[Tutor] Using type

[Dave Angel]

On 08/12/2011 05:31 AM, ALAN GAULD wrote:
>
>
>> It'll work on lists and tuples, for simple examples.  But beware that
>> when you iterate through strings, the individual characters are also
>> strings, and this function will fail with an error like:
>>     RuntimeError: maximum recursion depth exceeded
>
> Oh, good catch Dave.
>
> So you'd want to add an explicit check for a string.
> And in  fact you probably don't want to "flatten" a string into
> individual characters anyway!
>
> Alan G.
>
I agree with you about probably not wanting to flatten a string.  So I'd 
add an isinstance() for strings and unicode.  But the following is an 
interesting problem.

I like duck typing in general.  But I dislike having negative 
isinstance() checks just for such a case.  For example, one might 
remember to check for a string, but not unicode or byte, which have the 
same problem.  Not to mention a user written iterable that might also 
behave the same way.

For many functions, I'd simply document which types it's known to work 
for, and let the caller beware.  But infinite recursion is a 
particularly nasty failure mode.

Any suggestions how to solve a problem of this kind without having an 
explicit list of exceptions?

The usual rule for avoiding infinite recursion is to assure that each 
recursion "reduces" the problem in some way, so that the termination 
condition can be assured.  The following is the best I could come up to 
implement that.  It refuses to recurse if the iterable won't change.


def flatten(my_data):
     gut = []
     for item in my_data:
        try:
           iter(item)  # test for iterability
           if len(item) == 1 and item == item[0]:
               gut.append(item)
           else:
               gut = gut + flatten(item)
        except TypeError:
           gut.append(item)
     return gut



-- 

DaveA