I just wanted a simple and straightforward hash table implementation that I could drop into my own C-based projects on whatever platform.
I haven't implemented one of these before, so it may be super naive, but it does appear to work pretty well.
NOTE: After exposure on HN, and seeing other hash implementations, I'm planning to restructure the code to a macro-based style, which should cut down the duplication.
You can create a hash table, and add strings, long ints, doubles and pointers to it, keyed by strings or long ints.
You can retrieve strings, long ints, doubles and pointers via the get functions.
All strings saved in a hash table are copied, and copies of strings are returned on retrieval.
I added locking on hash buckets which only minorly affects performance, and allows safe retrieval and storing of key value pairs.
Performance seems decent. Saving 1,000,000 int values by string key is done in about 0.28 secs, and multi-threaded performance scales pretty closely with number of processors.
Type make in the folder to build the code. Type ./test to run the demo.
The following were key to getting various aspects working:
Hash function for integer keys. http://stackoverflow.com/a/12996028
Hash function for string keys. http://www.cse.yorku.ca/~oz/hash.html
Efficient lock when low-contention is expected. http://stackoverflow.com/questions/1383363/is-my-spin-lock-implementation-correct-and-optimal
typedef struct jwHashTable jwHashTable;
struct jwHashTable
{
jwHashEntry **bucket; // pointer to array of buckets
size_t buckets;
size_t bucketsinitial; // if we resize, may need to hash multiple times
HASHRESULT lastError;
#ifdef HASHTHREADED
volatile int *locks; // array of locks
volatile int lock; // lock for entire table
#endif
};
typedef struct jwHashEntry jwHashEntry;
struct jwHashEntry
{
union
{
char *strValue;
double dblValue;
int intValue;
} key;
HASHVALTAG valtag;
union
{
char *strValue;
double dblValue;
int intValue;
void *ptrValue;
} value;
jwHashEntry *next;
};
jwHashTable *create_hash( size_t buckets );
void *delete_hash( jwHashTable *table ); // clean up all memory
HASHRESULT add_str_by_str( jwHashTable*, char *key, char *value );
HASHRESULT add_int_by_str( jwHashTable*, char *key, long int value );
HASHRESULT add_dbl_by_str( jwHashTable*, char *key, double value );
HASHRESULT add_ptr_by_str( jwHashTable*, char *key, void *value );
HASHRESULT del_by_str( jwHashTable*, char *key );
HASHRESULT get_str_by_str( jwHashTable *table, char *key, char **value );
HASHRESULT get_int_by_str( jwHashTable *table, char *key, int *i );
HASHRESULT get_dbl_by_str( jwHashTable *table, char *key, double *val );
HASHRESULT get_ptr_by_str( jwHashTable *table, char *key, void **val );
[Similar for long int keys]
- Support multi-threading, -- this started, and implemented for the test
- Implement clean-up,
- Implement re-hashing to a larger hash table,
- Implement a callback to allow iterating through keys, values
// Test hashing by string
char * strv1 = "Jonathan";
char * strv2 = "Zevi";
char * strv3 = "Jude";
char * strv4 = "Voldemort";
add_str_by_str(table,"oldest",strv1);
add_str_by_str(table,"2ndoldest",strv2);
add_str_by_str(table,"3rdoldest",strv3);
add_str_by_str(table,"4tholdest",strv4);
char * sstrv1; get_str_by_str(table,"oldest",&sstrv1);
char * sstrv2; get_str_by_str(table,"2ndoldest",&sstrv2);
char * sstrv3; get_str_by_str(table,"3rdoldest",&sstrv3);
char * sstrv4; get_str_by_str(table,"4tholdest",&sstrv4);
printf("got strings:\noldest->%s \n2ndoldest->%s \n3rdoldest->%s \n4tholdest->%s\n",
sstrv1,sstrv2,sstrv3,sstrv4);
#define NUMTHREADS 8
#define HASHCOUNT 1000000
typedef struct threadinfo {jwHashTable *table; int start;} threadinfo;
void * thread_func(void *arg)
{
threadinfo *info = arg;
char buffer[512];
int i = info->start;
jwHashTable *table = info->table;
free(info);
for(;i<HASHCOUNT;i+=NUMTHREADS) {
sprintf(buffer,"%d",i);
add_int_by_str(table,buffer,i);
}
}
int thread_test()
{
// create
jwHashTable * table = create_hash(HASHCOUNT>>2);
// hash a million strings into various sizes of table
struct timeval tval_before, tval_done1, tval_done2, tval_writehash, tval_readhash;
gettimeofday(&tval_before, NULL);
int t;
pthread_t * threads[NUMTHREADS];
for(t=0;t<NUMTHREADS;++t) {
pthread_t * pth = malloc(sizeof(pthread_t));
threads[t] = pth;
threadinfo *info = (threadinfo*)malloc(sizeof(threadinfo));
info->table = table; info->start = t;
pthread_create(pth,NULL,thread_func,info);
}
for(t=0;t<NUMTHREADS;++t) {
pthread_join(*threads[t], NULL);
}
gettimeofday(&tval_done1, NULL);
int i,j;
int error = 0;
char buffer[512];
for(i=0;i<HASHCOUNT;++i) {
sprintf(buffer,"%d",i);
get_int_by_str(table,buffer,&j);
if(i!=j) {
printf("Error: %d != %d\n",i,j);
error = 1;
}
}
if(!error) {
printf("No errors.\n");
}
gettimeofday(&tval_done2, NULL);
timersub(&tval_done1, &tval_before, &tval_writehash);
timersub(&tval_done2, &tval_done1, &tval_readhash);
printf("\n%d threads.\n",NUMTHREADS);
printf("Store %d ints by string: %ld.%06ld sec, read %d ints: %ld.%06ld sec\n",HASHCOUNT,
(long int)tval_writehash.tv_sec, (long int)tval_writehash.tv_usec,HASHCOUNT,
(long int)tval_readhash.tv_sec, (long int)tval_readhash.tv_usec);
return 0;
}