/*
* Compute the visible end position of a contig. This isn't just the extents
* of start_used / end_used in the bins as this can included invisible
* data such as cached consensus sequences.
*/
int contig_visible_end(GapIO *io, tg_rec crec) {
rangec_t *r;
contig_iterator *ci;
ci = contig_iter_new_by_type(io, crec, 1, CITER_LAST | CITER_IEND,
CITER_CSTART, CITER_CEND,
GRANGE_FLAG_ISANY);
if (!ci) {
contig_t *c = cache_search(io, GT_Contig, crec);
return c->end;
}
while (r = contig_iter_prev(io, ci)) {
int v;
if ((r->flags & GRANGE_FLAG_ISMASK) == GRANGE_FLAG_ISCONS)
continue;
v = r->end;
contig_iter_del(ci);
return v;
}
contig_iter_del(ci);
return 0;
}
int edview_search_tag_type(edview *xx, int dir, int strand, char *value) {
contig_iterator *iter;
int start, end;
rangec_t *r;
rangec_t *(*ifunc)(GapIO *io, contig_iterator *ci);
int type = str2type(value);
contig_t *c = cache_search(xx->io, GT_Contig, xx->cnum);
if (dir) {
start = xx->cursor_apos + (dir ? 1 : -1);
end = c->end;
ifunc = contig_iter_next;
} else {
start = c->start;
end = xx->cursor_apos + (dir ? 1 : -1);
ifunc = contig_iter_prev;
}
iter = contig_iter_new_by_type(xx->io, xx->cnum, 1,
dir == 1 ? CITER_FIRST : CITER_LAST,
start, end, GRANGE_FLAG_ISANNO);
if (!iter)
/* Can happen legitimately when we're already at the end of contig */
return -1;
while (r = ifunc(xx->io, iter)) {
if ((dir && r->start < start) ||
(!dir && r->start > end))
continue;
if (r->mqual == type)
break;
}
if (r) {
if (r->flags & GRANGE_FLAG_TAG_SEQ) {
int pos;
sequence_get_position(xx->io, r->pair_rec, NULL, &pos, NULL, NULL);
pos = r->start - pos;
edSetCursorPos(xx, GT_Seq, r->pair_rec, pos, 1);
} else {
edSetCursorPos(xx, GT_Contig, xx->cnum, r->start, 1);
}
contig_iter_del(iter);
return 0;
}
contig_iter_del(iter);
return -1;
}
static void tag_shift_for_delete(GapIO *io, tg_rec crec, tg_rec srec,
int start, int end, int pos, tg_rec brec) {
contig_iterator *ci;
rangec_t *r;
contig_t *c = cache_search(io, GT_Contig, crec);;
//printf("< tag in seq %"PRIrec" at %d\n", srec, pos);
cache_incr(io, c);
ci = contig_iter_new_by_type(io, crec, 0, CITER_FIRST | CITER_ISTART,
start+pos, end, GRANGE_FLAG_ISANNO);
if (!ci) {
cache_decr(io, c);
return;
}
while ((r = contig_iter_next(io, ci))) {
range_t r2, *r_out;
anno_ele_t *a;
bin_index_t *bin;
if (r->pair_rec != srec)
continue;
bin_remove_item(io, &c, GT_AnnoEle, r->rec);
r2.start = (r->start > start+pos) ? r->start-1 : r->start;
r2.end = r->end-1;
r2.mqual = r->mqual;
r2.rec = r->rec;
r2.pair_rec = r->pair_rec;
r2.flags = r->flags;
if (r2.end < r2.start) {
/* Tag entirely removed now, it must have been on a pad */
a = cache_search(io, GT_AnnoEle, r->rec);
a = cache_rw(io, a);
cache_deallocate(io, a);
continue;
}
bin = bin_add_to_range(io, &c, brec, &r2, &r_out, NULL, 0);
a = cache_search(io, GT_AnnoEle, r->rec);
if (a->bin != bin->rec /*||
a->idx != r_out - ArrayBase(range_t, bin->rng)*/) {
/* Annotation moved bins */
a = cache_rw(io, a);
a->bin = bin->rec;
//a->bin_idx = r_out - ArrayBase(range_t, bin->rng);
}
}
cache_decr(io, c);
contig_iter_del(ci);
}
int edview_search_tag_indel(edview *xx, int dir, int strand, char *value) {
contig_iterator *iter;
int start, end;
rangec_t *r;
rangec_t *(*ifunc)(GapIO *io, contig_iterator *ci);
contig_t *c = cache_search(xx->io, GT_Contig, xx->cnum);
if (dir) {
start = xx->cursor_apos + (dir ? 1 : -1);
end = c->end;
ifunc = contig_iter_next;
} else {
start = c->start;
end = xx->cursor_apos + (dir ? 1 : -1);
ifunc = contig_iter_prev;
}
iter = contig_iter_new_by_type(xx->io, xx->cnum, 1,
dir == 1 ? CITER_FIRST : CITER_LAST,
start, end, GRANGE_FLAG_ISREFPOS);
if (!iter)
return -1;
while (r = ifunc(xx->io, iter)) {
if ((dir && r->start < start) ||
(!dir && r->start > end))
continue;
break;
}
if (r) {
edSetCursorPos(xx, GT_Contig, xx->cnum, r->start, 1);
contig_iter_del(iter);
return 0;
}
contig_iter_del(iter);
return -1;
}
/*
* Removes all tags of specific types (hashed in h, or all if h == NULL)
* from a specified contig.
*
* Returns 0 on success
* -1 on failure
*/
static int delete_tag_single_contig(GapIO *io, tg_rec crec,
HashTable *h, int verbose) {
contig_iterator *ci;
rangec_t *r;
contig_t *c;
int ret = -1;
ci = contig_iter_new_by_type(io, crec, 1, CITER_FIRST,
CITER_CSTART, CITER_CEND,
GRANGE_FLAG_ISANNO);
if (!ci)
return -1;
if (!(c = cache_search(io, GT_Contig, crec))) {
contig_iter_del(ci);
return -1;
}
cache_incr(io, c);
while (NULL != (r = contig_iter_next(io, ci))) {
char t[5];
(void)type2str(r->mqual, t);
if (!h || HashTableSearch(h, t, 4)) {
anno_ele_t *e;
if (verbose)
vmessage("Removing anno %s #%"PRIrec"\tContig %s\t%d..%d\n",
t, r->rec, c->name, r->start, r->end);
if (bin_remove_item(io, &c, GT_AnnoEle, r->rec)) goto fail;
/* FIXME: Need to reclaim the GT_AnnoEle record itself */
}
}
ret = 0;
fail:
contig_iter_del(ci);
cache_decr(io, c);
return ret;
}
static void test_mode3(GapIO *io, int cnum, int xpos) {
rangec_t *r;
contig_iterator *ci;
ci = contig_iter_new(io, cnum, 0, CITER_FIRST, CITER_CSTART, CITER_CEND);
while (r = contig_iter_next(io, ci)) {
seq_t *s = get_seq(io, r->rec);
char name[256];
sprintf(name, "%.*s", s->name_len, s->name);
printf("%c%-22s\t%8d..%-8d\t%.*s\n",
"+-"[s->len<0], name, r->start, r->end, ABS(s->len), s->seq);
}
contig_iter_del(ci);
exit(0);
}
/*
* Attempt to find edits. It's not 100% reliable, but works for most cases.
* We look for lowercase bases and confidence 100 and 0 (if not N).
* We cannot find deleted bases though.
*/
int edview_search_edit(edview *xx, int dir, int strand, char *value) {
int start, end;
contig_iterator *iter;
rangec_t *(*ifunc)(GapIO *io, contig_iterator *ci);
rangec_t *r;
int best_pos, found = 0;
int fpos;
tg_rec fseq;
if (dir) {
start = xx->cursor_apos + 1;
end = CITER_CEND;
iter = contig_iter_new(xx->io, xx->cnum, 1,
CITER_FIRST | CITER_ISTART,
start, end);
ifunc = contig_iter_next;
best_pos = INT_MAX;
} else {
start = CITER_CSTART;
end = xx->cursor_apos -1;
iter = contig_iter_new(xx->io, xx->cnum, 1,
CITER_LAST | CITER_IEND,
start, end);
ifunc = contig_iter_prev;
best_pos = INT_MIN;
}
if (!iter)
return -1;
while ((r = ifunc(xx->io, iter))) {
seq_t *s, *sorig;
char *seq, *qual;
int seq_len, comp, off = 0, i;
if (found && dir && r->start > best_pos)
break;
if (found && !dir && r->end < best_pos)
break;
if (NULL == (s = sorig = cache_search(xx->io, GT_Seq, r->rec)))
break;
if (r->comp ^ (s->len < 0)) {
s = dup_seq(s);
complement_seq_t(s);
}
seq = s->seq;
qual = s->conf;
seq_len = ABS(s->len);
if (r->start < start) {
off = start - r->start;
seq += off;
qual += off;
seq_len -= off;
}
for (i = 0; i < seq_len; i++) {
if (islower(seq[i]) ||
qual[i] == 100 ||
(qual[i] == 0 && seq[i] != 'N' && seq[i] != '-'
&& seq[i] != '*')) {
int pos = r->start + i + off;
if (dir) {
if (best_pos > pos && pos > xx->cursor_apos) {
found = 1;
best_pos = pos;
fpos = i + off;
fseq = r->rec;
}
break;
} else {
if (best_pos < pos && pos < xx->cursor_apos) {
found = 1;
best_pos = pos;
fpos = i + off;
fseq = r->rec;
}
}
}
}
if (s != sorig)
free(s);
}
if (found) {
edSetCursorPos(xx, fseq == xx->cnum ? GT_Contig : GT_Seq,
fseq, fpos, 1);
}
contig_iter_del(iter);
return found ? 0 : -1;
}
int edview_search_tag_anno(edview *xx, int dir, int strand, char *value) {
contig_iterator *iter;
int start, end;
rangec_t *r;
rangec_t *(*ifunc)(GapIO *io, contig_iterator *ci);
char *r_exp = NULL;
contig_t *c = cache_search(xx->io, GT_Contig, xx->cnum);
if (value) {
if (NULL == (r_exp = REGCMP(xx->interp, value))) {
verror(ERR_WARN, "Search by anno", "invalid regular expression");
return -1;
}
}
if (dir) {
start = xx->cursor_apos + (dir ? 1 : -1);
end = c->end;
ifunc = contig_iter_next;
} else {
start = c->start;
end = xx->cursor_apos + (dir ? 1 : -1);
ifunc = contig_iter_prev;
}
iter = contig_iter_new_by_type(xx->io, xx->cnum, 1,
dir == 1 ? CITER_FIRST : CITER_LAST,
start, end, GRANGE_FLAG_ISANNO);
if (!iter)
return -1;
while (r = ifunc(xx->io, iter)) {
anno_ele_t *ae;
if ((dir && r->start < start) ||
(!dir && r->start > end))
continue;
if (!r_exp)
break; /* blank expr => match all */
ae = cache_search(xx->io, GT_AnnoEle, r->rec);
if (!ae->comment)
continue;
if (REGEX(xx->interp, ae->comment, r_exp))
break;
}
REGFREE(xx->interp, r_exp);
if (r) {
if (r->flags & GRANGE_FLAG_TAG_SEQ) {
int pos;
sequence_get_position(xx->io, r->pair_rec, NULL, &pos, NULL, NULL);
pos = r->start - pos;
edSetCursorPos(xx, GT_Seq, r->pair_rec, pos, 1);
} else {
edSetCursorPos(xx, GT_Contig, xx->cnum, r->start, 1);
}
contig_iter_del(iter);
return 0;
}
contig_iter_del(iter);
return -1;
}
int edview_search_name(edview *xx, int dir, int strand, char *value)
{
tg_rec rec, *rp, cnum = -1, best_rec;
int best_pos, best_off;
int nr, i;
rangec_t *(*ifunc)(GapIO *io, contig_iterator *ci);
int start, end, cstart;
contig_iterator *iter;
contig_t *c;
/* Check for #num where num is a sequence record in this contig */
if (*value == '#') {
char *endp;
int64_t v = strtol64(value+1, &endp, 10);
rec = v;
if (*endp == '\0' && cache_exists(xx->io, GT_Seq, rec)) {
sequence_get_clipped_position(xx->io, rec, &cnum,
&start, NULL, &cstart, NULL, NULL);
if (cnum == xx->cnum) {
edSetCursorPos(xx, GT_Seq, rec, cstart - start, 1);
return 0;
}
}
}
/* Find all hits matching this name */
rp = sequence_index_query_all(xx->io, value, 1, &nr);
/* Also get an position-based iterator */
c = cache_search(xx->io, GT_Contig, xx->cnum);
if (dir) {
start = xx->cursor_apos + 1;
end = c->end;
ifunc = contig_iter_next;
best_pos = end + 1;
best_off = 0;
} else {
start = c->start;
end = xx->cursor_apos - 1;
ifunc = contig_iter_prev;
best_pos = start - 1;
best_off = 0;
}
iter = contig_iter_new_by_type(xx->io, xx->cnum, 1,
dir == 1 ? CITER_FIRST : CITER_LAST,
start-1, end+1, GRANGE_FLAG_ISSEQ);
if (!iter)
return -1;
/*
* The iterator also finds overlapping objects, not just ones beyond this
* point. That's fine if we're on the consensus as we probably want to
* jump to the first seq-name overlapping this point.
*
* However if we're on a sequence already, we want the first one
* after or before that sequence. So we skip along iterator until we're
* at the current record.
*/
if (xx->cursor_type == GT_Seq) {
rangec_t *r;
while ((r = ifunc(xx->io, iter))) {
if (r->rec == xx->cursor_rec)
break;
}
}
/* Alternate between the by-name and by-position scan */
best_rec = -1;
for (i = 0; i < nr; i++) {
int start, end;
rangec_t *r;
/* From name index */
rec = rp[i++];
sequence_get_clipped_position(xx->io, rec, &cnum, &start, &end,
&cstart, NULL, NULL);
if (cnum == xx->cnum) {
if ((dir && best_pos > cstart && cstart > xx->cursor_apos) ||
(!dir && best_pos < cstart && cstart < xx->cursor_apos)) {
best_pos = cstart;
best_off = cstart - start;
best_rec = rec;
}
}
/* From iterator */
if ((r = ifunc(xx->io, iter))) {
seq_t *s;
if (NULL == (s = cache_search(xx->io, GT_Seq, r->rec))) {
/* No match */
best_rec = -1;
break;
}
if (strncmp(s->name, value, strlen(value)) == 0) {
/* prefix match */
puts("Found by pos iterator");
best_rec = r->rec;
break;
}
} else {
/* End of contig - bail out early */
best_rec = -1;
break;
}
}
contig_iter_del(iter);
if (rp)
free(rp);
if (best_rec != -1) {
edSetCursorPos(xx, GT_Seq, best_rec, best_off, 1);
return 0;
}
return -1;
}
int edview_search_sequence(edview *xx, int dir, int strand, char *value) {
int mismatches = 0; /* exact match */
int where = 2; /* consensus */
char *p;
int start, end;
int patlen;
char *uppert, *upperb;
int found = 0, at_end = 0;
tg_rec fseq;
int fpos, i, j;
contig_t *c;
contig_iterator *iter;
rangec_t *(*ifunc)(GapIO *io, contig_iterator *ci);
rangec_t *r;
int best_pos;
if (dir) {
start = xx->cursor_apos + 1;
end = CITER_CEND;
iter = contig_iter_new(xx->io, xx->cnum, 1,
CITER_FIRST | CITER_ISTART,
start, end);
ifunc = contig_iter_next;
best_pos = INT_MAX;
} else {
start = CITER_CSTART;
end = xx->cursor_apos -1;
iter = contig_iter_new(xx->io, xx->cnum, 1,
CITER_LAST | CITER_IEND,
start, end);
ifunc = contig_iter_prev;
best_pos = INT_MIN;
}
if (!iter)
return -1;
/*
* Parse value search string. It optionally includes two extra params
* separated by #. Ie:
* <string>#<N.mismatches>#<where>.
* <where> is 1 for readings, 2 for consensus, 3 for both.
*/
if (p = strchr(value, '#')) {
mismatches = atoi(p+1);
*p = 0;
if (p = strchr(p+1, '#'))
where = atoi(p+1);
}
/* uppercase search string, remove pads, and store fwd/rev copies */
patlen = strlen(value);
depad_seq(value, &patlen, NULL);
if (NULL == (uppert = (char *)xmalloc(patlen + 1)))
return 0;
if (NULL == (upperb = (char *)xmalloc(patlen + 1)))
return 0;
uppert[patlen] = upperb[patlen] = 0;
for (i = patlen-1; i >= 0; i--) {
upperb[i] = uppert[i] = toupper(value[i]);
}
complement_seq(upperb, patlen);
while ((r = ifunc(xx->io, iter))) {
seq_t *s, *sorig;
char *ind, *indt = NULL, *indb = NULL, *seq;
int seq_len, comp, off = 0;
if (found && dir && r->start > best_pos)
break;
if (found && !dir && r->end < best_pos)
break;
if (NULL == (s = sorig = cache_search(xx->io, GT_Seq, r->rec)))
break;
if (r->comp ^ (s->len < 0)) {
s = dup_seq(s);
complement_seq_t(s);
}
seq = s->seq;
seq_len = ABS(s->len);
if (r->start < start) {
off = start - r->start;
seq += off;
seq_len -= off;
}
if (r->end - (patlen-1) > end)
seq_len -= r->end - (patlen-1) - end;
if (dir) {
if (strand == '+' || strand == '=')
indt = pstrnstr_inexact(seq, seq_len, uppert, patlen,
mismatches, NULL);
if (strand == '-' || strand == '=')
indb = pstrnstr_inexact(seq, seq_len, upperb, patlen,
mismatches, NULL);
} else {
if (strand == '+' || strand == '=')
indt = prstrnstr_inexact(seq, seq_len, uppert, patlen,
mismatches, NULL);
if (strand == '-' || strand == '=')
indb = prstrnstr_inexact(seq, seq_len, upperb, patlen,
mismatches, NULL);
}
if (indt && indb)
ind = MIN(indt, indb);
else if (indt)
ind = indt;
else if (indb)
ind = indb;
else
ind = NULL;
if (ind) {
int pos = r->start + ind - seq + off;
if (dir) {
if (best_pos > pos) {
found = 1;
best_pos = pos;
fpos = ind - s->seq;
fseq = r->rec;
}
} else {
if (best_pos < pos) {
found = 1;
best_pos = pos;
fpos = ind - s->seq;
fseq = r->rec;
}
}
//printf("Matches #%"PRIrec": at abs pos %d\n", r->rec, pos);
}
if (s != sorig)
free(s);
}
if (found) {
edSetCursorPos(xx, fseq == xx->cnum ? GT_Contig : GT_Seq,
fseq, fpos, 1);
}
free(uppert);
free(upperb);
contig_iter_del(iter);
return found ? 0 : -1;
}
/*
* Scans through one or more contigs checking each reading for correct
* assembly. This is simply a check for misaligned data, not looking into
* cutoff data. (The gap4 method did this, but it hasn't yet been implemented
* in gap5).
*
* Returns -1 for failure, 0 for success.
*/
int check_assembly(GapIO *io, int num_contigs, contig_list_t *contigs,
int winsize, float maxperc, int ignore_N) {
int i, sc, count = 0, allocated = 0;
char *con;
tg_rec *reads = NULL, *conts = NULL;
int *score = NULL, *length = NULL, *pos = NULL;
for (i = 0; i < num_contigs; i++) {
tg_rec crec = contigs[i].contig;
contig_iterator *ci;
rangec_t *r;
int start = contigs[i].start, end = contigs[i].end;
if (NULL == (con = (char *)xmalloc(end-start+1)))
return -1;
calculate_consensus_simple(io, crec, start, end, con, NULL);
ci = contig_iter_new(io, crec, 0, CITER_FIRST, start, end);
while (NULL != (r = contig_iter_next(io, ci))) {
UpdateTextOutput();
sc = check_uassembly_single(io, con - start, crec, r,
maxperc, winsize, ignore_N);
if (count >= allocated) {
allocated = allocated ? allocated * 2 : 256;
reads = xrealloc(reads, allocated * sizeof(*reads));
conts = xrealloc(conts, allocated * sizeof(*conts));
score = xrealloc(score, allocated * sizeof(*score));
length = xrealloc(length, allocated * sizeof(*length));
pos = xrealloc(pos, allocated * sizeof(*pos));
if (!reads || !conts || !score || !length || !pos)
goto error;
}
if (sc > 0) {
reads[count] = r->rec;
score[count] = sc * 100;
pos[count] = r->start;
length[count] = r->end - r->start+1;
conts[count++] = crec;
}
}
contig_iter_del(ci);
xfree(con);
}
if (-1 == check_assembly_plot(io, reads, conts, score, pos, length, count))
goto error;
if (reads)
xfree(reads);
if (conts)
xfree(conts);
if (pos)
xfree(pos);
if (length)
xfree(length);
if (score)
xfree(score);
return 0;
error:
if (reads)
xfree(reads);
if (conts)
xfree(conts);
if (pos)
xfree(pos);
if (length)
xfree(length);
if (score)
xfree(score);
return -1;
}
/**
* Builds and returns MALIGN from a Gap5 IO handle for the contig 'cnum'.
*/
MALIGN *build_malign(GapIO *io, tg_rec cnum, int start, int end) {
CONTIGL *contig, *first_contig = NULL, *last_contig = NULL;
int i, j;
contig_iterator *citer;
rangec_t *r;
/* Expand start and end to the range covered by seqs overlapping
* start .. end
*/
{
seq_t *s;
citer = contig_iter_new(io, cnum, 0,
CITER_FIRST | CITER_ICLIPPEDSTART,
start, start);
r = contig_iter_next(io, citer);
if (r) {
s = cache_search(io, GT_Seq, r->rec);
start = ((s->len < 0) ^ r->comp)
? r->end - s->right - 2
: r->start + s->left - 2;
}
contig_iter_del(citer);
}
{
seq_t *s;
citer = contig_iter_new(io, cnum, 0,
CITER_LAST | CITER_ICLIPPEDEND,
end, end);
r = contig_iter_next(io, citer);
if (r) {
s = cache_search(io, GT_Seq, r->rec);
end = ((s->len < 0) ^ r->comp)
? r->end - s->left + 2
: r->start + s->right + 2;
}
contig_iter_del(citer);
}
//printf("Generating data for %d..%d\n", start, end);
/* Generate contigl linked list */
//citer = contig_iter_new(io, cnum, 1, CITER_FIRST, CITER_CSTART, CITER_CEND);
citer = contig_iter_new(io, cnum, 0, CITER_FIRST, start, end);
while ((r = contig_iter_next(io, citer))) {
seq_t *s, *sorig;
char *seq;
int len;
assert((r->flags & GRANGE_FLAG_ISMASK) == GRANGE_FLAG_ISSEQ);
contig = create_contig_link();
contig->id = r->rec;
contig->mseg = create_mseg();
sorig = s = cache_search(io, GT_Seq, r->rec);
/* Check for out-of-bounds clip points. It shouldn't happen, but
gap5 databases have been seen with this problem, and we
don't want to crash if there are any. */
if (s->left < 1) s->left = 1;
if (s->right > ABS(s->len)) s->right = ABS(s->len);
/* Fix reads of zero length */
if (s->right < s->left) {
sorig = s = cache_rw(io, s);
s->right = s->left;
if (s->right > ABS(s->len))
s->left = s->right = ABS(s->len);
}
if ((s->len < 0) ^ r->comp) {
s = dup_seq(s);
complement_seq_t(s);
}
len = s->right - s->left + 1;
if (NULL == (seq = malloc(len+1)))
return NULL;
for (j = 0, i = s->left-1; i < s->right; i++, j++) {
/* Protect against the sequence containing "."; our pad sym */
if (s->seq[i] == '.')
seq[j] = 'N';
else
seq[j] = s->seq[i];
}
seq[j] = 0;
init_mseg(contig->mseg, seq, len, r->start-1 + s->left-1);
contig->mseg->comp = (s != sorig);
if (last_contig) {
last_contig->next = contig;
} else {
first_contig = contig;
}
last_contig = contig;
if (s != sorig)
free(s);
}
contig_iter_del(citer);
/* for 454 data -6 to -10 seem to work fine */
return contigl_to_malign(first_contig, -7, -7);
}