PQ *
enlarge_pqueue(PQ **pq, puint32 nbucks)
{
PQ *q1, *q=*pq;
puint32 i;
int first_i;
puint32 bucket;
q1=new_pqueue(nbucks, q->l.nelems, q->l.cost);
if (!q1) {
animal_err_register ("enlarge_pqueue", ANIMAL_ERROR_FAILURE,"");
return NULL;
}
q1->c.mincost = q->c.mincost;
q1->c.maxcost = q->c.maxcost;
q1->c.tiebreak = q->c.tiebreak;
for (i=0; i < q->c.nbuckets+1; i++)
if (q->c.first[i] != NIL) {
first_i = q->c.first[i];
bucket = q->l.cost[first_i] % q->c.nbuckets;
q1->c.first[bucket] = first_i;
q1->c.last[bucket] = q->c.last[i];
}
for (i=0; i < q->l.nelems; i++)
q1->l.elem[i] = q->l.elem[i];
free_pqueue(pq);
return q1;
}
void prediction_stat (char *AB, pst_type T, const int n, const char *sfile,
const char *ifile, const char *s0file, const char *i0file, const int s1,
const int s2, int s3, double *non1p, double *non2p, double *non12p,
double *iso1p, double *iso2p, double *iso12p)
{
int e1, e2, e3, i, cbuflen, dum1, dum2;
char cbuf[STRINGLEN_MAX+1], f1, f2, f12;
double a, amin, iso3;
pQ_type Q;
FILE *fp, *ifp;
amin = -DBL_MAX;
Q = make_empty_pqueue();
e1 = 0; /* train */
e2 = 0; /* test */
e3 = 0; /* others */
if ((fp = fopen(s0file, "r")) == NULL)
{
perror("failed to open others strings file for read");
exit(1);
}
if ((ifp = fopen(i0file, "r")) == NULL)
{
perror("failed to open others names file for read");
exit(1);
}
dum1 = 0;
while ((dum1<s3) && (fscanf(fp,"%s",cbuf) != EOF))
{
if ((cbuflen = strlen(cbuf)) > STRINGLEN_MAX)
{
fprintf(stderr, "single string size exceeds STRINGLEN_MAX = %d (may update and re-compile)\n", STRINGLEN_MAX);
exit(1);
}
if (strspn(cbuf,AB)!=cbuflen)
{
fprintf(stderr,"found a rejected string\n");
exit(1);
}
if ((fscanf(ifp,"%*s %*s %*d %d",&i)) == EOF)
/* id acc len cln */
{
fprintf(stderr,"names file is shorter then others string file\n");
exit(1);
}
if (i != n)
{
dum1++;
a = log10like_on_pst(AB, T, cbuf)/cbuflen;
if (a>amin)
amin = a;
enqueue_pqueue(&Q, NULL, a, 3, 0,0);
}
}
fclose(fp);
fclose(ifp);
/* we disable so that PFAM could be run...
if (dum1 < s3)
{
fprintf(stderr,"not enough strings in 0.strings\n");
exit(1);
}
*/
if ((fp = fopen(sfile, "r")) == NULL)
{
perror("failed to open cluster <strings_file> for read");
exit(1);
}
if ((ifp = fopen(ifile, "r")) == NULL)
{
perror("failed to open cluster <index_file> for read");
exit(1);
}
while (fscanf(fp,"%s",cbuf) != EOF)
{
if ((cbuflen = strlen(cbuf)) > STRINGLEN_MAX)
{
fprintf(stderr, "single string size exceeds STRINGLEN_MAX = %d (may update and re-compile)\n", STRINGLEN_MAX);
exit(1);
}
if (strspn(cbuf,AB)!=cbuflen)
{
fprintf(stderr,"found a rejected string\n");
exit(1);
}
if ((fscanf(ifp,"%d",&i)) == EOF) /* 0=train and 1=test */
{
fprintf(stderr,"<index_file> is shorter then <string_file>\n");
exit(1);
}
a = log10like_on_pst(AB, T, cbuf)/cbuflen;
if (a<=amin)
{
enqueue_pqueue(&Q, NULL, a, i+1, 0,0);
((i == 0) ? e1++ : e2++);
}
}
fclose(fp);
fclose(ifp);
*non1p = 1.0*e1; /* /s1; */
*non2p = 1.0*e2; /* /s2; */
*non12p = 1.0*(e1+e2); /* /(s1+s2); */
*iso1p = *non1p;
*iso2p = *non2p;
*iso12p = *non12p;
iso3 = 0.0;
f1 = (*iso1p > iso3);
f2 = (*iso2p > iso3);
f12 = (*iso12p > iso3);
while (f1+f2+f12 > 0)
{
dequeue_pqueue(&Q, NULL, &a, &i, &dum1, &dum2);
switch(i)
{
case 1:
e1--;
if (f1)
*iso1p = 1.0*e1; /* /s1; */
if (f12)
*iso12p = 1.0*(e1+e2); /* /(s1+s2); */
break;
case 2:
e2--;
if (f2)
*iso2p = 1.0*e2; /* /s2; */
if (f12)
*iso12p = 1.0*(e1+e2); /* /(s1+s2); */
break;
case 3:
e3++;
iso3 = 1.0*e3; /* /s3; */
break;
}
f1 = (*iso1p > iso3);
f2 = (*iso2p > iso3);
f12 = (*iso12p > iso3);
}
free_pqueue(Q);
free(cbuf);
return;
}
void pred_stat_save (char *AB, pst_type T, const int n, const char *sfile,
const char *ifile, const char *s0file, const char *i0file, int s1,
int s2, int s3, double *non1p, double *non2p, double *non12p,
double *iso1p, double *iso2p, double *iso12p, const char *wfile,
const char *w0file, const int tsize)
{
int e1, e2, e3, i, cbuflen, dum1, dum2;
char cbuf[STRINGLEN_MAX+1], f1, f2, f12, nam[20], acc[20]; /* PFAM disabled comb[40]; */
double a, amin, iso3, y;
pQ_type Q; /* PFAM disabled Qh; */
FILE *fp, *ifp, *wfp;
amin = -DBL_MAX;
Q = make_empty_pqueue();
e1 = 0; /* train */
e2 = 0; /* test */
e3 = 0; /* others */
if ((wfp = fopen(w0file, "w")) == NULL)
{
perror("failed to open others output file for write");
exit(1);
}
if ((fp = fopen(s0file, "r")) == NULL)
{
perror("failed to open others strings file for read");
exit(1);
}
if ((ifp = fopen(i0file, "r")) == NULL)
{
perror("failed to open others names file for read");
exit(1);
}
dum1 = 0;
while (fscanf(fp,"%s",cbuf) != EOF)
{
if ((cbuflen = strlen(cbuf)) > STRINGLEN_MAX)
{
fprintf(stderr, "single string size exceeds STRINGLEN_MAX = %d (may update and re-compile)\n", STRINGLEN_MAX);
exit(1);
}
if (strspn(cbuf,AB)!=cbuflen)
{
fprintf(stderr,"found a rejected string\n");
exit(1);
}
if ((fscanf(ifp,"%s %s %d %d",nam, acc, &dum2, &i)) == EOF)
{
fprintf(stderr,"names file is shorter then others string file\n");
exit(1);
}
if ((i != n) && (dum1<s3))
{
dum1++;
y = log10like_on_pst(AB, T, cbuf);
fprintf(wfp, "%f\n", y);
/* a = y/cbuflen;
if (a>amin)
amin = a;
sprintf(comb, "%15s %10s", nam, acc);
enqueue_pqueue(&Q, comb, a, 3, i, dum2); */
} /* id_acc kind cluster len */
else
fprintf(wfp, "%f\n", 1.0);
}
fclose(fp);
fclose(ifp);
fclose(wfp);
/*
printf("\nthe %d most related others strings:\n \t slope\t len\tcluster\tname\n", tsize/2);
Qh = Q;
for (i=1; (i<=tsize/2) && (Qh!=NULL); i++)
{
printf("(%d)\t%5.3f\t%5d\t%5d\t%s\n", i, -Qh->p_S, Qh->chi_sufSx,
Qh->chi_xS,Qh->S);
Qh = Qh->nextp;
}
if (dum1 < s3)
{
fprintf(stderr,"not enough strings in 0.strings\n");
exit(1);
}
*/
if ((wfp = fopen(wfile, "w")) == NULL)
{
perror("failed to open cluster output file for write");
exit(1);
}
if ((fp = fopen(sfile, "r")) == NULL)
{
perror("failed to open cluster <strings_file> for read");
exit(1);
}
if ((ifp = fopen(ifile, "r")) == NULL)
{
perror("failed to open cluster <index_file> for read");
exit(1);
}
while (fscanf(fp,"%s",cbuf) != EOF)
{
if ((cbuflen = strlen(cbuf)) > STRINGLEN_MAX)
{
fprintf(stderr, "single string size exceeds STRINGLEN_MAX = %d (may update and re-compile)\n", STRINGLEN_MAX);
exit(1);
}
if (strspn(cbuf,AB)!=cbuflen)
{
fprintf(stderr,"found a rejected string\n");
exit(1);
}
if ((fscanf(ifp,"%d",&i)) == EOF) /* 0=train and 1=test */
{
fprintf(stderr,"<index_file> is shorter then <string_file>\n");
exit(1);
}
y = log10like_on_pst(AB, T, cbuf);
fprintf (wfp, "%f\n", y);
/* a = y/cbuflen;
if (a<=amin)
{
enqueue_pqueue(&Q, NULL, a, i+1, 0,0);
((i == 0) ? e1++ : e2++);
} */
}
fclose(fp);
fclose(ifp);
fclose(wfp);
return; /* PFAM!!! */
*non1p = 1.0*e1; /* /s1; */
*non2p = 1.0*e2; /* /s2; */
*non12p = 1.0*(e1+e2); /* /(s1+s2); */
*iso1p = *non1p;
*iso2p = *non2p;
*iso12p = *non12p;
iso3 = 0.0;
f1 = (*iso1p > iso3);
f2 = (*iso2p > iso3);
f12 = (*iso12p > iso3);
while (f1+f2+f12 > 0)
{
dequeue_pqueue(&Q, NULL, &a, &i, &dum1, &dum2);
switch(i)
{
case 1:
e1--;
if (f1)
*iso1p = 1.0*e1; /* /s1; */
if (f12)
*iso12p = 1.0*(e1+e2); /* /(s1+s2); */
break;
case 2:
e2--;
if (f2)
*iso2p = 1.0*e2; /* /s2; */
if (f12)
*iso12p = 1.0*(e1+e2); /* /(s1+s2); */
break;
case 3:
e3++;
iso3 = 1.0*e3; /* /s3; */
break;
}
f1 = (*iso1p > iso3);
f2 = (*iso2p > iso3);
f12 = (*iso12p > iso3);
}
free_pqueue(Q);
free(cbuf);
return;
}
/*
* Single source shortest path algorithm (Dijkstra's algorithm).
*/
static int shortest_path(double *nodeArray, int num_nodes, double *edgeArray, int num_edges,
int start_index, double dimdist[3], double radius, double *lastNodeList, int geoflag)
{
int i, cnt, num_nbhrs, curNodeIdx, curNhbrIdx, firstNodeOffset;
double *node, *new_node, *nbhrs, *nbhrdists, new_dist;
PQueue *PQ;
/* Initialize distance to HUGE_VAL. */
for (i=0, node=&nodeArray[DIST]; i<num_nodes;
i++, node+=NUM_ATTRS) *node = HUGE_VAL;
/* Allocate priority queue and insert start_index as first node. */
PQ = make_pqueue(num_nodes);
node = NODEN(nodeArray,start_index-1);
node[DIST] = 0.0;
if (lastNodeList != NULL) {
lastNodeList[start_index-1] = 0;
}
pqueue_insert(PQ, (PQueueNode)node);
firstNodeOffset = ((int)NODEN(nodeArray,0));
/* Dijkstra's algorithm. */
cnt = 0;
while (!pqueue_empty_p(PQ)) {
node = (double *) pqueue_extract_min(PQ);
curNodeIdx = (int)(((int)(node) - firstNodeOffset) / (8*NUM_ATTRS)) + 1;
node[DIST] = -node[DIST]; /* computed */
cnt++;
/* Relax all the neighboring nodes. */
num_nbhrs = (int) node[NUM_NBHRS];
/* Get the geodesic distances to the neighbors. */
if (geoflag == 1) {
nbhrs = EDGEN(edgeArray,(((int)node[NBHRS])-1));
nbhrdists = EDGEN(edgeArray,(((int)node[NBHRS])-1)) + 1;
} else {
nbhrs = &edgeArray[((int)node[NBHRS])-1];
}
for (i=0; i<num_nbhrs; i++) {
if (geoflag == 1) {
curNhbrIdx = ((int)nbhrs[2*i]);
new_node = NODEN(nodeArray,((int)nbhrs[2*i])-1);
} else {
curNhbrIdx = (((int)nbhrs[i]));
new_node = NODEN(nodeArray,((int)nbhrs[i])-1);
}
if (new_node[DIST]>=0) { /* not computed yet */
/* node[DIST] has been negated a couple of lines above.
* so, we need to negate it again to get the actual
* (positive) distance.
* -node[DIST] is the distance from our working node
* to the start point. node_dist() is the distance
* between our working node and its ith neighbor. */
/* 08.05.98 SJC
* If geoflag == 1, use the geodesic edge distance. */
if (geoflag == 1) {
new_dist = -node[DIST] + nbhrdists[2*i];
} else {
new_dist = -node[DIST] + node_dist(node, new_node, dimdist);
}
if (new_dist<radius) {
/* If this is the first time that we've gotten here, we
* use this new distance and add the node into the
* priority queue. */
if (new_node[DIST]==HUGE_VAL) {
new_node[DIST] = new_dist;
if (lastNodeList != NULL) {
lastNodeList[curNhbrIdx - 1] = curNodeIdx;
}
pqueue_insert(PQ, (PQueueNode)new_node);
}
/* Otherwise, if the new distance is less than
* the previously computed distance, then we pick
* the new (shorter) distance and adjust its
* position in the priority queue. */
else {
if (new_dist<new_node[DIST]) {
new_node[DIST] = new_dist;
if (lastNodeList != NULL) {
lastNodeList[curNhbrIdx - 1] = curNodeIdx;
}
pqueue_deckey(PQ, (PQueueNode)new_node);
}
}
}
}
}
}
free_pqueue(PQ);
return(cnt);
}
