[Python-Dev] Validators (was Re: Python-Dev Digest, Vol 9, Issue 17) (original) (raw)

Isaac ishnigarrab at earthlink.net
Mon Apr 5 21:30:32 EDT 2004

Guido:

Please note that you seem to be using the syntax ``type: variable''. We've settled on using ``variable: type'' instead. (How to express types is quite a different story...)

I haven't been following the thread on the subject much at all, I've just heard some rumbling about introducing static typing into python, and frankly I'm a bit wary about the whole thing. I was just thinking that in keeping with python's dynamic nature that something more like a generalized way of validating the nature of something as opposed to it's underlying type would be more appropriate. It's just that it rings more pythonic to me (not that I have in any way graduated to the rank of Pythonist to make such judgments). As for the order of things, it's really not that important, "name: validator" would be just as useful. The only problem would be that

def func(name: int): pass

works in terms of type checking, as in

def func(name):
    if not isinstance(name, int):
        raise TypeError

where as translating it as a validator:

def func(name):
    int(name)

doesn't make as much sense, int would have to be overloaded or something like is_int would have to be implemented.

But the advantages of validators would be numerous, IMO:

i.e.:

def add(a: int, b: int): return a+b

add('45', 5) => 50

if it were established that validators raise exceptions if the input is invalid, and return the value (possibly modified) if valid. Similar to implicit upcasting, i.e.:

class MyInt(int):
   def __init__(self, i):
      self.i = i
   def add(self, other):
      self.i += other.i
      return self

def add(a: MyInt, b: MyInt):
   return a.add(b)

add(4, 4) => MyInt(8)

Again, I'm mostly just thinking aloud, hoping to be productive.

Isaac

More information about the Python-Dev mailing list