int main(int argc, char **argv)
{
cmph_uint32 verbosity = 0;
char generate = 0;
char *mphf_file = NULL;
FILE *mphf_fd = stdout;
const char *keys_file = NULL;
FILE *keys_fd;
cmph_uint32 nkeys = UINT_MAX;
cmph_uint32 seed = UINT_MAX;
CMPH_HASH *hashes = NULL;
cmph_uint32 nhashes = 0;
cmph_uint32 i;
CMPH_ALGO mph_algo = CMPH_CHM;
double c = 0;
cmph_config_t *config = NULL;
cmph_t *mphf = NULL;
char * tmp_dir = NULL;
cmph_io_adapter_t *source;
cmph_uint32 memory_availability = 0;
cmph_uint32 b = 0;
cmph_uint32 keys_per_bin = 1;
while (1)
{
char ch = (char)getopt(argc, argv, "hVvgc:k:a:M:b:t:f:m:d:s:");
if (ch == -1) break;
switch (ch)
{
case 's':
{
char *cptr;
seed = (cmph_uint32)strtoul(optarg, &cptr, 10);
if(*cptr != 0) {
fprintf(stderr, "Invalid seed %s\n", optarg);
exit(1);
}
}
break;
case 'c':
{
char *endptr;
c = strtod(optarg, &endptr);
if(*endptr != 0) {
fprintf(stderr, "Invalid c value %s\n", optarg);
exit(1);
}
}
break;
case 'g':
generate = 1;
break;
case 'k':
{
char *endptr;
nkeys = (cmph_uint32)strtoul(optarg, &endptr, 10);
if(*endptr != 0) {
fprintf(stderr, "Invalid number of keys %s\n", optarg);
exit(1);
}
}
break;
case 'm':
mphf_file = strdup(optarg);
break;
case 'd':
tmp_dir = strdup(optarg);
break;
case 'M':
{
char *cptr;
memory_availability = (cmph_uint32)strtoul(optarg, &cptr, 10);
if(*cptr != 0) {
fprintf(stderr, "Invalid memory availability %s\n", optarg);
exit(1);
}
}
break;
case 'b':
{
char *cptr;
b = (cmph_uint32)strtoul(optarg, &cptr, 10);
if(*cptr != 0) {
fprintf(stderr, "Parameter b was not found: %s\n", optarg);
exit(1);
}
}
break;
case 't':
{
char *cptr;
keys_per_bin = (cmph_uint32)strtoul(optarg, &cptr, 10);
if(*cptr != 0) {
fprintf(stderr, "Parameter t was not found: %s\n", optarg);
exit(1);
}
}
break;
case 'v':
++verbosity;
break;
case 'V':
printf("%s\n", VERSION);
return 0;
case 'h':
usage_long(argv[0]);
return 0;
case 'a':
{
char valid = 0;
for (i = 0; i < CMPH_COUNT; ++i)
{
if (strcmp(cmph_names[i], optarg) == 0)
{
mph_algo = (CMPH_ALGO)i;
valid = 1;
break;
}
}
if (!valid)
{
fprintf(stderr, "Invalid mph algorithm: %s. It is not available in version %s\n", optarg, VERSION);
return -1;
}
}
break;
case 'f':
{
char valid = 0;
for (i = 0; i < CMPH_HASH_COUNT; ++i)
{
if (strcmp(cmph_hash_names[i], optarg) == 0)
{
hashes = (CMPH_HASH *)realloc(hashes, sizeof(CMPH_HASH) * ( nhashes + 2 ));
hashes[nhashes] = (CMPH_HASH)i;
hashes[nhashes + 1] = CMPH_HASH_COUNT;
++nhashes;
valid = 1;
break;
}
}
if (!valid)
{
fprintf(stderr, "Invalid hash function: %s\n", optarg);
return -1;
}
}
break;
default:
usage(argv[0]);
return 1;
}
}
if (optind != argc - 1)
{
usage(argv[0]);
return 1;
}
keys_file = argv[optind];
if (seed == UINT_MAX) seed = (cmph_uint32)time(NULL);
srand(seed);
int ret = 0;
if (mphf_file == NULL)
{
mphf_file = (char *)malloc(strlen(keys_file) + 5);
memcpy(mphf_file, keys_file, strlen(keys_file));
memcpy(mphf_file + strlen(keys_file), ".mph\0", (size_t)5);
}
keys_fd = fopen(keys_file, "r");
if (keys_fd == NULL)
{
fprintf(stderr, "Unable to open file %s: %s\n", keys_file, strerror(errno));
return -1;
}
if (seed == UINT_MAX) seed = (cmph_uint32)time(NULL);
if(nkeys == UINT_MAX) source = cmph_io_nlfile_adapter(keys_fd);
else source = cmph_io_nlnkfile_adapter(keys_fd, nkeys);
if (generate)
{
//Create mphf
mphf_fd = fopen(mphf_file, "w");
config = cmph_config_new(source);
cmph_config_set_algo(config, mph_algo);
if (nhashes) cmph_config_set_hashfuncs(config, hashes);
cmph_config_set_verbosity(config, verbosity);
cmph_config_set_tmp_dir(config, (cmph_uint8 *) tmp_dir);
cmph_config_set_mphf_fd(config, mphf_fd);
cmph_config_set_memory_availability(config, memory_availability);
cmph_config_set_b(config, b);
cmph_config_set_keys_per_bin(config, keys_per_bin);
//if((mph_algo == CMPH_BMZ || mph_algo == CMPH_BRZ) && c >= 2.0) c=1.15;
if(mph_algo == CMPH_BMZ && c >= 2.0) c=1.15;
if (c != 0) cmph_config_set_graphsize(config, c);
mphf = cmph_new(config);
cmph_config_destroy(config);
if (mphf == NULL)
{
fprintf(stderr, "Unable to create minimum perfect hashing function\n");
//cmph_config_destroy(config);
free(mphf_file);
return -1;
}
if (mphf_fd == NULL)
{
fprintf(stderr, "Unable to open output file %s: %s\n", mphf_file, strerror(errno));
free(mphf_file);
return -1;
}
cmph_dump(mphf, mphf_fd);
cmph_destroy(mphf);
fclose(mphf_fd);
}
else
{
cmph_uint8 * hashtable = NULL;
mphf_fd = fopen(mphf_file, "r");
if (mphf_fd == NULL)
{
fprintf(stderr, "Unable to open input file %s: %s\n", mphf_file, strerror(errno));
free(mphf_file);
return -1;
}
mphf = cmph_load(mphf_fd);
fclose(mphf_fd);
if (!mphf)
{
fprintf(stderr, "Unable to parser input file %s\n", mphf_file);
free(mphf_file);
return -1;
}
cmph_uint32 siz = cmph_size(mphf);
hashtable = (cmph_uint8*)calloc(siz, sizeof(cmph_uint8));
memset(hashtable, 0,(size_t) siz);
//check all keys
for (i = 0; i < source->nkeys; ++i)
{
cmph_uint32 h;
char *buf;
cmph_uint32 buflen = 0;
source->read(source->data, &buf, &buflen);
h = cmph_search(mphf, buf, buflen);
if (!(h < siz))
{
fprintf(stderr, "Unknown key %*s in the input.\n", buflen, buf);
ret = 1;
} else if(hashtable[h] >= keys_per_bin)
{
fprintf(stderr, "More than %u keys were mapped to bin %u\n", keys_per_bin, h);
fprintf(stderr, "Duplicated or unknown key %*s in the input\n", buflen, buf);
ret = 1;
} else hashtable[h]++;
if (verbosity)
{
printf("%s -> %u\n", buf, h);
}
source->dispose(source->data, buf, buflen);
}
cmph_destroy(mphf);
free(hashtable);
}
fclose(keys_fd);
free(mphf_file);
free(tmp_dir);
cmph_io_nlfile_adapter_destroy(source);
return ret;
}
cmph_t *chd_new(cmph_config_t *mph, double c)
{
cmph_t *mphf = NULL;
chd_data_t *chdf = NULL;
chd_config_data_t *chd = (chd_config_data_t *)mph->data;
chd_ph_config_data_t * chd_ph = (chd_ph_config_data_t *)chd->chd_ph->data;
compressed_rank_t cr;
register cmph_t * chd_phf = NULL;
register cmph_uint32 packed_chd_phf_size = 0;
cmph_uint8 * packed_chd_phf = NULL;
register cmph_uint32 packed_cr_size = 0;
cmph_uint8 * packed_cr = NULL;
register cmph_uint32 i, idx, nkeys, nvals, nbins;
cmph_uint32 * vals_table = NULL;
register cmph_uint32 * occup_table = NULL;
#ifdef CMPH_TIMING
double construction_time_begin = 0.0;
double construction_time = 0.0;
ELAPSED_TIME_IN_SECONDS(&construction_time_begin);
#endif
cmph_config_set_verbosity(chd->chd_ph, mph->verbosity);
cmph_config_set_graphsize(chd->chd_ph, c);
if (mph->verbosity)
{
fprintf(stderr, "Generating a CHD_PH perfect hash function with a load factor equal to %.3f\n", c);
}
chd_phf = cmph_new(chd->chd_ph);
if(chd_phf == NULL)
{
return NULL;
}
packed_chd_phf_size = cmph_packed_size(chd_phf);
DEBUGP("packed_chd_phf_size = %u\n", packed_chd_phf_size);
/* Make sure that we have enough space to pack the mphf. */
packed_chd_phf = calloc((size_t)packed_chd_phf_size,(size_t)1);
/* Pack the mphf. */
cmph_pack(chd_phf, packed_chd_phf);
cmph_destroy(chd_phf);
if (mph->verbosity)
{
fprintf(stderr, "Compressing the range of the resulting CHD_PH perfect hash function\n");
}
compressed_rank_init(&cr);
nbins = chd_ph->n;
nkeys = chd_ph->m;
nvals = nbins - nkeys;
vals_table = (cmph_uint32 *)calloc(nvals, sizeof(cmph_uint32));
occup_table = (cmph_uint32 *)chd_ph->occup_table;
for(i = 0, idx = 0; i < nbins; i++)
{
if(!GETBIT32(occup_table, i))
{
vals_table[idx++] = i;
}
}
compressed_rank_generate(&cr, vals_table, nvals);
free(vals_table);
packed_cr_size = compressed_rank_packed_size(&cr);
packed_cr = (cmph_uint8 *) calloc(packed_cr_size, sizeof(cmph_uint8));
compressed_rank_pack(&cr, packed_cr);
compressed_rank_destroy(&cr);
mphf = (cmph_t *)malloc(sizeof(cmph_t));
mphf->algo = mph->algo;
chdf = (chd_data_t *)malloc(sizeof(chd_data_t));
chdf->packed_cr = packed_cr;
packed_cr = NULL; //transfer memory ownership
chdf->packed_chd_phf = packed_chd_phf;
packed_chd_phf = NULL; //transfer memory ownership
chdf->packed_chd_phf_size = packed_chd_phf_size;
chdf->packed_cr_size = packed_cr_size;
mphf->data = chdf;
mphf->size = nkeys;
DEBUGP("Successfully generated minimal perfect hash\n");
if (mph->verbosity)
{
fprintf(stderr, "Successfully generated minimal perfect hash function\n");
}
#ifdef CMPH_TIMING
ELAPSED_TIME_IN_SECONDS(&construction_time);
register cmph_uint32 space_usage = chd_packed_size(mphf)*8;
construction_time = construction_time - construction_time_begin;
fprintf(stdout, "%u\t%.2f\t%u\t%.4f\t%.4f\n", nkeys, c, chd_ph->keys_per_bucket, construction_time, space_usage/(double)nkeys);
#endif
return mphf;
}
