>> I guess this would be helpful
No, not a bit. Your code is wrong.

>> for (n=0;inp[n];n++)
Assuming that n was defined somewhere, the test for imp[n] against 0 is dngerous because imp is uninitialized. There could be a null character straight away, or 5000 characters later. You're really risking an access violation with this loop.

>> char *arr = new char [sizeof(char)];
This alloctes memory for one char. sizeof(char) is guaranteed to return 1, everywhere, without fail. That's one of the few absolutes when it comes to type sizes in C++.

>> for (n=0;arr[n];n++)
You have the same problem here as the previous loop.

Yes, your idea is valid assuming there's some sentinel value at the end of the array to stop the loop on. With the original question that's difficult because any string object is valid in the general case. That's probably why you took it upon yourself to change the example to char so that you could use a null character as the sentinel.

The best solution is to avoid using arrays in the first place because they're unsafe and most people don't understand them well enough to avoid the pitfalls, as displayed by your flawed example. The std::vector class provides a good container that grows dynamically.