char *oph_predicate2(UDF_INIT * initid, UDF_ARGS * args, char *result, unsigned long *length, char *is_null, char *error)
{
oph_string measure;
char *buffer;
char **names;
int count;
oph_predicate2_param *param;
int i = 0;
if (core_set_type(&(measure), args->args[0], &(args->lengths[0]))) {
pmesg(1, __FILE__, __LINE__, "Type not recognized\n");
*length = 0;
*is_null = 0;
*error = 1;
return NULL;
}
measure.content = args->args[2];
measure.length = &(args->lengths[2]);
measure.missingvalue = NULL;
core_set_elemsize(&(measure));
if (core_set_numelem(&(measure))) {
pmesg(1, __FILE__, __LINE__, "Error on counting elements\n");
*length = 0;
*is_null = 0;
*error = 1;
return NULL;
}
if (!initid->ptr) {
initid->ptr = (char *) malloc(sizeof(oph_predicate2_param));
if (!initid->ptr) {
pmesg(1, __FILE__, __LINE__, "Error allocating result string\n");
*length = 0;
*is_null = 0;
*error = 1;
return NULL;
}
param = (oph_predicate2_param *) initid->ptr;
for (i = 0; i < 4; ++i)
param->f[i] = NULL;
param->occurrence = 0; // ALL
buffer = (char *) malloc(1 + args->lengths[3]);
strncpy(buffer, args->args[3], args->lengths[3]);
buffer[args->lengths[3]] = '\0';
pthread_rwlock_wrlock(&lock);
param->f[1] = evaluator_create(buffer);
pthread_rwlock_unlock(&lock);
free(buffer);
if (!param->f[1]) {
pmesg(1, __FILE__, __LINE__, "Error allocating evaluator\n");
*length = 0;
*is_null = 0;
*error = 1;
return NULL;
}
evaluator_get_variables(param->f[1], &names, &count);
if (count > 1) {
pmesg(1, __FILE__, __LINE__, "Too variables in expression\n");
*length = 0;
*is_null = 0;
*error = 1;
return NULL;
}
// Comparison operator
if (core_set_comp(¶m->op, args->args[4], &(args->lengths[4]))) {
pmesg(1, __FILE__, __LINE__, "Comparison operator not recognized\n");
*length = 0;
*is_null = 0;
*error = 1;
return NULL;
}
if (args->args[5] && args->lengths[5]) {
buffer = (char *) malloc(1 + args->lengths[5]);
core_strncpy(buffer, args->args[5], &(args->lengths[5]));
if (!strcasecmp(buffer, "nan"))
sprintf(buffer, "0/0");
} else
buffer = strdup("0/0");
pthread_rwlock_wrlock(&lock);
param->f[2] = evaluator_create(buffer);
pthread_rwlock_unlock(&lock);
free(buffer);
if (!param->f[2]) {
pmesg(1, __FILE__, __LINE__, "Error allocating evaluator\n");
*length = 0;
*is_null = 0;
*error = 1;
return NULL;
}
evaluator_get_variables(param->f[2], &names, &count);
if (count > 1) {
pmesg(1, __FILE__, __LINE__, "Too variables in expression\n");
*length = 0;
*is_null = 0;
*error = 1;
return NULL;
}
if (args->args[6] && args->lengths[6]) {
buffer = (char *) malloc(1 + args->lengths[6]);
core_strncpy(buffer, args->args[6], &(args->lengths[6]));
if (!strcasecmp(buffer, "nan"))
sprintf(buffer, "0/0");
} else
buffer = strdup("0/0");
pthread_rwlock_wrlock(&lock);
param->f[3] = evaluator_create(buffer);
pthread_rwlock_unlock(&lock);
free(buffer);
if (!param->f[3]) {
pmesg(1, __FILE__, __LINE__, "Error allocating evaluator\n");
*length = 0;
*is_null = 0;
*error = 1;
return NULL;
}
evaluator_get_variables(param->f[3], &names, &count);
if (count > 1) {
pmesg(1, __FILE__, __LINE__, "Too variables in expression\n");
*length = 0;
*is_null = 0;
*error = 1;
return NULL;
}
oph_string output_array;
core_set_type(&output_array, args->args[1], &(args->lengths[1]));
if (!output_array.type) {
pmesg(1, __FILE__, __LINE__, "Unable to recognize measures type\n");
*length = 0;
*is_null = 0;
*error = 1;
return NULL;
}
if (core_set_elemsize(&output_array)) {
pmesg(1, __FILE__, __LINE__, "Unable to recognize measures type\n");
*length = 0;
*is_null = 0;
*error = 1;
return NULL;
}
param->result_type = output_array.type;
param->result_elemsize = output_array.elemsize;
param->length = output_array.elemsize * measure.numelem;
param->f[0] = malloc(param->length);
if (!param->f[0]) {
pmesg(1, __FILE__, __LINE__, "Error allocating result string\n");
*length = 0;
*is_null = 0;
*error = 1;
return NULL;
}
if (args->arg_count > 7) {
buffer = (char *) malloc(1 + args->lengths[7]);
core_strncpy(buffer, args->args[7], &(args->lengths[7]));
if (strcasecmp(buffer, OPH_PREDICATE2_ALL_OCCURRENCE)) {
if (!strcasecmp(buffer, OPH_PREDICATE2_FIRST_OCCURRENCE) || !strcasecmp(buffer, OPH_PREDICATE2_BEGIN_OCCURRENCE))
param->occurrence = 1;
else if (!strcasecmp(buffer, OPH_PREDICATE2_LAST_OCCURRENCE) || !strcasecmp(buffer, OPH_PREDICATE2_END_OCCURRENCE))
param->occurrence = -1;
else {
if (args->arg_type[7] == STRING_RESULT)
param->occurrence = (long) strtol(buffer, NULL, 10);
else
param->occurrence = *((long long *) args->args[7]);
if (param->occurrence < 1) {
free(buffer);
pmesg(1, __FILE__, __LINE__, "Unable to read occurrence\n");
*length = 0;
*is_null = 0;
*error = 1;
return NULL;
}
}
}
free(buffer);
}
} else
param = (oph_predicate2_param *) initid->ptr;
i = core_oph_predicate2(&measure, initid->ptr);
if (i) {
pmesg(1, __FILE__, __LINE__, "Unable to compute result\n");
*length = 0;
*is_null = 0;
*error = 1;
return NULL;
}
*length = param->length;
*error = 0;
*is_null = 0;
return (char *) param->f[0];
}
void oph_aggregate_stats_partial_add(UDF_INIT * initid, UDF_ARGS * args, char *is_null, char *error)
{
if (*error != 0)
return;
if (args->args[2]) {
oph_agg_stats_partial_data *dat = (oph_agg_stats_partial_data *) initid->ptr;
/* Setting of the aggregate result */
if (!dat->result.content) {
//It's the first row
// default values
unsigned long def_mask_len = MASK_LEN;
dat->mask.content = (char *) calloc(MASK_LEN + 1, sizeof(char));
if (!dat->mask.content) {
pmesg(1, __FILE__, __LINE__, "Error allocating mask content\n");
*error = 1;
return;
}
core_strncpy(dat->mask.content, DEFAULT_MASK, &def_mask_len);
dat->mask.length = (unsigned long *) calloc(1, sizeof(unsigned long));
if (!dat->mask.length) {
pmesg(1, __FILE__, __LINE__, "Error allocating mask length\n");
*error = 1;
return;
}
*(dat->mask.length) = def_mask_len;
core_set_type(&(dat->measure), NULL, 0);
dat->measure.length = (unsigned long *) calloc(1, sizeof(unsigned long));
if (!dat->measure.length) {
pmesg(1, __FILE__, __LINE__, "Error allocating measure length\n");
*error = 1;
return;
}
*(dat->measure.length) = args->lengths[2];
core_set_type(&(dat->measure), args->args[0], &(args->lengths[0]));
if (dat->measure.type != OPH_SHORT && dat->measure.type != OPH_BYTE && dat->measure.type != OPH_INT && dat->measure.type != OPH_LONG && dat->measure.type != OPH_FLOAT
&& dat->measure.type != OPH_DOUBLE) {
pmesg(1, __FILE__, __LINE__, "Invalid type\n");
*error = 1;
return;
}
if (args->arg_count > 3) {
core_strncpy(dat->mask.content, args->args[3], &(args->lengths[3]));
*(dat->mask.length) = args->lengths[3];
}
if (core_set_elemsize(&(dat->measure))) {
pmesg(1, __FILE__, __LINE__, "Error on setting measure elements size\n");
*error = 1;
return;
}
if (core_set_numelem(&(dat->measure))) {
pmesg(1, __FILE__, __LINE__, "Error on counting measure elements\n");
*error = 1;
return;
}
core_set_type(&dat->result, args->args[1], &(args->lengths[1]));
if (!dat->result.type) {
pmesg(1, __FILE__, __LINE__, "Unable to recognize measures type\n");
*error = 1;
return;
}
if (core_set_elemsize(&dat->result)) {
pmesg(1, __FILE__, __LINE__, "Unable to recognize measures type\n");
*error = 1;
return;
}
// mask processing
int i;
for (i = 0; i < *(dat->mask.length); i++) {
if (dat->mask.content[i] == '1') {
(dat->mask.numelem)++; // count 1s
switch (i) {
case 0: // mean
dat->sum1 = 1; // sum{x_i} needed
break;
case 1: // variance
dat->sum1 = 1;
dat->sum2 = 1; // sum{(x_i)^2} needed
break;
case 2: // std dev
dat->sum1 = 1;
dat->sum2 = 1;
break;
case 3: // skew
dat->sum1 = 1;
dat->sum2 = 1;
dat->sum3 = 1; // sum{(x_i)^3} needed
break;
case 4: // kurtosis
dat->sum1 = 1;
dat->sum2 = 1;
dat->sum3 = 1;
dat->sum4 = 1; // sum{(x_i)^4} needed
break;
case 5: // max
dat->max = 1; // array with max values needed
break;
case 6: // min
dat->min = 1; // array with min values needed
}
}
}
if (dat->mask.numelem == 0) {
pmesg(1, __FILE__, __LINE__, "Invalid mask\n");
*error = 1;
return;
}
int size = 1;
if (dat->sum1)
size++;
if (dat->sum2)
size++;
if (dat->sum3)
size++;
if (dat->sum4)
size++;
if (dat->max)
size++;
if (dat->min)
size++;
// output array allocation
dat->result.numelem = size * dat->measure.numelem; // In future use a structed type of 'size' fields
unsigned long outlen = dat->result.numelem * dat->result.elemsize;
dat->result.length = (unsigned long *) calloc(1, sizeof(unsigned long));
if (!dat->result.length) {
pmesg(1, __FILE__, __LINE__, "Error allocating result length\n");
*error = 1;
return;
}
*(dat->result.length) = outlen;
dat->result.content = (char *) calloc(1, *(dat->result.length));
if (!dat->result.content) {
pmesg(1, __FILE__, __LINE__, "Error allocating output array\n");
*error = 1;
return;
}
// partial results array allocation
dat->partials = (oph_stringPtr) calloc((size - 1), sizeof(oph_string));
if (!dat->partials) {
pmesg(1, __FILE__, __LINE__, "Error allocating intermediate arrays\n");
*error = 1;
return;
}
for (i = 0; i < (size - 1); i++) {
dat->partials[i].type = dat->measure.type;
dat->partials[i].elemsize = dat->measure.elemsize;
dat->partials[i].numelem = dat->measure.numelem;
dat->partials[i].length = dat->measure.length;
dat->partials[i].content = (char *) calloc(1, *(dat->partials[i].length));
if (!dat->partials[i].content) {
pmesg(1, __FILE__, __LINE__, "Error allocating intermediate array\n");
*error = 1;
return;
}
}
}
if (!dat->count) {
//It's the first row or the first row in the group of the GROUP BY clause
int i = 0;
int j;
// copy input values
if (dat->sum1) {
memcpy((void *) (dat->partials[i].content), (void *) (args->args[2]), *(dat->partials[i].length));
i++;
}
// copy input values squared
if (dat->sum2) {
switch (dat->measure.type) {
case OPH_INT:
{
int val_i;
for (j = 0; j < dat->measure.numelem; j++) {
val_i = ((int *) (args->args[2]))[j];
((int *) (dat->partials[i].content))[j] = val_i * val_i;
}
break;
}
case OPH_SHORT:
{
int val_i;
for (j = 0; j < dat->measure.numelem; j++) {
val_i = ((short *) (args->args[2]))[j];
((short *) (dat->partials[i].content))[j] = val_i * val_i;
}
break;
}
case OPH_BYTE:
{
int val_i;
for (j = 0; j < dat->measure.numelem; j++) {
val_i = ((char *) (args->args[2]))[j];
((char *) (dat->partials[i].content))[j] = val_i * val_i;
}
break;
}
case OPH_LONG:
{
long long val_l;
for (j = 0; j < dat->measure.numelem; j++) {
val_l = ((long long *) (args->args[2]))[j];
((long long *) (dat->partials[i].content))[j] = val_l * val_l;
}
break;
}
case OPH_FLOAT:
{
float val_f;
for (j = 0; j < dat->measure.numelem; j++) {
val_f = ((float *) (args->args[2]))[j];
((float *) (dat->partials[i].content))[j] = val_f * val_f;
}
break;
}
case OPH_DOUBLE:
{
double val_d;
for (j = 0; j < dat->measure.numelem; j++) {
val_d = ((double *) (args->args[2]))[j];
((double *) (dat->partials[i].content))[j] = val_d * val_d;
}
}
case OPH_COMPLEX_INT:
case OPH_COMPLEX_LONG:
case OPH_COMPLEX_FLOAT:
case OPH_COMPLEX_DOUBLE:
case INVALID_TYPE:
break;
}
i++;
}
// copy input values raised to 3
if (dat->sum3) {
switch (dat->measure.type) {
case OPH_INT:
{
int val_i;
for (j = 0; j < dat->measure.numelem; j++) {
val_i = ((int *) (args->args[2]))[j];
((int *) (dat->partials[i].content))[j] = val_i * val_i * val_i;
}
break;
}
case OPH_SHORT:
{
short val_i;
for (j = 0; j < dat->measure.numelem; j++) {
val_i = ((short *) (args->args[2]))[j];
((short *) (dat->partials[i].content))[j] = val_i * val_i * val_i;
}
break;
}
case OPH_BYTE:
{
char val_i;
for (j = 0; j < dat->measure.numelem; j++) {
val_i = ((char *) (args->args[2]))[j];
((char *) (dat->partials[i].content))[j] = val_i * val_i * val_i;
}
break;
}
case OPH_LONG:
{
long long val_l;
for (j = 0; j < dat->measure.numelem; j++) {
val_l = ((long long *) (args->args[2]))[j];
((long long *) (dat->partials[i].content))[j] = val_l * val_l * val_l;
}
break;
}
case OPH_FLOAT:
{
float val_f;
for (j = 0; j < dat->measure.numelem; j++) {
val_f = ((float *) (args->args[2]))[j];
((float *) (dat->partials[i].content))[j] = val_f * val_f * val_f;
}
break;
}
case OPH_DOUBLE:
{
double val_d;
for (j = 0; j < dat->measure.numelem; j++) {
val_d = ((double *) (args->args[2]))[j];
((double *) (dat->partials[i].content))[j] = val_d * val_d * val_d;
}
}
case OPH_COMPLEX_INT:
case OPH_COMPLEX_LONG:
case OPH_COMPLEX_FLOAT:
case OPH_COMPLEX_DOUBLE:
case INVALID_TYPE:
break;
}
i++;
}
// copy input values raised to 4
if (dat->sum4) {
switch (dat->measure.type) {
case OPH_INT:
{
int val_i;
for (j = 0; j < dat->measure.numelem; j++) {
val_i = ((int *) (args->args[2]))[j];
((int *) (dat->partials[i].content))[j] = val_i * val_i * val_i * val_i;
}
break;
}
case OPH_SHORT:
{
short val_i;
for (j = 0; j < dat->measure.numelem; j++) {
val_i = ((short *) (args->args[2]))[j];
((short *) (dat->partials[i].content))[j] = val_i * val_i * val_i * val_i;
}
break;
}
case OPH_BYTE:
{
char val_i;
for (j = 0; j < dat->measure.numelem; j++) {
val_i = ((char *) (args->args[2]))[j];
((char *) (dat->partials[i].content))[j] = val_i * val_i * val_i * val_i;
}
break;
}
case OPH_LONG:
{
long long val_l;
for (j = 0; j < dat->measure.numelem; j++) {
val_l = ((long long *) (args->args[2]))[j];
((long long *) (dat->partials[i].content))[j] = val_l * val_l * val_l * val_l;
}
break;
}
case OPH_FLOAT:
{
float val_f;
for (j = 0; j < dat->measure.numelem; j++) {
val_f = ((float *) (args->args[2]))[j];
((float *) (dat->partials[i].content))[j] = val_f * val_f * val_f * val_f;
}
break;
}
case OPH_DOUBLE:
{
double val_d;
for (j = 0; j < dat->measure.numelem; j++) {
val_d = ((double *) (args->args[2]))[j];
((double *) (dat->partials[i].content))[j] = val_d * val_d * val_d * val_d;
}
}
case OPH_COMPLEX_INT:
case OPH_COMPLEX_LONG:
case OPH_COMPLEX_FLOAT:
case OPH_COMPLEX_DOUBLE:
case INVALID_TYPE:
break;
}
i++;
}
// copy input values
if (dat->max) {
memcpy(dat->partials[i].content, (void *) (args->args[2]), *(dat->partials[i].length));
i++;
}
// copy input values
if (dat->min) {
memcpy(dat->partials[i].content, (void *) (args->args[2]), *(dat->partials[i].length));
i++;
}
} else {
// Not the first row in the group => execute next aggregation step
dat->measure.content = args->args[2];
int i = 0;
if (dat->sum1) {
if (core_oph_sum_array(&(dat->measure), &(dat->partials[i]), dat->partials[i].content)) {
pmesg(1, __FILE__, __LINE__, "Unable to compute result\n");
*error = 1;
return;
}
i++;
}
if (dat->sum2) {
if (core_oph_sum2_array(&(dat->measure), &(dat->partials[i]), dat->partials[i].content)) {
pmesg(1, __FILE__, __LINE__, "Unable to compute result\n");
*error = 1;
return;
}
i++;
}
if (dat->sum3) {
if (core_oph_sum3_array(&(dat->measure), &(dat->partials[i]), dat->partials[i].content)) {
pmesg(1, __FILE__, __LINE__, "Unable to compute result\n");
*error = 1;
return;
}
i++;
}
if (dat->sum4) {
if (core_oph_sum4_array(&(dat->measure), &(dat->partials[i]), dat->partials[i].content)) {
pmesg(1, __FILE__, __LINE__, "Unable to compute result\n");
*error = 1;
return;
}
i++;
}
if (dat->max) {
if (core_oph_max_array(&(dat->measure), &(dat->partials[i]), dat->partials[i].content)) {
pmesg(1, __FILE__, __LINE__, "Unable to compute result\n");
*error = 1;
return;
}
i++;
}
if (dat->min) {
if (core_oph_min_array(&(dat->measure), &(dat->partials[i]), dat->partials[i].content)) {
pmesg(1, __FILE__, __LINE__, "Unable to compute result\n");
*error = 1;
return;
}
i++;
}
}
dat->count++;
}
}
