Question

I was asked this in an interview,
If you to find out if a string consists of only a given set of characters.
For example, let the set of strings be all strings over {0,1,2,3,4,5,6,7,8,9} ie all “numeric” strings. Among this, if the set of strings over {3,8,5} are only valid ones, how do I check if the string consists of only valid characters.
Say:

Input 8888338385
     Output VALID
Input 887837348234 
Output : Invalid

The way that I suggested was brute force, that required checking every character in the given string against a list of invalid characters. If any one of the character was invalid, I’d skip checking all other characters and display the failure message.
However, as suggested here, there may be better algorithms.
Please help.

Answer 1

EDIT: Thanks to Luc Touraille for a vast improvement to the original algorithm.

Create an array a[10] of booleans. For each expected digit e, set a[e] = true.

Now for each digit d in your input, check if a[d] is true. If it’s not, return false. If they all succeed, return true.

You can generalise this to all ASCII characters with a 256-element array.

If your input string is length N, your comparison string is length M, and the number of letters in your alphabet is A, then the complexity is O(N+M) (to scan your two strings) plus O(A) (to initialise the boolean array). So unless your string length is close to or greater than your alphabet size, this might not be optimal.

It’s worth pointing out, with respect to Niklas Baumstark’s excellent performance comparison, that our two solutions are actually the same. The boolean array constructed here is identical to the transition table you’d build in a two-state DFA accepting [c₁c₂…]*. I’d imagine the only difference is that Java’s implementation, being much more general, carries a lot more overhead.