gk_i2cc2i_t *gk_i2cc2i_create_common(char *alphabet)
{
int nsymbols;
gk_idx_t i;
gk_i2cc2i_t *t;
nsymbols = strlen(alphabet);
t = gk_malloc(sizeof(gk_i2cc2i_t),"gk_i2c_create_common");
t->n = nsymbols;
t->i2c = gk_cmalloc(256, "gk_i2c_create_common");
t->c2i = gk_imalloc(256, "gk_i2c_create_common");
gk_cset(256, -1, t->i2c);
gk_iset(256, -1, t->c2i);
for(i=0;i<nsymbols;i++){
t->i2c[i] = alphabet[i];
t->c2i[(int)alphabet[i]] = i;
}
return t;
}
gk_graph_t *gk_graph_ExtractPartition(gk_graph_t *graph, int *part, int pid)
{
ssize_t i, j, nnz;
gk_graph_t *ngraph;
ngraph = gk_graph_Create();
ngraph->nrows = 0;
ngraph->ncols = graph->ncols;
for (nnz=0, i=0; i<graph->nrows; i++) {
if (part[i] == pid) {
ngraph->nrows++;
nnz += graph->rowptr[i+1]-graph->rowptr[i];
}
}
ngraph->rowptr = gk_zmalloc(ngraph->nrows+1, "gk_graph_ExtractPartition: rowptr");
ngraph->rowind = gk_imalloc(nnz, "gk_graph_ExtractPartition: rowind");
ngraph->rowval = gk_fmalloc(nnz, "gk_graph_ExtractPartition: rowval");
ngraph->rowptr[0] = 0;
for (nnz=0, j=0, i=0; i<graph->nrows; i++) {
if (part[i] == pid) {
gk_icopy(graph->rowptr[i+1]-graph->rowptr[i], graph->rowind+graph->rowptr[i], ngraph->rowind+nnz);
gk_fcopy(graph->rowptr[i+1]-graph->rowptr[i], graph->rowval+graph->rowptr[i], ngraph->rowval+nnz);
nnz += graph->rowptr[i+1]-graph->rowptr[i];
ngraph->rowptr[++j] = nnz;
}
}
ASSERT(j == ngraph->nrows);
return ngraph;
}
gk_graph_t *gk_graph_ExtractRows(gk_graph_t *graph, int nrows, int *rind)
{
ssize_t i, ii, j, nnz;
gk_graph_t *ngraph;
ngraph = gk_graph_Create();
ngraph->nrows = nrows;
ngraph->ncols = graph->ncols;
for (nnz=0, i=0; i<nrows; i++)
nnz += graph->rowptr[rind[i]+1]-graph->rowptr[rind[i]];
ngraph->rowptr = gk_zmalloc(ngraph->nrows+1, "gk_graph_ExtractPartition: rowptr");
ngraph->rowind = gk_imalloc(nnz, "gk_graph_ExtractPartition: rowind");
ngraph->rowval = gk_fmalloc(nnz, "gk_graph_ExtractPartition: rowval");
ngraph->rowptr[0] = 0;
for (nnz=0, j=0, ii=0; ii<nrows; ii++) {
i = rind[ii];
gk_icopy(graph->rowptr[i+1]-graph->rowptr[i], graph->rowind+graph->rowptr[i], ngraph->rowind+nnz);
gk_fcopy(graph->rowptr[i+1]-graph->rowptr[i], graph->rowval+graph->rowptr[i], ngraph->rowval+nnz);
nnz += graph->rowptr[i+1]-graph->rowptr[i];
ngraph->rowptr[++j] = nnz;
}
ASSERT(j == ngraph->nrows);
return ngraph;
}
void gk_find_frequent_itemsets(int ntrans, int *tranptr, int *tranind,
int minfreq, int maxfreq, int minlen, int maxlen,
void (*process_itemset)(void *stateptr, int nitems, int *itemids,
int ntrans, int *transids),
void *stateptr)
{
ssize_t i;
gk_csr_t *mat, *pmat;
isparams_t params;
int *pattern;
/* Create the matrix */
mat = gk_csr_Create();
mat->nrows = ntrans;
mat->ncols = tranind[gk_iargmax(tranptr[ntrans], tranind)]+1;
mat->rowptr = gk_zmalloc(ntrans+1, "gk_find_frequent_itemsets: mat.rowptr");
for (i=0; i<ntrans+1; i++)
mat->rowptr[i] = tranptr[i];
mat->rowind = gk_icopy(tranptr[ntrans], tranind, gk_imalloc(tranptr[ntrans], "gk_find_frequent_itemsets: mat.rowind"));
mat->colids = gk_iincset(mat->ncols, 0, gk_imalloc(mat->ncols, "gk_find_frequent_itemsets: mat.colids"));
/* Setup the parameters */
params.minfreq = minfreq;
params.maxfreq = (maxfreq == -1 ? mat->nrows : maxfreq);
params.minlen = minlen;
params.maxlen = (maxlen == -1 ? mat->ncols : maxlen);
params.tnitems = mat->ncols;
params.callback = process_itemset;
params.stateptr = stateptr;
params.rmarker = gk_ismalloc(mat->nrows, 0, "gk_find_frequent_itemsets: rmarker");
params.cand = gk_ikvmalloc(mat->ncols, "gk_find_frequent_itemsets: cand");
/* Perform the initial projection */
gk_csr_CreateIndex(mat, GK_CSR_COL);
pmat = itemsets_project_matrix(¶ms, mat, -1);
gk_csr_Free(&mat);
pattern = gk_imalloc(pmat->ncols, "gk_find_frequent_itemsets: pattern");
itemsets_find_frequent_itemsets(¶ms, pmat, 0, pattern);
gk_csr_Free(&pmat);
gk_free((void **)&pattern, ¶ms.rmarker, ¶ms.cand, LTERM);
}
gk_seq_t *gk_seq_ReadGKMODPSSM(char *filename)
{
gk_seq_t *seq;
gk_idx_t i, j, ii;
size_t ntokens, nbytes, len;
FILE *fpin;
gk_Tokens_t tokens;
static char *AAORDER = "ARNDCQEGHILKMFPSTWYVBZX*";
static int PSSMWIDTH = 20;
char *header, line[MAXLINELEN];
gk_i2cc2i_t *converter;
header = gk_cmalloc(PSSMWIDTH, "gk_seq_ReadGKMODPSSM: header");
converter = gk_i2cc2i_create_common(AAORDER);
gk_getfilestats(filename, &len, &ntokens, NULL, &nbytes);
len --;
seq = gk_malloc(sizeof(gk_seq_t),"gk_seq_ReadGKMODPSSM");
gk_seq_init(seq);
seq->len = len;
seq->sequence = gk_imalloc(len, "gk_seq_ReadGKMODPSSM");
seq->pssm = gk_iAllocMatrix(len, PSSMWIDTH, 0, "gk_seq_ReadGKMODPSSM");
seq->psfm = gk_iAllocMatrix(len, PSSMWIDTH, 0, "gk_seq_ReadGKMODPSSM");
seq->nsymbols = PSSMWIDTH;
seq->name = gk_getbasename(filename);
fpin = gk_fopen(filename,"r","gk_seq_ReadGKMODPSSM");
/* Read the header line */
if (fgets(line, MAXLINELEN-1, fpin) == NULL)
errexit("Unexpected end of file: %s\n", filename);
gk_strtoupper(line);
gk_strtokenize(line, " \t\n", &tokens);
for (i=0; i<PSSMWIDTH; i++)
header[i] = tokens.list[i][0];
gk_freetokenslist(&tokens);
/* Read the rest of the lines */
for (i=0, ii=0; ii<len; ii++) {
if (fgets(line, MAXLINELEN-1, fpin) == NULL)
errexit("Unexpected end of file: %s\n", filename);
gk_strtoupper(line);
gk_strtokenize(line, " \t\n", &tokens);
seq->sequence[i] = converter->c2i[(int)tokens.list[1][0]];
for (j=0; j<PSSMWIDTH; j++) {
seq->pssm[i][converter->c2i[(int)header[j]]] = atoi(tokens.list[2+j]);
seq->psfm[i][converter->c2i[(int)header[j]]] = atoi(tokens.list[2+PSSMWIDTH+j]);
}
gk_freetokenslist(&tokens);
i++;
}
seq->len = i; /* Reset the length if certain characters were skipped */
gk_free((void **)&header, LTERM);
gk_fclose(fpin);
return seq;
}
gk_csr_t *itemsets_project_matrix(isparams_t *params, gk_csr_t *mat, int cid)
{
ssize_t i, j, k, ii, pnnz;
int nrows, ncols, pnrows, pncols;
ssize_t *colptr, *pcolptr;
int *colind, *colids, *pcolind, *pcolids, *rmarker;
gk_csr_t *pmat;
gk_ikv_t *cand;
nrows = mat->nrows;
ncols = mat->ncols;
colptr = mat->colptr;
colind = mat->colind;
colids = mat->colids;
rmarker = params->rmarker;
cand = params->cand;
/* Allocate space for the projected matrix based on what you know thus far */
pmat = gk_csr_Create();
pmat->nrows = pnrows = (cid == -1 ? nrows : colptr[cid+1]-colptr[cid]);
/* Mark the rows that will be kept and determine the prowids */
if (cid == -1) { /* Initial projection */
gk_iset(nrows, 1, rmarker);
}
else { /* The other projections */
for (i=colptr[cid]; i<colptr[cid+1]; i++)
rmarker[colind[i]] = 1;
}
/* Determine the length of each column that will be left in the projected matrix */
for (pncols=0, pnnz=0, i=cid+1; i<ncols; i++) {
for (k=0, j=colptr[i]; j<colptr[i+1]; j++) {
k += rmarker[colind[j]];
}
if (k >= params->minfreq && k <= params->maxfreq) {
cand[pncols].val = i;
cand[pncols++].key = k;
pnnz += k;
}
}
/* Sort the columns in increasing order */
gk_ikvsorti(pncols, cand);
/* Allocate space for the remaining fields of the projected matrix */
pmat->ncols = pncols;
pmat->colids = pcolids = gk_imalloc(pncols, "itemsets_project_matrix: pcolids");
pmat->colptr = pcolptr = gk_zmalloc(pncols+1, "itemsets_project_matrix: pcolptr");
pmat->colind = pcolind = gk_imalloc(pnnz, "itemsets_project_matrix: pcolind");
/* Populate the projected matrix */
pcolptr[0] = 0;
for (pnnz=0, ii=0; ii<pncols; ii++) {
i = cand[ii].val;
for (j=colptr[i]; j<colptr[i+1]; j++) {
if (rmarker[colind[j]])
pcolind[pnnz++] = colind[j];
}
pcolids[ii] = colids[i];
pcolptr[ii+1] = pnnz;
}
/* Reset the rmarker array */
if (cid == -1) { /* Initial projection */
gk_iset(nrows, 0, rmarker);
}
else { /* The other projections */
for (i=colptr[cid]; i<colptr[cid+1]; i++)
rmarker[colind[i]] = 0;
}
return pmat;
}
gk_graph_t *gk_graph_Prune(gk_graph_t *graph, int what, int minf, int maxf)
{
ssize_t i, j, nnz;
int nrows, ncols;
ssize_t *rowptr, *nrowptr;
int *rowind, *nrowind, *collen;
float *rowval, *nrowval;
gk_graph_t *ngraph;
ngraph = gk_graph_Create();
nrows = ngraph->nrows = graph->nrows;
ncols = ngraph->ncols = graph->ncols;
rowptr = graph->rowptr;
rowind = graph->rowind;
rowval = graph->rowval;
nrowptr = ngraph->rowptr = gk_zmalloc(nrows+1, "gk_graph_Prune: nrowptr");
nrowind = ngraph->rowind = gk_imalloc(rowptr[nrows], "gk_graph_Prune: nrowind");
nrowval = ngraph->rowval = gk_fmalloc(rowptr[nrows], "gk_graph_Prune: nrowval");
switch (what) {
case GK_CSR_COL:
collen = gk_ismalloc(ncols, 0, "gk_graph_Prune: collen");
for (i=0; i<nrows; i++) {
for (j=rowptr[i]; j<rowptr[i+1]; j++) {
ASSERT(rowind[j] < ncols);
collen[rowind[j]]++;
}
}
for (i=0; i<ncols; i++)
collen[i] = (collen[i] >= minf && collen[i] <= maxf ? 1 : 0);
nrowptr[0] = 0;
for (nnz=0, i=0; i<nrows; i++) {
for (j=rowptr[i]; j<rowptr[i+1]; j++) {
if (collen[rowind[j]]) {
nrowind[nnz] = rowind[j];
nrowval[nnz] = rowval[j];
nnz++;
}
}
nrowptr[i+1] = nnz;
}
gk_free((void **)&collen, LTERM);
break;
case GK_CSR_ROW:
nrowptr[0] = 0;
for (nnz=0, i=0; i<nrows; i++) {
if (rowptr[i+1]-rowptr[i] >= minf && rowptr[i+1]-rowptr[i] <= maxf) {
for (j=rowptr[i]; j<rowptr[i+1]; j++, nnz++) {
nrowind[nnz] = rowind[j];
nrowval[nnz] = rowval[j];
}
}
nrowptr[i+1] = nnz;
}
break;
default:
gk_graph_Free(&ngraph);
gk_errexit(SIGERR, "Unknown prunning type of %d\n", what);
return NULL;
}
return ngraph;
}
