Question

A common interview question asks to write an algorithm that detects duplicates in a string.
Using a character array of length 128 to keep track of the characters already seen is a good way to solve this problem in linear time.

In C we would type something like

char seen_chars[128];
unsigned char c;

/* set seen_chars to all zeros, assign c */

seen_chars[ c ] = 1;

To mark character c as seen. Of course this relies on

(int) c 

returning a value between 0 and 127.

I’m wondering when would this fail? What are the assumptions that make this code work correctly?

Answer 1

The code will fail (and cause undefined behavior) every time when the integer value of the given char c is not between 0 and 127 (inclusive).

C does in no way limit the maximum range of char – you are only guaranteed that it can hold at least 256 distinct values – so in any given C implementation a valid char value can be out of that boundary. On most desktop systems a char can hold values from -128 to 127, or from 0 to 255. However, as an example:

char aFunction(void);
char c = aFunction();
if ((int)c > 1000000000)
    printf("This could be true on some systems\n").

The following would be valid (although it may exhaust your stack on systems with large chars):

#include <limits.h>
_Bool seen[1<<CHAR_BIT] = {0};
seen[(unsigned char)c] = 1;
/* etc. */