kwargs[arg_name]
                    if not isinstance(arg, allowed_type):
                        kwargs[arg_name]=allowed_type(kwargs[arg_name])

            return f(*args, **kwargs)

        wrapper.wrapped_args = wrapped_args
        return wrapper

    return make_wrapper
Created by Per Vognsen on Tue, 15 Nov 2005 (PSF)
Python recipes (4591)
Per Vognsen's recipes (1)

Required Modules

  • (none specified)

Other Information and Tasks