/*Takes in as input the vector of iterations from coarse BLAST, an index into
*the vector (representing the index of which query we want to get the results
*from), and the database we are using for search and returns a vector of
*every original sequence section re-created from the calls to
*cb_coarse_expand for the hits in the iteration we are expanding.
*/
struct DSVector *expand_blast_hits(struct DSVector *iterations, int index,
struct cb_database_r *db){
struct DSVector *expanded_hits = ds_vector_create(),
*hits = get_blast_hits(
(xmlNode *)ds_vector_get(iterations, index));
int i = 0, j = 0, k = 0;
for (i = 0; i < hits->size; i++) {
struct hit *current_hit = (struct hit *)ds_vector_get(hits, i);
struct DSVector *hsps = current_hit->hsps;
for (j = 0; j < hsps->size; j++) {
struct DSVector *oseqs;
struct hsp *h = (struct hsp *)ds_vector_get(hsps, j);
int32_t coarse_start = h->hit_from-1, coarse_end = h->hit_to-1,
coarse_seq_id = current_hit->accession;
oseqs = cb_coarse_expand(db->coarse_db, db->com_db, coarse_seq_id,
coarse_start, coarse_end, 50);
for (k = 0; k < oseqs->size; k++)
ds_vector_append(expanded_hits, ds_vector_get(oseqs, k));
ds_vector_free_no_data(oseqs);
}
ds_vector_free(current_hit->hsps);
free(current_hit);
}
ds_vector_free_no_data(hits);
return expanded_hits;
}
static zval *php_ds_vector_read_dimension(zval *obj, zval *offset, int type, zval *return_value)
{
ds_vector_t *vector = Z_DS_VECTOR_P(obj);
zval *value;
// Dereference the offset if it's a reference.
ZVAL_DEREF(offset);
// `??`
if (type == BP_VAR_IS) {
if (Z_TYPE_P(offset) != IS_LONG || ! ds_vector_isset(vector, Z_LVAL_P(offset), 0)) {
return &EG(uninitialized_zval);
}
}
// Enforce strict integer index.
if (Z_TYPE_P(offset) != IS_LONG) {
INTEGER_INDEX_REQUIRED(offset);
return NULL;
}
// Access the value at the given index.
value = ds_vector_get(vector, Z_LVAL_P(offset));
// If we're accessing by reference we have to create a reference.
// This is for access like $deque[$a][$b] = $c
if (value && type != BP_VAR_R) {
ZVAL_MAKE_REF(value);
}
return value;
}
/*Takes in a vector of expansion structs for the expanded BLAST hits from a
query and outputs them to the FASTA file CaBLAST_fine.fasta.*/
void write_fine_fasta(struct DSVector *oseqs){
FILE *temp;
int i;
if (NULL == (temp = fopen("CaBLAST_fine.fasta", "w"))) {
fprintf(stderr, "fopen: 'fopen %s' failed: %s\n",
"CaBLAST_fine.fasta", strerror(errno));
exit(1);
}
if (!temp) {
fprintf(stderr, "Could not open CaBLAST_fine.fasta for writing\n");
return;
}
for (i = 0; i < oseqs->size; i++) {
struct cb_seq *current_seq =
((struct cb_hit_expansion *)ds_vector_get(oseqs, i))->seq;
fprintf(temp, "> %s\n%s\n", current_seq->name, current_seq->residues);
}
fclose(temp);
if (search_flags.fine_blast_db)
system("makeblastdb -dbtype nucl -in CaBLAST_fine.fasta -out "
"CaBLAST_fine.fasta");
}
/* tries to register current frame onto previous frame.
* Algorithm:
* discards fields with low contrast
* select maxfields fields according to their contrast
* check theses fields for vertical and horizontal transformation
* use minimal difference of all possible positions
* calculate shift as cleaned mean of all remaining fields
* calculate rotation angle of each field in respect to center of fields
* after shift removal
* calculate rotation angle as cleaned mean of all angles
* compensate for possibly off-center rotation
*/
Transform calcTransFields(MotionDetect* md, calcFieldTransFunc fieldfunc,
contrastSubImgFunc contrastfunc) {
Transform* ts = (Transform*) ds_malloc(sizeof(Transform) * md->fieldNum);
Field** fs = (Field**) ds_malloc(sizeof(Field*) * md->fieldNum);
double *angles = (double*) ds_malloc(sizeof(double) * md->fieldNum);
int i, index = 0, num_trans;
Transform t;
#ifdef STABVERBOSE
FILE *file = NULL;
char buffer[32];
ds_snprintf(buffer, sizeof(buffer), "k%04i.dat", md->frameNum);
file = fopen(buffer, "w");
fprintf(file, "# plot \"%s\" w l, \"\" every 2:1:0\n", buffer);
#endif
DSVector goodflds = selectfields(md, contrastfunc);
// use all "good" fields and calculate optimal match to previous frame
#ifdef USE_OMP
#pragma omp parallel for shared(goodflds, md, ts, fs) // does not bring speedup
#endif
for(index=0; index < ds_vector_size(&goodflds); index++){
int i = ((contrast_idx*)ds_vector_get(&goodflds,index))->index;
t = fieldfunc(md, &md->fields[i], i); // e.g. calcFieldTransYUV
#ifdef STABVERBOSE
fprintf(file, "%i %i\n%f %f %i\n \n\n", md->fields[i].x, md->fields[i].y,
md->fields[i].x + t.x, md->fields[i].y + t.y, t.extra);
#endif
if (t.extra != -1) { // ignore if extra == -1 (unused at the moment)
ts[index] = t;
fs[index] = md->fields + i;
}
}
t = null_transform();
num_trans = ds_vector_size(&goodflds); // amount of transforms we actually have
ds_vector_del(&goodflds);
if (num_trans < 1) {
ds_log_warn(md->modName, "too low contrast! No field remains.\n"
"(no translations are detected in frame %i)", md->frameNum);
return t;
}
int center_x = 0;
int center_y = 0;
// calc center point of all remaining fields
for (i = 0; i < num_trans; i++) {
center_x += fs[i]->x;
center_y += fs[i]->y;
}
center_x /= num_trans;
center_y /= num_trans;
if (md->show) { // draw fields and transforms into frame.
// this has to be done one after another to handle possible overlap
if (md->show > 1) {
for (i = 0; i < num_trans; i++)
drawFieldScanArea(md, fs[i], &ts[i]);
}
for (i = 0; i < num_trans; i++)
drawField(md, fs[i], &ts[i]);
for (i = 0; i < num_trans; i++)
drawFieldTrans(md, fs[i], &ts[i]);
}
/* median over all transforms
t= median_xy_transform(ts, md->field_num);*/
// cleaned mean
t = cleanmean_xy_transform(ts, num_trans);
// substract avg
for (i = 0; i < num_trans; i++) {
ts[i] = sub_transforms(&ts[i], &t);
}
// figure out angle
if (md->fieldNum < 6) {
// the angle calculation is inaccurate for 5 and less fields
t.alpha = 0;
} else {
for (i = 0; i < num_trans; i++) {
angles[i] = calcAngle(md, fs[i], &ts[i], center_x, center_y);
}
double min, max;
t.alpha = -cleanmean(angles, num_trans, &min, &max);
if (max - min > md->maxAngleVariation) {
t.alpha = 0;
ds_log_info(md->modName, "too large variation in angle(%f)\n",
max-min);
}
}
// compensate for off-center rotation
double p_x = (center_x - md->fi.width / 2);
double p_y = (center_y - md->fi.height / 2);
t.x += (cos(t.alpha) - 1) * p_x - sin(t.alpha) * p_y;
t.y += sin(t.alpha) * p_x + (cos(t.alpha) - 1) * p_y;
#ifdef STABVERBOSE
fclose(file);
#endif
return t;
}
int main(int argc, char **argv){
FILE *query_file = NULL, *test_hits_file = NULL;
struct cb_database_r *db = NULL;
struct opt_config *conf;
struct opt_args *args;
struct DSVector *iterations = NULL, *expanded_hits = NULL, *queries = NULL,
*oseqs = ds_vector_create();
struct fasta_seq *query = NULL;
xmlDoc *doc = NULL;
xmlNode *root = NULL;
uint64_t dbsize = 0;
int i = 0, j = 0;
conf = load_search_args();
args = opt_config_parse(conf, argc, argv);
if (args->nargs < 2) {
fprintf(stderr,
"Usage: %s [flags] database-dir fasta-file "
"[ --blast_args BLASTN_ARGUMENTS ]\n", argv[0]);
opt_config_print_usage(conf);
exit(1);
}
system("rm CaBLAST_results.xml");
if (!search_flags.hide_progress)
fprintf(stderr, "Loading database data\n\n");
db = cb_database_r_init(args->args[0],
(search_flags.load_coarse_db ||
search_flags.load_coarse_residues),
(search_flags.load_coarse_db ||
search_flags.load_coarse_links),
search_flags.load_compressed_db,
search_flags.link_block_size);
dbsize = read_int_from_file(8, db->coarse_db->db->file_params);
if (!search_flags.hide_progress)
fprintf(stderr, "Running coarse BLAST\n\n");
blast_coarse(args, dbsize);
if (NULL == (query_file = fopen(args->args[1], "r"))) {
fprintf(stderr, "fopen: 'fopen %s' failed: %s\n",
args->args[1], strerror(errno));
exit(1);
}
queries = ds_vector_create();
query = fasta_read_next(query_file, "");
while (query) {
ds_vector_append(queries, (void *)query);
query = fasta_read_next(query_file, "");
}
fclose(query_file);
if (!search_flags.hide_progress)
fprintf(stderr, "Processing coarse BLAST hits for fine BLAST\n\n");
if (search_flags.show_hit_info)
if (NULL == (test_hits_file = fopen("CaBLAST_hits.txt", "w"))) {
fprintf(stderr, "fopen: 'fopen %s' failed: %s\n",
"CaBLAST_hits.txt", strerror(errno));
exit(1);
}
//Parse the XML file generated from coarse BLAST and get its iterations.
doc = xmlReadFile("CaBLAST_temp_blast_results.xml", NULL, 0);
if (doc == NULL) {
fprintf(stderr, "Could not parse CaBLAST_temp_blast_results.xml\n");
return 0;
}
root = xmlDocGetRootElement(doc);
iterations = get_blast_iterations(root);
for (i = 0; i < iterations->size; i++) {
if (!search_flags.hide_progress) {
int32_t digits_full = floor(log10((double)iterations->size)),
digits_i = floor(log10((double)i)), spaces;
char *bar = progress_bar(i, iterations->size);
spaces = digits_full - digits_i;
fprintf(stderr, "\r");
fprintf(stderr, "iteration: %d/%d", i+1, iterations->size);
for (j = 0; j < spaces; j++)
putc(' ', stderr);
fprintf(stderr, " %s ", bar);
free(bar);
}
/*Expand any BLAST hits we got from the current query sequence during
coarse BLAST.*/
expanded_hits = expand_blast_hits(iterations, i, db);
for (j = 0; j < expanded_hits->size; j++)
ds_vector_append(oseqs, ds_vector_get(expanded_hits, j));
ds_vector_free_no_data(expanded_hits);
}
write_fine_fasta(oseqs);
for (i = 0; i < oseqs->size; i++)
cb_hit_expansion_free(
(struct cb_hit_expansion *)ds_vector_get(oseqs, i));
ds_vector_free_no_data(oseqs);
blast_fine(args, dbsize);
if (!search_flags.hide_progress)
fprintf(stderr, "\n"); //Make a newline after the progress bar
for (i = 0; i < queries->size; i++)
fasta_free_seq((struct fasta_seq *)ds_vector_get(queries, i));
ds_vector_free_no_data(queries);
if (search_flags.show_hit_info)
fclose(test_hits_file);
//Free the XML data and expanded hits
for (i = 0; i < iterations->size; i++) {
struct DSVector *iteration =
(struct DSVector *)ds_vector_get(iterations, i);
for (j = 0; j < iteration->size; j++) {
struct hit *h = (struct hit *)ds_vector_get(iteration, j);
ds_vector_free(h->hsps);
free(h);
}
}
ds_vector_free_no_data(iterations);
cb_database_r_free(db);
xmlFreeDoc(doc);
/*Free the coarse BLAST results file if the --no-cleanup flag is not being
used.*/
if (!search_flags.no_cleanup)
system("rm CaBLAST_temp_blast_results.xml");
opt_args_free(args);
opt_config_free(conf);
return 0;
}
