In C++ case
statements require a constant integer value and cannot be used with values calculated at runtime. However if you are using C++11 you can use a constexpr
function to generate case
values simulate using strings with a case
statement.
This uses a hash function that takes a pointer to a string and generates a value at compile time instead of runtime. If more than one string generates the same value (a hash collision) you get the familiar error message about multiple case statements using the same value.
constexpr unsigned int djb2Hash(const char* str, int index = 0)
{
return !str[index] ? 0x1505 : (djb2Hash(str, index + 1) * 0x21) ^ str[index];
}
The djb2Hash
function can then be used directly in both the switch
and case
statements. There is one caveat however, the hash function can result in a collision at runtime. The probability of this happening is driven primarily by the quality of the hash function. The solution presented here does not attempt to address this problem (but may in the future).
void DoSomething(const char *str)
{
switch(djb2Hash(str))
{
case djb2Hash("Hello"): SayHello(); break;
case djb2Hash("World"): SayWorld(); break;
}
}
This works very well but might be considered ugly. You can simplify this further by declaring a user defined literal that handles invoking the hash function.
// Create a literal type for short-hand case strings
constexpr unsigned int operator"" _C ( const char str[], size_t size)
{
return djb2Hash(str);
}
void DoSomething(const char *str)
{
switch(djb2Hash(str))
{
case "Hello"_C: SayHello(); break;
case "World"_C: SayWorld(); break;
}
}
This provides a more intuitive usage of strings in a switch statements but may also be considered slightly confusing because of the user defined literal.
[Edit: Added note about runtime hash collisions. Much Kudos to R. Martinho Fernandes for bringing it to my attention!]