Thread: Dissecting numbers

Originally Posted by epidemic

you could ask the user to enter that number in a character array, and just do - 48.

a[0] = 1 , to get the number , do a[0] - 48
a[1] = 2 " "
a[2 ] = 3, " "

Don't use magic constants... If a[0] == '1' to get a number use a[0] - '0'

I would write your loops as:

Code:

for (Iterator p = c.begin(); p != c.end(); ++p) {}

and

Code:

for (ReverseIterator p = c.rbegin(); p != c.rend(); ++p) {}

In the second loop above, in the last test, p is set to c.begin() - 1 which is invalid.

That is true.

It turns out that

Code:

for (p=c.end()-1; p >= c.begin(); p--) {}

is invalid for the same reason.

I disagree. For random access iterators:

Code:

p = c.end() - 1;

must have the same net effect with respect to p as:

Code:

p = c.end();
p--;

So either they are both invalid, or they are both valid. In my opinion, they are both valid. The C++ standard states that the pre-condition for --r (and thus r--) for some iterator r is that "there exists s such that r == ++s". The post-condition is that "s is dereferenceable". But the pre-condition for ++r is "r is dereferenceable" with the post-condition that "r is dereferenceable or r is
past-the-end". As far as I can tell, the code snippets above satisfy the pre-conditions.

An uninitialized local variable can normally take on any possible value. Does this mean that a local pointer variable HAS to be initialized immediately to avoid having an invalid value? Or is it only if one explicitly assigns an invalid value to it that one has a problem?

I think that there is no problem since it is uninitialised. Of course, it is good practice to initialise a pointer to null in such a case.

What I was referring to was not the validity of

Code:

p = c.end() - 1;

although there actually is another issue there - if c has zero length, then c.end() == c.begin() so even the initial value is invalid. I was thinking of the fact that in the last iteration of the loop, p == c.begin(), and then at the end of the loop, p-- tries to set p to c.begin() - 1. The loop assumes that the next test will fail, but if the code is running on something that checks each pointer assignment and traps immediately if the value is invalid, the program may crash right there. Or there may be some kind of wraparound so the test doesn't fail. The point is, as soon as the invalid pointer is generated, all bets are off.

Edit: Please replace the word "pointer" with "iterator" in the above - I'm much more familiar with pointers than general iterators, although since pointers are a special case, the issues are the same.