motiejus
/
turbonss
1
Fork 0
Code Releases Activity

Removed fingerprint methods and fixed pending bugs.

This commit is contained in:
davi 2008-04-28 01:18:23 +00:00
parent 9cb2e56a2f
commit 35b0a816dc
21 changed files with 6 additions and 743 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
src/bdz.c
View File

@ -599,33 +599,6 @@ void bdz_destroy(cmph_t *mphf)
free(mphf);
}
/** cmph_uint32 bdz_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
* \brief Computes the mphf value and a fingerprint of 12 bytes (i.e., figerprint should be a prealocated area to fit three 4-byte integers).
* \param mphf pointer to the resulting function
* \param key is the key to be hashed
* \param keylen is the key legth in bytes
* \return The mphf value
*
* Computes the mphf value and a fingerprint of 12 bytes. The figerprint pointer should be
* a prealocated area to fit three 4-byte integers. You don't need to use all the 12 bytes
* as fingerprint. According to the application, just few bits can be enough, once mphf does
* not allow collisions for the keys previously known.
*/
cmph_uint32 bdz_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint)
{
register cmph_uint32 vertex;
register bdz_data_t *bdz = mphf->data;
cmph_uint32 hl[3];
hash_vector(bdz->hl, key, keylen, hl);
memcpy(fingerprint, hl, sizeof(hl));
hl[0] = hl[0] % bdz->r;
hl[1] = hl[1] % bdz->r + bdz->r;
hl[2] = hl[2] % bdz->r + (bdz->r << 1);
vertex = hl[(GETVALUE(bdz->g, hl[0]) + GETVALUE(bdz->g, hl[1]) + GETVALUE(bdz->g, hl[2])) % 3];
return rank(bdz->b, bdz->ranktable, bdz->g, vertex);
}
/** \fn void bdz_pack(cmph_t *mphf, void *packed_mphf);
* \brief Support the ability to pack a perfect hash function into a preallocated contiguous memory space pointed by packed_mphf.
* \param mphf pointer to the resulting mphf

14
src/bdz.h
View File

@ -17,20 +17,6 @@ int bdz_dump(cmph_t *mphf, FILE *f);
void bdz_destroy(cmph_t *mphf);
cmph_uint32 bdz_search(cmph_t *mphf, const char *key, cmph_uint32 keylen);
/** cmph_uint32 bdz_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
* \brief Computes the mphf value and a fingerprint of 12 bytes (i.e., figerprint should be a prealocated area to fit three 4-byte integers).
* \param mphf pointer to the resulting function
* \param key is the key to be hashed
* \param keylen is the key legth in bytes
* \return The mphf value
*
* Computes the mphf value and a fingerprint of 12 bytes. The figerprint pointer should be
* a prealocated area to fit three 4-byte integers. You don't need to use all the 12 bytes
* as fingerprint. According to the application, just few bits can be enough, once mphf does
* not allow collisions for the keys previously known.
*/
cmph_uint32 bdz_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
/** \fn void bdz_pack(cmph_t *mphf, void *packed_mphf);
* \brief Support the ability to pack a perfect hash function into a preallocated contiguous memory space pointed by packed_mphf.
* \param mphf pointer to the resulting mphf

38
src/bdz_ph.c
View File

@ -527,44 +527,6 @@ void bdz_ph_destroy(cmph_t *mphf)
free(mphf);
}
/** cmph_uint32 bdz_ph_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
* \brief Computes the mphf value and a fingerprint of 12 bytes (i.e., figerprint should be a prealocated area to fit three 4-byte integers).
* \param mphf pointer to the resulting function
* \param key is the key to be hashed
* \param keylen is the key legth in bytes
* \return The mphf value
*
* Computes the mphf value and a fingerprint of 12 bytes. The figerprint pointer should be
* a prealocated area to fit three 4-byte integers. You don't need to use all the 12 bytes
* as fingerprint. According to the application, just few bits can be enough, once mphf does
* not allow collisions for the keys previously known.
*/
cmph_uint32 bdz_ph_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint)
{
register bdz_ph_data_t *bdz_ph = mphf->data;
cmph_uint32 hl[3];
register cmph_uint8 byte0, byte1, byte2;
register cmph_uint32 vertex;
hash_vector(bdz_ph->hl, key, keylen,hl);
memcpy(fingerprint, hl, sizeof(hl));
hl[0] = hl[0] % bdz_ph->r;
hl[1] = hl[1] % bdz_ph->r + bdz_ph->r;
hl[2] = hl[2] % bdz_ph->r + (bdz_ph->r << 1);
byte0 = bdz_ph->g[hl[0]/5];
byte1 = bdz_ph->g[hl[1]/5];
byte2 = bdz_ph->g[hl[2]/5];
byte0 = lookup_table[hl[0]%5][byte0];
byte1 = lookup_table[hl[1]%5][byte1];
byte2 = lookup_table[hl[2]%5][byte2];
vertex = hl[(byte0 + byte1 + byte2)%3];
return vertex;
}
/** \fn void bdz_ph_pack(cmph_t *mphf, void *packed_mphf);
* \brief Support the ability to pack a perfect hash function into a preallocated contiguous memory space pointed by packed_mphf.
* \param mphf pointer to the resulting mphf

14
src/bdz_ph.h
View File

@ -16,20 +16,6 @@ int bdz_ph_dump(cmph_t *mphf, FILE *f);
void bdz_ph_destroy(cmph_t *mphf);
cmph_uint32 bdz_ph_search(cmph_t *mphf, const char *key, cmph_uint32 keylen);
/** cmph_uint32 bdz_ph_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
* \brief Computes the mphf value and a fingerprint of 12 bytes (i.e., figerprint should be a prealocated area to fit three 4-byte integers).
* \param mphf pointer to the resulting function
* \param key is the key to be hashed
* \param keylen is the key legth in bytes
* \return The mphf value
*
* Computes the mphf value and a fingerprint of 12 bytes. The figerprint pointer should be
* a prealocated area to fit three 4-byte integers. You don't need to use all the 12 bytes
* as fingerprint. According to the application, just few bits can be enough, once mphf does
* not allow collisions for the keys previously known.
*/
cmph_uint32 bdz_ph_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
/** \fn void bdz_ph_pack(cmph_t *mphf, void *packed_mphf);
* \brief Support the ability to pack a perfect hash function into a preallocated contiguous memory space pointed by packed_mphf.
* \param mphf pointer to the resulting mphf

31
src/bmz.c
View File

@ -22,7 +22,7 @@ static void bmz_traverse_non_critical_nodes(bmz_config_data_t *bmz, cmph_uint8 *
bmz_config_data_t *bmz_config_new()
{
bmz_config_data_t *bmz;
bmz_config_data_t *bmz = NULL;
bmz = (bmz_config_data_t *)malloc(sizeof(bmz_config_data_t));
assert(bmz);
memset(bmz, 0, sizeof(bmz_config_data_t));
@ -539,35 +539,6 @@ void bmz_destroy(cmph_t *mphf)
free(mphf);
}
/** cmph_uint32 bmz_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
* \brief Computes the mphf value and a fingerprint of 12 bytes (i.e., figerprint should be a prealocated area to fit three 4-byte integers).
* \param mphf pointer to the resulting function
* \param key is the key to be hashed
* \param keylen is the key legth in bytes
* \return The mphf value
*
* Computes the mphf value and a fingerprint of 12 bytes. The figerprint pointer should be
* a prealocated area to fit three 4-byte integers. You don't need to use all the 12 bytes
* as fingerprint. According to the application, just few bits can be enough, once mphf does
* not allow collisions for the keys previously known.
*/
cmph_uint32 bmz_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint)
{
bmz_data_t *bmz = mphf->data;
cmph_uint32 h1, h2;
hash_vector(bmz->hashes[0], key, keylen, fingerprint);
h1 = fingerprint[2] % bmz->n;
hash_vector(bmz->hashes[1], key, keylen, fingerprint);
h2 = fingerprint[2] % bmz->n;
DEBUGP("key: %s h1: %u h2: %u\n", key, h1, h2);
if (h1 == h2 && ++h2 > bmz->n) h2 = 0;
DEBUGP("key: %s g[h1]: %u g[h2]: %u edges: %u\n", key, bmz->g[h1], bmz->g[h2], bmz->m);
return bmz->g[h1] + bmz->g[h2];
}
/** \fn void bmz_pack(cmph_t *mphf, void *packed_mphf);
* \brief Support the ability to pack a perfect hash function into a preallocated contiguous memory space pointed by packed_mphf.
* \param mphf pointer to the resulting mphf

14
src/bmz.h
View File

@ -16,20 +16,6 @@ int bmz_dump(cmph_t *mphf, FILE *f);
void bmz_destroy(cmph_t *mphf);
cmph_uint32 bmz_search(cmph_t *mphf, const char *key, cmph_uint32 keylen);
/** cmph_uint32 bmz_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
* \brief Computes the mphf value and a fingerprint of 12 bytes (i.e., figerprint should be a prealocated area to fit three 4-byte integers).
* \param mphf pointer to the resulting function
* \param key is the key to be hashed
* \param keylen is the key legth in bytes
* \return The mphf value
*
* Computes the mphf value and a fingerprint of 12 bytes. The figerprint pointer should be
* a prealocated area to fit three 4-byte integers. You don't need to use all the 12 bytes
* as fingerprint. According to the application, just few bits can be enough, once mphf does
* not allow collisions for the keys previously known.
*/
cmph_uint32 bmz_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
/** \fn void bmz_pack(cmph_t *mphf, void *packed_mphf);
* \brief Support the ability to pack a perfect hash function into a preallocated contiguous memory space pointed by packed_mphf.
* \param mphf pointer to the resulting mphf

29
src/bmz8.c
View File

@ -548,35 +548,6 @@ void bmz8_destroy(cmph_t *mphf)
free(mphf);
}
/** cmph_uint8 bmz8_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
* \brief Computes the mphf value and a fingerprint of 12 bytes (i.e., figerprint should be a prealocated area to fit three 4-byte integers).
* \param mphf pointer to the resulting function
* \param key is the key to be hashed
* \param keylen is the key legth in bytes
* \return The mphf value
*
* Computes the mphf value and a fingerprint of 12 bytes. The figerprint pointer should be
* a prealocated area to fit three 4-byte integers. You don't need to use all the 12 bytes
* as fingerprint. According to the application, just few bits can be enough, once mphf does
* not allow collisions for the keys previously known.
*/
cmph_uint8 bmz8_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint)
{
bmz8_data_t *bmz8 = mphf->data;
cmph_uint8 h1, h2;
hash_vector(bmz8->hashes[0], key, keylen, fingerprint);
h1 = fingerprint[2] % bmz8->n;
hash_vector(bmz8->hashes[1], key, keylen, fingerprint);
h2 = fingerprint[2] % bmz8->n;
DEBUGP("key: %s h1: %u h2: %u\n", key, h1, h2);
if (h1 == h2 && ++h2 > bmz8->n) h2 = 0;
DEBUGP("key: %s g[h1]: %u g[h2]: %u edges: %u\n", key, bmz8->g[h1], bmz8->g[h2], bmz8->m);
return bmz8->g[h1] + bmz8->g[h2];
}
/** \fn void bmz8_pack(cmph_t *mphf, void *packed_mphf);
* \brief Support the ability to pack a perfect hash function into a preallocated contiguous memory space pointed by packed_mphf.
* \param mphf pointer to the resulting mphf

14
src/bmz8.h
View File

@ -16,20 +16,6 @@ int bmz8_dump(cmph_t *mphf, FILE *f);
void bmz8_destroy(cmph_t *mphf);
cmph_uint8 bmz8_search(cmph_t *mphf, const char *key, cmph_uint32 keylen);
/** cmph_uint8 bmz8_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
* \brief Computes the mphf value and a fingerprint of 12 bytes (i.e., figerprint should be a prealocated area to fit three 4-byte integers).
* \param mphf pointer to the resulting function
* \param key is the key to be hashed
* \param keylen is the key legth in bytes
* \return The mphf value
*
* Computes the mphf value and a fingerprint of 12 bytes. The figerprint pointer should be
* a prealocated area to fit three 4-byte integers. You don't need to use all the 12 bytes
* as fingerprint. According to the application, just few bits can be enough, once mphf does
* not allow collisions for the keys previously known.
*/
cmph_uint8 bmz8_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
/** \fn void bmz8_pack(cmph_t *mphf, void *packed_mphf);
* \brief Support the ability to pack a perfect hash function into a preallocated contiguous memory space pointed by packed_mphf.
* \param mphf pointer to the resulting mphf

28
src/brz.c
View File

@ -712,32 +712,6 @@ void brz_destroy(cmph_t *mphf)
free(mphf);
}
/** cmph_uint32 brz_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
* \brief Computes the mphf value and a fingerprint of 12 bytes (i.e., figerprint should be a prealocated area to fit three 4-byte integers).
* \param mphf pointer to the resulting function
* \param key is the key to be hashed
* \param keylen is the key legth in bytes
* \return The mphf value
*
* Computes the mphf value and a fingerprint of 12 bytes. The figerprint pointer should be
* a prealocated area to fit three 4-byte integers. You don't need to use all the 12 bytes
* as fingerprint. According to the application, just few bits can be enough, once mphf does
* not allow collisions for the keys previously known.
*/
cmph_uint32 brz_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint)
{
brz_data_t *brz = mphf->data;
switch(brz->algo)
{
case CMPH_FCH:
return brz_fch_search(brz, key, keylen, fingerprint);
case CMPH_BMZ8:
return brz_bmz8_search(brz, key, keylen, fingerprint);
default: assert(0);
}
return 0;
}
/** \fn void brz_pack(cmph_t *mphf, void *packed_mphf);
* \brief Support the ability to pack a perfect hash function into a preallocated contiguous memory space pointed by packed_mphf.
* \param mphf pointer to the resulting mphf
@ -791,7 +765,7 @@ void brz_pack(cmph_t *mphf, void *packed_mphf)
#if defined (__ia64) || defined (__x86_64__)
cmph_uint64 * g_is_ptr = (cmph_uint64 *)ptr;
#else
cmph_uint32 * g_is_ptr = ptr;
cmph_uint32 * g_is_ptr = (cmph_uint32 *)ptr;
#endif
cmph_uint8 * g_i = (cmph_uint8 *) (g_is_ptr + data->k);

14
src/brz.h
View File

@ -21,20 +21,6 @@ int brz_dump(cmph_t *mphf, FILE *f);
void brz_destroy(cmph_t *mphf);
cmph_uint32 brz_search(cmph_t *mphf, const char *key, cmph_uint32 keylen);
/** cmph_uint32 brz_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
* \brief Computes the mphf value and a fingerprint of 12 bytes (i.e., figerprint should be a prealocated area to fit three 4-byte integers).
* \param mphf pointer to the resulting function
* \param key is the key to be hashed
* \param keylen is the key legth in bytes
* \return The mphf value
*
* Computes the mphf value and a fingerprint of 12 bytes. The figerprint pointer should be
* a prealocated area to fit three 4-byte integers. You don't need to use all the 12 bytes
* as fingerprint. According to the application, just few bits can be enough, once mphf does
* not allow collisions for the keys previously known.
*/
cmph_uint32 brz_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
/** \fn void brz_pack(cmph_t *mphf, void *packed_mphf);
* \brief Support the ability to pack a perfect hash function into a preallocated contiguous memory space pointed by packed_mphf.
* \param mphf pointer to the resulting mphf

34
src/chm.c
View File

@ -19,10 +19,10 @@ static void chm_traverse(chm_config_data_t *chm, cmph_uint8 *visited, cmph_uint3
chm_config_data_t *chm_config_new()
{
chm_config_data_t *chm;
chm_config_data_t *chm = NULL;
chm = (chm_config_data_t *)malloc(sizeof(chm_config_data_t));
assert(chm);
memset(chm,0,sizeof(chm_config_data_t));
memset(chm, 0, sizeof(chm_config_data_t));
chm->hashfuncs[0] = CMPH_HASH_JENKINS;
chm->hashfuncs[1] = CMPH_HASH_JENKINS;
chm->g = NULL;
@ -293,36 +293,6 @@ void chm_destroy(cmph_t *mphf)
free(mphf);
}
/** cmph_uint32 chm_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
* \brief Computes the mphf value and a fingerprint of 12 bytes (i.e., figerprint should be a prealocated area to fit three 4-byte integers).
* \param mphf pointer to the resulting function
* \param key is the key to be hashed
* \param keylen is the key legth in bytes
* \return The mphf value
*
* Computes the mphf value and a fingerprint of 12 bytes. The figerprint pointer should be
* a prealocated area to fit three 4-byte integers. You don't need to use all the 12 bytes
* as fingerprint. According to the application, just few bits can be enough, once mphf does
* not allow collisions for the keys previously known.
*/
cmph_uint32 chm_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint)
{
chm_data_t *chm = mphf->data;
cmph_uint32 h1, h2;
hash_vector(chm->hashes[0], key, keylen, fingerprint);
h1 = fingerprint[2] % chm->n;
hash_vector(chm->hashes[1], key, keylen, fingerprint);
h2 = fingerprint[2] % chm->n;
DEBUGP("key: %s h1: %u h2: %u\n", key, h1, h2);
if (h1 == h2 && ++h2 >= chm->n) h2 = 0;
DEBUGP("key: %s g[h1]: %u g[h2]: %u edges: %u\n", key, chm->g[h1], chm->g[h2], chm->m);
return (chm->g[h1] + chm->g[h2]) % chm->m;
}
/** \fn void chm_pack(cmph_t *mphf, void *packed_mphf);
* \brief Support the ability to pack a perfect hash function into a preallocated contiguous memory space pointed by packed_mphf.
* \param mphf pointer to the resulting mphf

14
src/chm.h
View File

@ -16,20 +16,6 @@ int chm_dump(cmph_t *mphf, FILE *f);
void chm_destroy(cmph_t *mphf);
cmph_uint32 chm_search(cmph_t *mphf, const char *key, cmph_uint32 keylen);
/** cmph_uint32 chm_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
* \brief Computes the mphf value and a fingerprint of 12 bytes (i.e., figerprint should be a prealocated area to fit three 4-byte integers).
* \param mphf pointer to the resulting function
* \param key is the key to be hashed
* \param keylen is the key legth in bytes
* \return The mphf value
*
* Computes the mphf value and a fingerprint of 12 bytes. The figerprint pointer should be
* a prealocated area to fit three 4-byte integers. You don't need to use all the 12 bytes
* as fingerprint. According to the application, just few bits can be enough, once mphf does
* not allow collisions for the keys previously known.
*/
cmph_uint32 chm_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
/** \fn void chm_pack(cmph_t *mphf, void *packed_mphf);
* \brief Support the ability to pack a perfect hash function into a preallocated contiguous memory space pointed by packed_mphf.
* \param mphf pointer to the resulting mphf

48
src/cmph.c
View File

@ -612,54 +612,6 @@ cmph_uint32 cmph_search(cmph_t *mphf, const char *key, cmph_uint32 keylen)
return 0;
}
/** cmph_uint32 cmph_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
* \brief Computes the mphf value and a fingerprint of 12 bytes (i.e., figerprint should be a prealocated area to fit three 4-byte integers).
* \param mphf pointer to the resulting function
* \param key is the key to be hashed
* \param keylen is the key legth in bytes
* \return The mphf value
*
* Computes the mphf value and a fingerprint of 12 bytes. The figerprint pointer should be
* a prealocated area to fit three 4-byte integers. You don't need to use all the 12 bytes
* as fingerprint. According to the application, just few bits can be enough, once mphf does
* not allow collisions for the keys previously known.
*/
cmph_uint32 cmph_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint)
{
DEBUGP("mphf algorithm: %u \n", mphf->algo);
switch(mphf->algo)
{
case CMPH_CHM:
return chm_search_fingerprint(mphf, key, keylen, fingerprint);
case CMPH_BMZ: /* included -- Fabiano */
DEBUGP("bmz algorithm search\n");
return bmz_search_fingerprint(mphf, key, keylen, fingerprint);
case CMPH_BMZ8: /* included -- Fabiano */
DEBUGP("bmz8 algorithm search\n");
return bmz8_search_fingerprint(mphf, key, keylen, fingerprint);
case CMPH_BRZ: /* included -- Fabiano */
DEBUGP("brz algorithm search\n");
return brz_search_fingerprint(mphf, key, keylen, fingerprint);
case CMPH_FCH: /* included -- Fabiano */
DEBUGP("fch algorithm search\n");
return fch_search_fingerprint(mphf, key, keylen, fingerprint);
case CMPH_BDZ: /* included -- Fabiano */
DEBUGP("bdz algorithm search\n");
return bdz_search_fingerprint(mphf, key, keylen, fingerprint);
case CMPH_BDZ_PH: /* included -- Fabiano */
DEBUGP("bdz_ph algorithm search\n");
return bdz_ph_search_fingerprint(mphf, key, keylen, fingerprint);
default:
assert(0);
}
assert(0);
return 0;
}
cmph_uint32 cmph_size(cmph_t *mphf)
{
return mphf->size;

16
src/cmph.h
View File

@ -69,22 +69,6 @@ cmph_t *cmph_new(cmph_config_t *mph);
*/
cmph_uint32 cmph_search(cmph_t *mphf, const char *key, cmph_uint32 keylen);
/** cmph_uint32 cmph_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
* \brief Computes the mphf value and a fingerprint of 12 bytes (i.e.,
* \brief figerprint should be a prealocated area to fit three 4-byte integers).
* \param mphf pointer to the resulting function
* \param key is the key to be hashed
* \param keylen is the key legth in bytes
* \return The mphf value
*
* Computes the mphf value and a fingerprint of 12 bytes. The figerprint pointer should be
* a prealocated area to fit three 4-byte integers. You don't need to use all the 12 bytes
* as fingerprint. According to the application, just few bits can be enough, once mphf does
* not allow collisions for the keys previously known.
*/
cmph_uint32 cmph_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
cmph_uint32 cmph_size(cmph_t *mphf);
void cmph_destroy(cmph_t *mphf);

27
src/fch.c
View File

@ -411,33 +411,6 @@ void fch_destroy(cmph_t *mphf)
free(mphf);
}
/** cmph_uint32 fch_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
* \brief Computes the mphf value and a fingerprint of 12 bytes (i.e., figerprint should be a prealocated area to fit three 4-byte integers).
* \param mphf pointer to the resulting function
* \param key is the key to be hashed
* \param keylen is the key legth in bytes
* \return The mphf value
*
* Computes the mphf value and a fingerprint of 12 bytes. The figerprint pointer should be
* a prealocated area to fit three 4-byte integers. You don't need to use all the 12 bytes
* as fingerprint. According to the application, just few bits can be enough, once mphf does
* not allow collisions for the keys previously known.
*/
cmph_uint32 fch_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint)
{
register fch_data_t *fch = mphf->data;
hash_vector(fch->h1, key, keylen, fingerprint);
register cmph_uint32 h1 = fingerprint[2] % fch->m;
hash_vector(fch->h2, key, keylen, fingerprint);
register cmph_uint32 h2 = fingerprint[2] % fch->m;
h1 = mixh10h11h12 (fch->b, fch->p1, fch->p2, h1);
//DEBUGP("key: %s h1: %u h2: %u g[h1]: %u\n", key, h1, h2, fch->g[h1]);
return (h2 + fch->g[h1]) % fch->m;
}
/** \fn void fch_pack(cmph_t *mphf, void *packed_mphf);
* \brief Support the ability to pack a perfect hash function into a preallocated contiguous memory space pointed by packed_mphf.
* \param mphf pointer to the resulting mphf

14
src/fch.h
View File

@ -22,20 +22,6 @@ int fch_dump(cmph_t *mphf, FILE *f);
void fch_destroy(cmph_t *mphf);
cmph_uint32 fch_search(cmph_t *mphf, const char *key, cmph_uint32 keylen);
/** cmph_uint32 fch_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
* \brief Computes the mphf value and a fingerprint of 12 bytes (i.e., figerprint should be a prealocated area to fit three 4-byte integers).
* \param mphf pointer to the resulting function
* \param key is the key to be hashed
* \param keylen is the key legth in bytes
* \return The mphf value
*
* Computes the mphf value and a fingerprint of 12 bytes. The figerprint pointer should be
* a prealocated area to fit three 4-byte integers. You don't need to use all the 12 bytes
* as fingerprint. According to the application, just few bits can be enough, once mphf does
* not allow collisions for the keys previously known.
*/
cmph_uint32 fch_search_fingerprint(cmph_t *mphf, const char *key, cmph_uint32 keylen, cmph_uint32 * fingerprint);
/** \fn void fch_pack(cmph_t *mphf, void *packed_mphf);
* \brief Support the ability to pack a perfect hash function into a preallocated contiguous memory space pointed by packed_mphf.
* \param mphf pointer to the resulting mphf

2
src/main.c
View File

@ -1,5 +1,5 @@
#ifdef WIN32
#include "../wingetopt.h"
#include "wingetopt.h"
#else
#include <getopt.h>
#endif

5
tests/Makefile.am
View File

@ -1,4 +1,4 @@
noinst_PROGRAMS = graph_tests packed_mphf_tests mphf_tests mphf_fingerprint_tests
noinst_PROGRAMS = graph_tests packed_mphf_tests mphf_tests
INCLUDES = -I../src/
@ -10,6 +10,3 @@ packed_mphf_tests_LDADD = ../src/libcmph.la
mphf_tests_SOURCES = mphf_tests.c
mphf_tests_LDADD = ../src/libcmph.la
mphf_fingerprint_tests_SOURCES = mphf_fingerprint_tests.c
mphf_fingerprint_tests_LDADD = ../src/libcmph.la

162
tests/mphf_fingerprint_tests.c
View File

@ -1,162 +0,0 @@
#ifdef WIN32
#include "../wingetopt.h"
#else
#include <getopt.h>
#endif
#include <stdio.h>
#include <stdlib.h>
#include <errno.h>
#include <string.h>
#include <time.h>
#include <limits.h>
#include <assert.h>
#include <cmph.h>
#ifdef WIN32
#define VERSION "0.8"
#else
#include "config.h"
#endif
void usage(const char *prg)
{
fprintf(stderr, "usage: %s [-v] [-h] [-V] [-k nkeys] [-m file.mph] keysfile\n", prg);
}
void usage_long(const char *prg)
{
fprintf(stderr, "usage: %s [-v] [-h] [-V] [-k nkeys] [-m file.mph] keysfile\n", prg);
fprintf(stderr, "Packed MPHFs testing tool\n\n");
fprintf(stderr, " -h\t print this help message\n");
fprintf(stderr, " -V\t print version number and exit\n");
fprintf(stderr, " -v\t increase verbosity (may be used multiple times)\n");
fprintf(stderr, " -k\t number of keys\n");
fprintf(stderr, " -m\t minimum perfect hash function file \n");
fprintf(stderr, " keysfile\t line separated file with keys\n");
}
int main(int argc, char **argv)
{
char verbosity = 0;
char *mphf_file = NULL;
const char *keys_file = NULL;
FILE *mphf_fd = stdout;
FILE *keys_fd;
cmph_uint32 nkeys = UINT_MAX;
cmph_uint32 i = 0;
cmph_t *mphf = NULL;
cmph_io_adapter_t *source;
cmph_uint32 fingerprint[3];
while (1)
{
char ch = getopt(argc, argv, "hVvk:m:");
if (ch == -1) break;
switch (ch)
{
case 'k':
{
char *endptr;
nkeys = 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 'v':
++verbosity;
break;
case 'V':
printf("%s\n", VERSION);
return 0;
case 'h':
usage_long(argv[0]);
return 0;
default:
usage(argv[0]);
return 1;
}
}
if (optind != argc - 1)
{
usage(argv[0]);
return 1;
}
keys_file = argv[optind];
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(nkeys == UINT_MAX) source = cmph_io_nlfile_adapter(keys_fd);
else source = cmph_io_nlnkfile_adapter(keys_fd, nkeys);
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*)malloc(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_fingerprint(mphf, buf, buflen, fingerprint);
if (!(h < siz))
{
fprintf(stderr, "Unknown key %*s in the input.\n", buflen, buf);
ret = 1;
} else if(hashtable[h])
{
fprintf(stderr, "Duplicated or unknown key %*s in the input\n", buflen, buf);
ret = 1;
} else hashtable[h] = 1;
if (verbosity)
{
printf("%s -> %u -- fingerprint: %u %u %u\n", buf, h, fingerprint[0], fingerprint[1], fingerprint[2]);
}
source->dispose(source->data, buf, buflen);
}
cmph_destroy(mphf);
free(hashtable);
fclose(keys_fd);
free(mphf_file);
cmph_io_nlfile_adapter_destroy(source);
return ret;
}

179
wingetopt.c
View File

@ -1,179 +0,0 @@
#ifdef WIN32
/*****************************************************************************
*
* MODULE NAME : GETOPT.C
*
* COPYRIGHTS:
* This module contains code made available by IBM
* Corporation on an AS IS basis. Any one receiving the
* module is considered to be licensed under IBM copyrights
* to use the IBM-provided source code in any way he or she
* deems fit, including copying it, compiling it, modifying
* it, and redistributing it, with or without
* modifications. No license under any IBM patents or
* patent applications is to be implied from this copyright
* license.
*
* A user of the module should understand that IBM cannot
* provide technical support for the module and will not be
* responsible for any consequences of use of the program.
*
* Any notices, including this one, are not to be removed
* from the module without the prior written consent of
* IBM.
*
* AUTHOR: Original author:
* G. R. Blair (BOBBLAIR at AUSVM1)
* Internet: bobblair@bobblair.austin.ibm.com
*
* Extensively revised by:
* John Q. Walker II, Ph.D. (JOHHQ at RALVM6)
* Internet: johnq@ralvm6.vnet.ibm.com
*
*****************************************************************************/
/******************************************************************************
* getopt()
*
* The getopt() function is a command line parser. It returns the next
* option character in argv that matches an option character in opstring.
*
* The argv argument points to an array of argc+1 elements containing argc
* pointers to character strings followed by a null pointer.
*
* The opstring argument points to a string of option characters; if an
* option character is followed by a colon, the option is expected to have
* an argument that may or may not be separated from it by white space.
* The external variable optarg is set to point to the start of the option
* argument on return from getopt().
*
* The getopt() function places in optind the argv index of the next argument
* to be processed. The system initializes the external variable optind to
* 1 before the first call to getopt().
*
* When all options have been processed (that is, up to the first nonoption
* argument), getopt() returns EOF. The special option "--" may be used to
* delimit the end of the options; EOF will be returned, and "--" will be
* skipped.
*
* The getopt() function returns a question mark (?) when it encounters an
* option character not included in opstring. This error message can be
* disabled by setting opterr to zero. Otherwise, it returns the option
* character that was detected.
*
* If the special option "--" is detected, or all options have been
* processed, EOF is returned.
*
* Options are marked by either a minus sign (-) or a slash (/).
*
* No errors are defined.
*****************************************************************************/
#include <stdio.h> /* for EOF */
#include <string.h> /* for strchr() */
/* static (global) variables that are specified as exported by getopt() */
extern char *optarg; /* pointer to the start of the option argument */
extern int optind; /* number of the next argv[] to be evaluated */
extern int opterr; /* non-zero if a question mark should be returned
when a non-valid option character is detected */
/* handle possible future character set concerns by putting this in a macro */
#define _next_char(string) (char)(*(string+1))
int getopt(int argc, char *argv[], char *opstring)
{
static char *pIndexPosition = NULL; /* place inside current argv string */
char *pArgString = NULL; /* where to start from next */
char *pOptString; /* the string in our program */
if (pIndexPosition != NULL) {
/* we last left off inside an argv string */
if (*(++pIndexPosition)) {
/* there is more to come in the most recent argv */
pArgString = pIndexPosition;
}
}
if (pArgString == NULL) {
/* we didn't leave off in the middle of an argv string */
if (optind >= argc) {
/* more command-line arguments than the argument count */
pIndexPosition = NULL; /* not in the middle of anything */
return EOF; /* used up all command-line arguments */
}
/*---------------------------------------------------------------------
* If the next argv[] is not an option, there can be no more options.
*-------------------------------------------------------------------*/
pArgString = argv[optind++]; /* set this to the next argument ptr */
if (('/' != *pArgString) && /* doesn't start with a slash or a dash? */
('-' != *pArgString)) {
--optind; /* point to current arg once we're done */
optarg = NULL; /* no argument follows the option */
pIndexPosition = NULL; /* not in the middle of anything */
return EOF; /* used up all the command-line flags */
}
/* check for special end-of-flags markers */
if ((strcmp(pArgString, "-") == 0) ||
(strcmp(pArgString, "--") == 0)) {
optarg = NULL; /* no argument follows the option */
pIndexPosition = NULL; /* not in the middle of anything */
return EOF; /* encountered the special flag */
}
pArgString++; /* look past the / or - */
}
if (':' == *pArgString) { /* is it a colon? */
/*---------------------------------------------------------------------
* Rare case: if opterr is non-zero, return a question mark;
* otherwise, just return the colon we're on.
*-------------------------------------------------------------------*/
return (opterr ? (int)'?' : (int)':');
}
else if ((pOptString = strchr(opstring, *pArgString)) == 0) {
/*---------------------------------------------------------------------
* The letter on the command-line wasn't any good.
*-------------------------------------------------------------------*/
optarg = NULL; /* no argument follows the option */
pIndexPosition = NULL; /* not in the middle of anything */
return (opterr ? (int)'?' : (int)*pArgString);
}
else {
/*---------------------------------------------------------------------
* The letter on the command-line matches one we expect to see
*-------------------------------------------------------------------*/
if (':' == _next_char(pOptString)) { /* is the next letter a colon? */
/* It is a colon. Look for an argument string. */
if ('\0' != _next_char(pArgString)) { /* argument in this argv? */
optarg = &pArgString[1]; /* Yes, it is */
}
else {
/*-------------------------------------------------------------
* The argument string must be in the next argv.
* But, what if there is none (bad input from the user)?
* In that case, return the letter, and optarg as NULL.
*-----------------------------------------------------------*/
if (optind < argc)
optarg = argv[optind++];
else {
optarg = NULL;
return (opterr ? (int)'?' : (int)*pArgString);
}
}
pIndexPosition = NULL; /* not in the middle of anything */
}
else {
/* it's not a colon, so just return the letter */
optarg = NULL; /* no argument follows the option */
pIndexPosition = pArgString; /* point to the letter we're on */
}
return (int)*pArgString; /* return the letter that matched */
}
}
#endif //WIN32

25
wingetopt.h
View File

@ -1,25 +0,0 @@
#ifdef __cplusplus
extern "C" {
#endif
#ifndef WIN32
#include <getopt.h>
#else
#ifndef _GETOPT_
#define _GETOPT_
#include <stdio.h> /* for EOF */
#include <string.h> /* for strchr() */
char *optarg = NULL; /* pointer to the start of the option argument */
int optind = 1; /* number of the next argv[] to be evaluated */
int opterr = 1; /* non-zero if a question mark should be returned */
int getopt(int argc, char *argv[], char *opstring);
#endif //_GETOPT_
#endif //WIN32
#ifdef __cplusplus
}
#endif