void add_to_search_space( State S, int op, int father )
{
/* see if state is already a part of this search space
*/
if ( state_hashed( S ) ) {
return;
}
source_to_dest( &(lsearch_space[lspace_end].S), S );
lsearch_space[lspace_end].op = op;
lsearch_space[lspace_end].father = father;
if ( father == -1 ) {
lsearch_space[lspace_end].depth = 1;
} else {
lsearch_space[lspace_end].depth = lsearch_space[father].depth + 1;
}
lspace_end++;
if ( lspace_end == MAX_SPACE ) {
printf("\nsearch space size too small! currently %d\n\n", MAX_SPACE);
exit( 1 );
}
lspace_size++;
hash_state( S );
}
Bool search_A_for_better_state( State S, int h, State *S_, int *h_ )
{
int i, h__;
State S__;
lspace_start = 0;
lspace_end = 0;
lspace_size = 0;
hash_state( S );
for ( i = 0; i < gnum_A; i++ ) {
result_to_dest( &S__, S, gA[i] );
add_to_search_space( S__, gA[i], -1 );
}
while ( TRUE ) {
if ( lspace_start == lspace_end ) {
return FALSE;
}
h__ = expand_A_first_node( h );
if ( LESS( h__, h ) ) {
break;
}
}
reset_hash_entrys();
extract_plan_fragment();
source_to_dest( S_, lsearch_space[lspace_start].S );
*h_ = h__;
if ( lsearch_space[lspace_start].depth > gmax_search_depth ) {
gmax_search_depth = lsearch_space[lspace_start].depth;
}
if ( lspace_size > gmax_search_size ) {
gmax_search_size = lspace_size;
}
return TRUE;
}
void
encodeValue (value *theValue, outputWriter *ow)
{
assert(theValue != 0);
switch (theValue->type)
{
case VALUE_UNDEFINED :
return;
case VALUE_INTERNAL :
OUT_STRING(ow, "<internal>", 10);
break;
case VALUE_BUILT_IN :
OUT_STRING(ow, "<builtIn>", 9);
break;
case VALUE_SCALAR :
encodeDynstring(&theValue->v.scalar.scalar, ow);
break;
case VALUE_LIST :
{
int i;
OUT_CHAR(ow, metaChar);
OUT_STRING(ow, "list(", 5);
if (valueListLength(theValue) > 0)
{
encodeValue(valueListGetElement(theValue, 0), ow);
for (i = 1; i < valueListLength(theValue); ++i)
{
OUT_CHAR(ow, ',');
encodeValue(valueListGetElement(theValue, i), ow);
}
}
OUT_CHAR(ow, ')');
}
break;
case VALUE_HASH :
{
value *aValue,
*keyValue;
char *aKey;
hstate state;
OUT_CHAR(ow, metaChar);
OUT_STRING(ow, "hash(", 5);
state = hash_state(theValue->v.hash.hash);
if ((aValue = (value*)hash_next(&state, &aKey)) != 0)
{
keyValue = valueNewScalarFromCString(aKey);
encodeValue(keyValue, ow);
OUT_CHAR(ow, ',');
encodeValue(aValue, ow);
while ((aValue = (value*)hash_next(&state, &aKey)) != 0)
{
OUT_CHAR(ow, ',');
keyValue = valueNewScalarFromCString(aKey);
encodeValue(keyValue, ow);
OUT_CHAR(ow, ',');
encodeValue(aValue, ow);
}
}
OUT_CHAR(ow, ')');
}
break;
case VALUE_LAMBDA :
{
int i;
OUT_CHAR(ow, metaChar);
OUT_STRING(ow, "lambda(", 7);
for (i = 0; i < theValue->v.lambda.numParams; ++i)
{
encodeDynstring(&theValue->v.lambda.paramNames[i], ow);
OUT_CHAR(ow, ',');
}
encodeBytecode(theValue->v.lambda.code, ow);
OUT_CHAR(ow, ')');
}
break;
case VALUE_ENVIRONMENT :
OUT_STRING(ow, "<environment>", 13);
break;
case VALUE_BYTECODE :
OUT_STRING(ow, "<bytecode>", 10);
break;
case VALUE_WHATSIT :
OUT_STRING(ow, "<whatsit>", 9);
break;
default :
assert(0);
}
}
int main(int argc, char **argv)
{
// time the program
struct timeval sysTimeStart, sysTimeEnd;
gettimeofday(&sysTimeStart, NULL);
mkdir("bplus", 0777);
int i;
// open file count information
file_count_t *fc = read_file_count();
// keep track of roots
bplus_roots_t bpr;
// users
{
printf("\n");
printf("Making user B+ tree with %d users\n", fc->users);
// time this section
struct timeval startSysTimeSub, endSysTimeSub;
gettimeofday(&startSysTimeSub, NULL);
// create root
int_node_t root;
root.next = -1;
root.previous = -1;
root.tableType = TABLE_TYPE_USER;
root.nodeType = NODE_TYPE_LEAF;
root.count = 0;
root.firstFile = -1;
write_node(0, &root);
int rootFileNum = 0;
for (i = 0; i < fc->users; i++)
{
user_t *user = read_user(i);
rootFileNum = insert_node(rootFileNum, TABLE_TYPE_USER, user->stateId, i);
free_user(user);
}
bpr.user = rootFileNum;
// end timing this section
gettimeofday(&endSysTimeSub, NULL);
float totalTime = (endSysTimeSub.tv_sec - startSysTimeSub.tv_sec)
+ (endSysTimeSub.tv_usec - startSysTimeSub.tv_usec) / 1000000.0f;
printf("B+ tree creation time: %f seconds\n", totalTime);
}
// cities
{
printf("\n");
printf("Making city B+ tree with %d cities\n", fc->cities);
// time this section
struct timeval startSysTimeSub, endSysTimeSub;
gettimeofday(&startSysTimeSub, NULL);
// create root
int_node_t root;
root.next = -1;
root.previous = -1;
root.tableType = TABLE_TYPE_CITY;
root.nodeType = NODE_TYPE_LEAF;
root.count = 0;
root.firstFile = -1;
write_node(0, &root);
int rootFileNum = 0;
for (i = 0; i < fc->cities; i++)
{
city_t *city = read_city(i);
rootFileNum = insert_node(rootFileNum, TABLE_TYPE_CITY, city->stateId, i);
free_city(city);
}
bpr.city = rootFileNum;
// end timing this section
gettimeofday(&endSysTimeSub, NULL);
float totalTime = (endSysTimeSub.tv_sec - startSysTimeSub.tv_sec)
+ (endSysTimeSub.tv_usec - startSysTimeSub.tv_usec) / 1000000.0f;
printf("B+ tree creation time: %f seconds\n", totalTime);
}
//states
{
printf("\n");
printf("Making state B+ tree with %d states\n", fc->states);
// time this section
struct timeval startSysTimeSub, endSysTimeSub;
gettimeofday(&startSysTimeSub, NULL);
// create root
int_node_t root;
root.next = -1;
root.previous = -1;
root.tableType = TABLE_TYPE_STATE;
root.nodeType = NODE_TYPE_LEAF;
root.count = 0;
root.firstFile = -1;
write_node(0, &root);
int rootFileNum = 0;
for (i = 0; i < fc->states; i++)
{
state_t *state = read_state(i);
rootFileNum = insert_node(rootFileNum, TABLE_TYPE_STATE, (int) hash_state(state), i);
free_state(state);
}
bpr.state = rootFileNum;
// end timing this section
gettimeofday(&endSysTimeSub, NULL);
float totalTime = (endSysTimeSub.tv_sec - startSysTimeSub.tv_sec)
+ (endSysTimeSub.tv_usec - startSysTimeSub.tv_usec) / 1000000.0f;
printf("B+ tree creation time: %f seconds\n", totalTime);
}
// messages
{
printf("\n");
printf("Making message B+ tree with %d messages\n", fc->messages);
// time this section
struct timeval startSysTimeSub, endSysTimeSub;
gettimeofday(&startSysTimeSub, NULL);
// create root
int_node_t root;
root.next = -1;
root.previous = -1;
root.tableType = TABLE_TYPE_MESSAGE;
root.nodeType = NODE_TYPE_LEAF;
root.count = 0;
root.firstFile = -1;
write_node(0, &root);
int rootFileNum = 0;
for (i = 0; i < fc->messages; i++)
{
message_t *message = read_message(i);
rootFileNum = insert_node(rootFileNum, TABLE_TYPE_MESSAGE, message->timestampId, i);
free_message(message);
}
bpr.message = rootFileNum;
// end timing this section
gettimeofday(&endSysTimeSub, NULL);
float totalTime = (endSysTimeSub.tv_sec - startSysTimeSub.tv_sec)
+ (endSysTimeSub.tv_usec - startSysTimeSub.tv_usec) / 1000000.0f;
printf("B+ tree creation time: %f seconds\n", totalTime);
}
// timestamps
{
printf("\n");
printf("Making timestamp B+ tree with %d timestamps\n", fc->timestamps);
// time this section
struct timeval startSysTimeSub, endSysTimeSub;
gettimeofday(&startSysTimeSub, NULL);
// create root
int_node_t root;
root.next = -1;
root.previous = -1;
root.tableType = TABLE_TYPE_TIMESTAMP;
root.nodeType = NODE_TYPE_LEAF;
root.count = 0;
root.firstFile = -1;
write_node(0, &root);
int rootFileNum = 0;
for (i = 0; i < fc->timestamps; i++)
{
timestamp_t *timestamp = read_timestamp(i);
rootFileNum = insert_node(rootFileNum, TABLE_TYPE_TIMESTAMP, (int) hash_timestamp(timestamp), i);
free_timestamp(timestamp);
}
bpr.timestamp = rootFileNum;
// end timing this section
gettimeofday(&endSysTimeSub, NULL);
float totalTime = (endSysTimeSub.tv_sec - startSysTimeSub.tv_sec)
+ (endSysTimeSub.tv_usec - startSysTimeSub.tv_usec) / 1000000.0f;
printf("B+ tree creation time: %f seconds\n", totalTime);
}
// datestamps
{
printf("\n");
printf("Making datestamp B+ tree with %d datestamps\n", fc->datestamps);
// time this section
struct timeval startSysTimeSub, endSysTimeSub;
gettimeofday(&startSysTimeSub, NULL);
// create root
int_node_t root;
root.next = -1;
root.previous = -1;
root.tableType = TABLE_TYPE_DATESTAMP;
root.nodeType = NODE_TYPE_LEAF;
root.count = 0;
root.firstFile = -1;
write_node(0, &root);
int rootFileNum = 0;
for (i = 0; i < fc->datestamps; i++)
{
datestamp_t *datestamp = read_datestamp(i);
rootFileNum = insert_node(rootFileNum, TABLE_TYPE_DATESTAMP, (int) hash_datestamp(datestamp), i);
free_datestamp(datestamp);
}
bpr.datestamp = rootFileNum;
// end timing this section
gettimeofday(&endSysTimeSub, NULL);
float totalTime = (endSysTimeSub.tv_sec - startSysTimeSub.tv_sec)
+ (endSysTimeSub.tv_usec - startSysTimeSub.tv_usec) / 1000000.0f;
printf("B+ tree creation time: %f seconds\n", totalTime);
}
free_file_count(fc);
printf("\n");
print_bplus_roots(&bpr);
write_bplus_roots(&bpr);
// end timing the program
gettimeofday(&sysTimeEnd, NULL);
float totalTime = (sysTimeEnd.tv_sec - sysTimeStart.tv_sec)
+ (sysTimeEnd.tv_usec - sysTimeStart.tv_usec) / 1000000.0f;
printf("Process time %f seconds\n", totalTime);
return 0;
}
int main(int argc, char **argv)
{
//print usage if needed
if (argc != 2) {
fprintf(stderr, "Usage: %s totalRecordNumber\n", argv[0]);
exit(0);
}
//get total record number from argument
int totalRecordNumber = atoi(argv[1]);
// time the program
struct timeval sysTimeStart, sysTimeEnd;
gettimeofday(&sysTimeStart, NULL);
// set up directories;
mkdir("cities", 0777);
mkdir("datestamps", 0777);
mkdir("messages", 0777);
mkdir("states", 0777);
mkdir("timestamps", 0777);
mkdir("users", 0777);
mkdir("bplus", 0777);
// set up some counters, etc.
unsigned int recordCount = 0,
userCount = 0,
cityCount = 0,
stateCount = 0,
timestampCount = 0,
datestampCount = 0,
messageCount = 0,
i,j;
// SET UP HASH TABLES FOR CITIES, STATES, TIMESTAMPS, DATESTAMPS
// cities
city_node *cityHT[HASH_SIZE];
for (i = 0; i < HASH_SIZE; i++)
{
cityHT[i] = malloc(sizeof(city_node));
cityHT[i] = NULL;
}
// states
state_node *stateHT[HASH_SIZE];
for (i = 0; i < HASH_SIZE; i++)
{
stateHT[i] = malloc(sizeof(state_node));
stateHT[i] = NULL;
}
// timestamps
timestamp_node *timestampHT[HASH_SIZE];
for (i = 0; i < HASH_SIZE; i++)
{
timestampHT[i] = malloc(sizeof(timestamp_node));
timestampHT[i] = NULL;
}
// datestamps
datestamp_node *datestampHT[HASH_SIZE];
for (i = 0; i < HASH_SIZE; i++)
{
datestampHT[i] = malloc(sizeof(datestamp_node));
datestampHT[i] = NULL;
}
// LOOP OVER RECORD FILES
char filename[1024];
FILE *fp = NULL;
for (i = 0; i < totalRecordNumber; i++) {
//open the corresponding file
sprintf(filename, "record_%06d.dat", i);
fp = fopen(filename,"rb");
if (!fp) {
fprintf(stderr, "Cannot open %s\n", filename);
continue;
}
record_t *record = read_record(fp);
// split location into city and state, as best we can
char cityStr[TEXT_SHORT];
char stateStr[TEXT_SHORT];
// there's one record where the location is \0, which strtok breaks on
if (record->location[0] == '\0')
{
strncpy(cityStr, "", TEXT_SHORT);
strncpy(stateStr, "", TEXT_SHORT);
}
else
{
char loc[TEXT_SHORT];
strncpy(loc, record->location, TEXT_SHORT);
strncpy(cityStr, strtok(loc, ","), TEXT_SHORT);
strncpy(stateStr, strtok(NULL, ","), TEXT_SHORT);
}
// create state
state_t state;
strncpy(state.name, stateStr, TEXT_SHORT);
// get stateId from hash if we have it already
unsigned int stateHash = hash_state(&state) % HASH_SIZE;
state_node *s;
int stateId = -1;
for(s = stateHT[stateHash]; (s != NULL) && (stateId == -1); s = s->next)
{
if (compare_states(&state, &(s->state)) == 0)
{
stateId = s->state.stateId;
}
}
// assign stateId, add to hash table, and write file if we don't have it
if (stateId == -1)
{
state.stateId = stateCount;
stateId = stateCount;
write_state(stateCount, &state);
stateCount++;
s = malloc(sizeof(state_node));
s->state = state;
s->next = stateHT[stateHash];
stateHT[stateHash] = s;
}
// create city
city_t city;
city.stateId = stateId;
strncpy(city.name, cityStr, TEXT_SHORT);
// get cityId from hash if we have it already
unsigned int cityHash = hash_city(&city) % HASH_SIZE;
city_node *c;
int cityId = -1;
for(c = cityHT[cityHash]; (c != NULL) && (cityId == -1); c = c->next)
{
if (compare_cities(&city, &(c->city)) == 0)
{
cityId = c->city.cityId;
}
}
// assign cityId, add to hash table, and write file if we don't have it
if (cityId == -1)
{
city.cityId = cityCount;
cityId = cityCount;
write_city(cityCount, &city);
cityCount++;
c = malloc(sizeof(city_node));
c->city = city;
c->next = cityHT[cityHash];
cityHT[cityHash] = c;
}
// create and write user
user_t user;
user.userId = record->id;
user.cityId = cityId;
user.stateId = stateId;
strncpy(user.name, record->name, TEXT_SHORT);
write_user(userCount, &user);
userCount++;
// loop over messages
for(j = 0; j < record->message_num; j++) {
// create timestamp
timestamp_t timestamp;
timestamp.hour = record->messages[j].hour;
timestamp.minute = record->messages[j].minute;
// get timestampId from hash if we have it already
unsigned int timestampHash = hash_timestamp(×tamp) % HASH_SIZE;
timestamp_node *t;
int tsId = -1;
for(t = timestampHT[timestampHash]; (t != NULL) && (tsId == -1); t = t->next)
{
if (compare_timestamps(×tamp, &(t->timestamp)) == 0)
{
tsId = t->timestamp.timestampId;
}
}
// assign timestampId, add to hash table, and write file if we don't have it
if (tsId == -1)
{
timestamp.timestampId = timestampCount;
tsId = timestampCount;
write_timestamp(timestampCount, ×tamp);
timestampCount++;
t = malloc(sizeof(timestamp_node));
t->timestamp = timestamp;
t->next = timestampHT[timestampHash];
timestampHT[timestampHash] = t;
}
// create datestamp
datestamp_t datestamp;
datestamp.year = record->messages[j].year;
datestamp.month = record->messages[j].month;
datestamp.day = record->messages[j].day;
// get datestampId from hash if we have it already
unsigned int datestampHash = hash_datestamp(&datestamp) % HASH_SIZE;
datestamp_node *d;
int dsId = -1;
for(d = datestampHT[datestampHash]; (d != NULL) && (dsId == -1); d = d->next)
{
if (compare_datestamps(&datestamp, &(d->datestamp)) == 0)
{
dsId = d->datestamp.datestampId;
}
}
// assign datestampId, add to hash table, and write file if we don't have it
if (dsId == -1)
{
datestamp.datestampId = datestampCount;
dsId = datestampCount;
write_datestamp(datestampCount, &datestamp);
datestampCount++;
d = malloc(sizeof(datestamp_node));
d->datestamp = datestamp;
d->next = datestampHT[datestampHash];
datestampHT[datestampHash] = d;
}
// create and write message
message_t message;
strncpy(message.text, record->messages[j].text, TEXT_LONG);
message.userId = user.userId;
message.timestampId = tsId;
message.datestampId = dsId;
message.messageId = messageCount;
write_message(messageCount, &message);
messageCount++;
}
// free and close record
free_record(record);
fclose(fp);
}
// free city nodes
city_node *cNode;
for (i = 0; i < HASH_SIZE; i++)
{
cNode = cityHT[i];
while (cNode != NULL)
{
city_node* tmp = cNode;
cNode = cNode->next;
free (tmp);
}
}
// free state nodes
state_node *sNode;
for (i = 0; i < HASH_SIZE; i++)
{
sNode = stateHT[i];
while (sNode != NULL)
{
state_node* tmp = sNode;
sNode = sNode->next;
free (tmp);
}
}
// free timestamp nodes
timestamp_node *tNode;
for (i = 0; i < HASH_SIZE; i++)
{
tNode = timestampHT[i];
while (tNode != NULL)
{
timestamp_node* tmp = tNode;
tNode = tNode->next;
free (tmp);
}
}
// free datestamp nodes
datestamp_node *dNode;
for (i = 0; i < HASH_SIZE; i++)
{
dNode = datestampHT[i];
while (dNode != NULL)
{
datestamp_node* tmp = dNode;
dNode = dNode->next;
free (tmp);
}
}
// create, write, print file count information file
file_count_t fc;
fc.users = userCount;
fc.cities = cityCount;
fc.states = stateCount;
fc.messages = messageCount;
fc.timestamps = timestampCount;
fc.datestamps = datestampCount;
write_file_count(&fc);
print_file_count(&fc);
// end timing the program
gettimeofday(&sysTimeEnd, NULL);
float totaltime2 = (sysTimeEnd.tv_sec - sysTimeStart.tv_sec)
+ (sysTimeEnd.tv_usec - sysTimeStart.tv_usec) / 1000000.0f;
printf("Process time %f seconds\n", totaltime2);
return 0;
}
