[boost] Re: String Algo bug.

From: Stefan Slapeta (stefan_nospam__at_[hidden])
Date: 2004-07-19 04:50:57

Stefan Slapeta wrote:
> David Abrahams wrote:
> 
>> If STLPort is complaining about that, it's because what you're doing
>> is illegal.  A default constructed standard container iterator is
>> "singular", and no operations on it (including comparison) are legal
>> other than assigning a non-singular iterator value into it.  Although
>> a few specialized iterators in std:: work differently
>> (e.g. istream_iterator), in general you have to make the same
>> assumption for any other iterator.
>

Very strange: the behaviour that the default constructor equals the 
past-the-end value is used for the example but it's not even documented 
for the find iterator!

See the example: http://tinyurl.com/4vzlm

     typedef find_iterator<string::iterator> string_find_iterator;
     for(string_find_iterator It=
             make_find_iterator(str1, first_finder("abc", is_iequal()));
         It!=string_find_iterator();
         ++It)
     ...

See the find iterator documentation: http://tinyurl.com/53j5x

 >>
     find_iterator();
     Default constructor.

     Construct null iterator. All null iterators are equal.

     Postconditions: eof()==true
<<

The example above implies that a find iterator which points to the last 
element will become a 'null iterator' after incrementation. (However, I 
didn't find this statement anywhere in the documentation)

Thus, fixing this bug would not only mean to provide a second function 
like 'make_find_iterator_end(collection)' that returns a proper end 
iterator but also to change the behaviour of iterators for 
incrementation to the past-the-end value etc.!

I would really like to hear what Pavol says about that...

Stefan