Case-insensitive string comparison in C++

Additional pitfalls to watch out for when doing case insensitive compares:

Comparing as lower or as upper case? (common enough issue)

Both below will return 0 with strcicmpL("A", "a") and strcicmpU("A", "a").
Yet strcicmpL("A", "_") and strcicmpU("A", "_") can return different signed results as '_' is often between the upper and lower case letters.

This affects the sort order when used with qsort(..., ..., ..., strcicmp). Non-standard library C functions like the commonly available stricmp() or strcasecmp() tend to be well defined and favor comparing via lowercase. Yet variations exist.

int strcicmpL(char const *a, char const *b) {
  while (*b) {
    int d = tolower(*a) - tolower(*b);
    if (d) {
        return d;
    } 
    a++;
    b++;
  } 
  return tolower(*a);
}

int strcicmpU(char const *a, char const *b) {
  while (*b) {
    int d = toupper(*a) - toupper(*b);
    if (d) {
        return d;
    } 
    a++;
    b++;
  } 
  return toupper(*a);
}

char can have a negative value. (not rare)

touppper(int) and tolower(int) are specified for unsigned char values and the negative EOF. Further, strcmp() returns results as if each char was converted to unsigned char, regardless if char is signed or unsigned.

tolower(*a); // Potential UB
tolower((unsigned char) *a); // Correct (Almost - see following)

char can have a negative value and not 2's complement. (rare)

The above does not handle -0 nor other negative values properly as the bit pattern should be interpreted as unsigned char. To properly handle all integer encodings, change the pointer type first.

// tolower((unsigned char) *a);
tolower(*(const unsigned char *)a); // Correct

Locale (less common)

Although character sets using ASCII code (0-127) are ubiquitous, the remainder codes tend to have locale specific issues. So strcasecmp("\xE4", "a") might return a 0 on one system and non-zero on another.

Unicode (the way of the future)

If a solution needs to handle more than ASCII consider a unicode_strcicmp(). As C lib does not provide such a function, a pre-coded function from some alternate library is recommended. Writing your own unicode_strcicmp() is a daunting task.

Do all letters map one lower to one upper? (pedantic)

[A-Z] maps one-to-one with [a-z], yet various locales map various lower case chracters to one upper and visa-versa. Further, some uppercase characters may lack a lower case equivalent and again, visa-versa.

This obliges code to covert through both tolower() and tolower().

int d = tolower(toupper(*a)) - tolower(toupper(*b));

Again, potential different results for sorting if code did tolower(toupper(*a)) vs. toupper(tolower(*a)).

Portability

@B. Nadolson recommends to avoid rolling your own strcicmp() and this is reasonable, except when code needs high equivalent portable functionality.

Below is an approach that even performed faster than some system provided functions. It does a single compare per loop rather than two by using 2 different tables that differ with '\0'. Your results may vary.

static unsigned char low1[UCHAR_MAX + 1] = {
  0, 1, 2, 3, ...
  '@', 'a', 'b', 'c', ... 'z', `[`, ...  // @ABC... Z[...
  '`', 'a', 'b', 'c', ... 'z', `{`, ...  // `abc... z{...
}
static unsigned char low2[UCHAR_MAX + 1] = {
// v--- Not zero, but A which matches none in `low1[]`
  'A', 1, 2, 3, ...
  '@', 'a', 'b', 'c', ... 'z', `[`, ...
  '`', 'a', 'b', 'c', ... 'z', `{`, ...
}

int strcicmp_ch(char const *a, char const *b) {
  // compare using tables that differ slightly.
  while (low1[*(const unsigned char *)a] == low2[*(const unsigned char *)b]) {
    a++;
    b++;
  }
  // Either strings differ or null character detected.
  // Perform subtraction using same table.
  return (low1[*(const unsigned char *)a] - low1[*(const unsigned char *)b]);
}

I've found built-in such method named from which contains additional string functions to the standard header .

Here's the relevant signatures : 

int  strcasecmp(const char *, const char *);
int  strncasecmp(const char *, const char *, size_t);

I also found it's synonym in xnu kernel (osfmk/device/subrs.c) and it's implemented in the following code, so you wouldn't expect to have any change of behavior in number compared to the original strcmp function. 

tolower(unsigned char ch) {
    if (ch >= 'A' && ch <= 'Z')
        ch = 'a' + (ch - 'A');
    return ch;
 }

int strcasecmp(const char *s1, const char *s2) {
    const unsigned char *us1 = (const u_char *)s1,
                        *us2 = (const u_char *)s2;

    while (tolower(*us1) == tolower(*us2++))
        if (*us1++ == '\0')
            return (0);
    return (tolower(*us1) - tolower(*--us2));
}

There is no function that does this in the C standard. Unix systems that comply with POSIX are required to have strcasecmp in the header strings.h; Microsoft systems have stricmp. To be on the portable side, write your own:

int strcicmp(char const *a, char const *b)
{
    for (;; a++, b++) {
        int d = tolower((unsigned char)*a) - tolower((unsigned char)*b);
        if (d != 0 || !*a)
            return d;
    }
}

But note that none of these solutions will work with UTF-8 strings, only ASCII ones.


Take a look at strcasecmp() in strings.h.

Tags:

C++

String

Recent Posts