#include <assert.h>
#include "sam.h"
+typedef struct {
+ int k, x, y, end;
+} cstate_t;
+
+static cstate_t g_cstate_null = { -1, 0, 0, 0 };
+
typedef struct __linkbuf_t {
bam1_t b;
uint32_t beg, end;
+ cstate_t s;
struct __linkbuf_t *next;
} lbnode_t;
/* --- BEGIN: Auxiliary functions */
-static inline int resolve_cigar(bam_pileup1_t *p, uint32_t pos)
+/* s->k: the index of the CIGAR operator that has just been processed.
+ s->x: the reference coordinate of the start of s->k
+ s->y: the query coordiante of the start of s->k
+ */
+static inline int resolve_cigar2(bam_pileup1_t *p, uint32_t pos, cstate_t *s)
{
- unsigned k;
+#define _cop(c) ((c)&BAM_CIGAR_MASK)
+#define _cln(c) ((c)>>BAM_CIGAR_SHIFT)
+
bam1_t *b = p->b;
bam1_core_t *c = &b->core;
- uint32_t x = c->pos, y = 0;
- int ret = 1, is_restart = 1;
-
- if (c->flag&BAM_FUNMAP) return 0; // unmapped read
- assert(x <= pos); // otherwise a bug
- p->qpos = -1; p->indel = 0; p->is_del = p->is_head = p->is_tail = 0;
- for (k = 0; k < c->n_cigar; ++k) {
- int op = bam1_cigar(b)[k] & BAM_CIGAR_MASK; // operation
- int l = bam1_cigar(b)[k] >> BAM_CIGAR_SHIFT; // length
- if (op == BAM_CMATCH) { // NOTE: this assumes the first and the last operation MUST BE a match or a clip
- if (x + l > pos) { // overlap with pos
- p->indel = p->is_del = 0;
- p->qpos = y + (pos - x);
- if (x == pos && is_restart) p->is_head = 1;
- if (x + l - 1 == pos) { // come to the end of a match
- if (k < c->n_cigar - 1) { // there are additional operation(s)
- uint32_t cigar = bam1_cigar(b)[k+1]; // next CIGAR
- int op_next = cigar&BAM_CIGAR_MASK; // next CIGAR operation
- if (op_next == BAM_CDEL) p->indel = -(int32_t)(cigar>>BAM_CIGAR_SHIFT); // del
- else if (op_next == BAM_CINS) p->indel = cigar>>BAM_CIGAR_SHIFT; // ins
- if (op_next == BAM_CDEL || op_next == BAM_CINS) {
- if (k + 2 < c->n_cigar) op_next = bam1_cigar(b)[k+2]&BAM_CIGAR_MASK;
- else p->is_tail = 1;
- }
- if (op_next == BAM_CSOFT_CLIP || op_next == BAM_CREF_SKIP || op_next == BAM_CHARD_CLIP)
- p->is_tail = 1; // tail
- } else p->is_tail = 1; // this is the last operation; set tail
+ uint32_t *cigar = bam1_cigar(b);
+ int k, is_head = 0;
+ // determine the current CIGAR operation
+// fprintf(stderr, "%s\tpos=%d\tend=%d\t(%d,%d,%d)\n", bam1_qname(b), pos, s->end, s->k, s->x, s->y);
+ if (s->k == -1) { // never processed
+ is_head = 1;
+ if (c->n_cigar == 1) { // just one operation, save a loop
+ if (_cop(cigar[0]) == BAM_CMATCH) s->k = 0, s->x = c->pos, s->y = 0;
+ } else { // find the first match
+ for (k = 0, s->x = c->pos, s->y = 0; k < c->n_cigar; ++k) {
+ int op = _cop(cigar[k]);
+ int l = _cln(cigar[k]);
+ if (op == BAM_CMATCH) break;
+ else if (op == BAM_CDEL || op == BAM_CREF_SKIP) s->x += l;
+ else if (op == BAM_CINS || op == BAM_CSOFT_CLIP) s->y += l;
+ }
+ assert(k < c->n_cigar);
+ s->k = k;
+ }
+ } else { // the read has been processed before
+ int op, l = _cln(cigar[s->k]);
+ if (pos - s->x >= l) { // jump to the next operation
+ assert(s->k < c->n_cigar); // otherwise a bug: this function should not be called in this case
+ op = _cop(cigar[s->k+1]);
+ if (op == BAM_CMATCH || op == BAM_CDEL || op == BAM_CREF_SKIP) { // jump to the next without a loop
+ if (_cop(cigar[s->k]) == BAM_CMATCH) s->y += l;
+ s->x += l;
+ ++s->k;
+ } else { // find the next M/D/N
+ if (_cop(cigar[s->k]) == BAM_CMATCH) s->y += l;
+ s->x += l;
+ for (k = s->k + 1; k < c->n_cigar; ++k) {
+ op = _cop(cigar[k]), l = _cln(cigar[k]);
+ if (op == BAM_CMATCH || op == BAM_CDEL || op == BAM_CREF_SKIP) break;
+ else if (op == BAM_CINS || op == BAM_CSOFT_CLIP) s->y += l;
}
+ s->k = k;
}
- x += l; y += l;
- } else if (op == BAM_CDEL) { // then set ->is_del
- if (x + l > pos) {
- p->indel = 0; p->is_del = 1;
- p->qpos = y + (pos - x);
+ assert(s->k < c->n_cigar); // otherwise a bug
+ } // else, do nothing
+ }
+ { // collect pileup information
+ int op, l;
+ op = _cop(cigar[s->k]); l = _cln(cigar[s->k]);
+ p->is_del = p->indel = p->is_refskip = 0;
+ if (s->x + l - 1 == pos && s->k + 1 < c->n_cigar) { // peek the next operation
+ int op2 = _cop(cigar[s->k+1]);
+ int l2 = _cln(cigar[s->k+1]);
+ if (op2 == BAM_CDEL) p->indel = -(int)l2;
+ else if (op2 == BAM_CINS) p->indel = l2;
+ else if (op2 == BAM_CPAD && s->k + 2 < c->n_cigar) { // no working for adjacent padding
+ int l3 = 0;
+ for (k = s->k + 2; k < c->n_cigar; ++k) {
+ op2 = _cop(cigar[k]); l2 = _cln(cigar[k]);
+ if (op2 == BAM_CINS) l3 += l2;
+ else if (op2 == BAM_CDEL || op2 == BAM_CMATCH || op2 == BAM_CREF_SKIP) break;
+ }
+ if (l3 > 0) p->indel = l3;
}
- x += l;
- } else if (op == BAM_CREF_SKIP) x += l;
- else if (op == BAM_CINS || op == BAM_CSOFT_CLIP) y += l;
- is_restart = (op == BAM_CREF_SKIP || op == BAM_CSOFT_CLIP || op == BAM_CHARD_CLIP);
- if (x > pos) {
- if (op == BAM_CREF_SKIP) ret = 0; // then do not put it into pileup at all
- break;
}
+ if (op == BAM_CMATCH) {
+ p->qpos = s->y + (pos - s->x);
+ } else if (op == BAM_CDEL || op == BAM_CREF_SKIP) {
+ p->is_del = 1; p->qpos = s->y; // FIXME: distinguish D and N!!!!!
+ p->is_refskip = (op == BAM_CREF_SKIP);
+ } // cannot be other operations; otherwise a bug
+ p->is_head = (pos == c->pos); p->is_tail = (pos == s->end);
}
- assert(x > pos); // otherwise a bug
- return ret;
+ return 1;
}
/* --- END: Auxiliary functions */
mempool_t *mp;
lbnode_t *head, *tail, *dummy;
int32_t tid, pos, max_tid, max_pos;
- int is_eof, flag_mask, max_plp, error;
+ int is_eof, flag_mask, max_plp, error, maxcnt;
bam_pileup1_t *plp;
// for the "auto" interface only
bam1_t *b;
iter->dummy = mp_alloc(iter->mp);
iter->max_tid = iter->max_pos = -1;
iter->flag_mask = BAM_DEF_MASK;
+ iter->maxcnt = 8000;
if (func) {
iter->func = func;
iter->data = data;
iter->plp = (bam_pileup1_t*)realloc(iter->plp, sizeof(bam_pileup1_t) * iter->max_plp);
}
iter->plp[n_plp].b = &p->b;
- if (resolve_cigar(iter->plp + n_plp, iter->pos)) ++n_plp; // skip the read if we are looking at ref-skip
+ if (resolve_cigar2(iter->plp + n_plp, iter->pos, &p->s)) ++n_plp; // actually always true...
}
}
iter->head = iter->dummy->next; // dummy->next may be changed
if (b) {
if (b->core.tid < 0) return 0;
if (b->core.flag & iter->flag_mask) return 0;
+ if (iter->tid == b->core.tid && iter->pos == b->core.pos && iter->mp->cnt > iter->maxcnt) return 0;
bam_copy1(&iter->tail->b, b);
iter->tail->beg = b->core.pos; iter->tail->end = bam_calend(&b->core, bam1_cigar(b));
+ iter->tail->s = g_cstate_null; iter->tail->s.end = iter->tail->end - 1; // initialize cstate_t
if (b->core.tid < iter->max_tid) {
fprintf(stderr, "[bam_pileup_core] the input is not sorted (chromosomes out of order)\n");
iter->error = 1;
{
const bam_pileup1_t *plp;
if (iter->func == 0 || iter->error) { *_n_plp = -1; return 0; }
- if ((plp = bam_plp_next(iter, _n_plp, _tid, _pos)) != 0) return plp;
- else {
+ if ((plp = bam_plp_next(iter, _tid, _pos, _n_plp)) != 0) return plp;
+ else { // no pileup line can be obtained; read alignments
*_n_plp = 0;
- while (iter->func(iter->b, iter->data) >= 0) {
+ if (iter->is_eof) return 0;
+ while (iter->func(iter->data, iter->b) >= 0) {
if (bam_plp_push(iter, iter->b) < 0) {
*_n_plp = -1;
return 0;
}
- if ((plp = bam_plp_next(iter, _n_plp, _tid, _pos)) != 0) return plp;
+ if ((plp = bam_plp_next(iter, _tid, _pos, _n_plp)) != 0) return plp;
+ // otherwise no pileup line can be returned; read the next alignment.
}
+ bam_plp_push(iter, 0);
+ if ((plp = bam_plp_next(iter, _tid, _pos, _n_plp)) != 0) return plp;
return 0;
}
}
iter->flag_mask = mask < 0? BAM_DEF_MASK : (BAM_FUNMAP | mask);
}
+void bam_plp_set_maxcnt(bam_plp_t iter, int maxcnt)
+{
+ iter->maxcnt = maxcnt;
+}
+
/*****************
* callback APIs *
*****************/
iter->pos = calloc(n, 8);
iter->n_plp = calloc(n, sizeof(int));
iter->plp = calloc(n, sizeof(void*));
+ iter->iter = calloc(n, sizeof(void*));
iter->n = n;
iter->min = (uint64_t)-1;
- for (i = 0; i < n; ++i)
+ for (i = 0; i < n; ++i) {
iter->iter[i] = bam_plp_init(func, data[i]);
+ iter->pos[i] = iter->min;
+ }
return iter;
}
+void bam_mplp_set_maxcnt(bam_mplp_t iter, int maxcnt)
+{
+ int i;
+ for (i = 0; i < iter->n; ++i)
+ iter->iter[i]->maxcnt = maxcnt;
+}
+
void bam_mplp_destroy(bam_mplp_t iter)
{
int i;
for (i = 0; i < iter->n; ++i) bam_plp_destroy(iter->iter[i]);
- free(iter->pos); free(iter->n_plp); free(iter->plp);
+ free(iter->iter); free(iter->pos); free(iter->n_plp); free(iter->plp);
free(iter);
}
iter->plp[i] = bam_plp_auto(iter->iter[i], &tid, &pos, &iter->n_plp[i]);
iter->pos[i] = (uint64_t)tid<<32 | pos;
}
- if (iter->pos[i] < new_min) new_min = iter->pos[i];
+ if (iter->plp[i] && iter->pos[i] < new_min) new_min = iter->pos[i];
}
iter->min = new_min;
if (new_min == (uint64_t)-1) return 0;
*_tid = new_min>>32; *_pos = (uint32_t)new_min;
for (i = 0; i < iter->n; ++i) {
- if (iter->pos[i] == iter->min) {
+ if (iter->pos[i] == iter->min) { // FIXME: valgrind reports "uninitialised value(s) at this line"
n_plp[i] = iter->n_plp[i], plp[i] = iter->plp[i];
++ret;
} else n_plp[i] = 0, plp[i] = 0;