Question

stper** pages;
int tableSize;    
struct Person{

    string name; 
    int age;    
    string homeTown;
};


void fonk1 (int numberOfBuckets)
{
    pages = new stper*[numberOfBuckets]();
    tableSize = numberOfBuckets;
} 

   int hashPerson(Person& person)
   {
    int hashVal = 0;
    for (int i=0; i < (person.getName()).length() ; i++)
        hashVal = 37*hashVal + (person.getName())[i];

    for (int i=0; i < (person.getHomeTown()).length() ; i++)
        hashVal = 37*hashVal + (person.getHomeTown())[i];   
    hashVal+= person.getAge();  

    hashVal %= tableSize;
    if(hashVal < 0)
        hashVal += tableSize;
    return hashVal;
   }

Hello everyone, I am new at hashing. My hashing function is at above in hashPerson function and asyou can see there are three key. Is my function a good algorithm for hashing and how can I improve the function and decrease the number of collision? (Please ignore if there are any syntax mistakes)

Answer 1

I have a few suggestions:

1. Use unsigned instead of int. In my experience this has proven to perform better, as when the unsigned overflows, it still stays non-negative(otherwise %-ing may lead to big problems – you get a negative index and … a crash) and it also results in reduced collision rate(empirically proven). Also after all the function is supposed to return an index in a table so it is natural for the value to be unsigned – the index can not be negative.

2. Multiply hashVal by something when adding the age. I would suggest a value larger then any possible age for instance 200.

3. You never say what is tableSize but I would advice you to use some big(as big as possible) prime number, again to reduce collision rate.