X-Git-Url: http://woldlab.caltech.edu/gitweb/?p=samtools.git;a=blobdiff_plain;f=bam_pileup.c;h=3c41a169127fcc5d3977df3e8febd944b7ced77e;hp=f68f400ab5cd086dc88a33e14899abe6cd4e53bb;hb=cb12a866906ec4ac644de0e658679261c82ab098;hpb=3a3b72190c04cb076438f5027e13fcf5ddee77be

diff --git a/bam_pileup.c b/bam_pileup.c
index f68f400..3c41a16 100644
--- a/bam_pileup.c
+++ b/bam_pileup.c
@@ -73,18 +73,28 @@ static inline int resolve_cigar(bam_pileup1_t *p, uint32_t pos)
 				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)
+					int has_next_match = 0;
+					unsigned i;
+					for (i = k + 1; i < c->n_cigar; ++i) {
+						uint32_t cigar = bam1_cigar(b)[i];
+						int opi = cigar&BAM_CIGAR_MASK;
+						if (opi == BAM_CMATCH) {
+							has_next_match = 1;
+							break;
+						} else if (opi == BAM_CSOFT_CLIP || opi == BAM_CREF_SKIP || opi == BAM_CHARD_CLIP) break;
+					}
+					if (!has_next_match) p->is_tail = 1;
+					if (k < c->n_cigar - 1 && has_next_match) { // 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;
+						else if (op_next == BAM_CPAD && k + 2 < c->n_cigar) { // no working for adjacent padding
+							cigar = bam1_cigar(b)[k+2]; op_next = cigar&BAM_CIGAR_MASK;
+							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_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
+					}
 				}
 			}
 			x += l; y += l;
@@ -96,7 +106,8 @@ static inline int resolve_cigar(bam_pileup1_t *p, uint32_t pos)
 			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 (is_restart) is_restart ^= (op == BAM_CMATCH);
+		else 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;
@@ -108,119 +119,167 @@ static inline int resolve_cigar(bam_pileup1_t *p, uint32_t pos)
 
 /* --- END: Auxiliary functions */
 
-struct __bam_plbuf_t {
+/*******************
+ * pileup iterator *
+ *******************/
+
+struct __bam_plp_t {
 	mempool_t *mp;
 	lbnode_t *head, *tail, *dummy;
-	bam_pileup_f func;
-	void *func_data;
 	int32_t tid, pos, max_tid, max_pos;
-	int max_pu, is_eof;
-	bam_pileup1_t *pu;
-	int flag_mask;
+	int is_eof, flag_mask, max_plp, error;
+	bam_pileup1_t *plp;
+	// for the "auto" interface only
+	bam1_t *b;
+	bam_plp_auto_f func;
+	void *data;
 };
 
-void bam_plbuf_reset(bam_plbuf_t *buf)
+bam_plp_t bam_plp_init(bam_plp_auto_f func, void *data)
 {
-	lbnode_t *p, *q;
-	buf->max_tid = buf->max_pos = -1;
-	buf->tid = buf->pos = 0;
-	buf->is_eof = 0;
-	for (p = buf->head; p->next;) {
-		q = p->next;
-		mp_free(buf->mp, p);
-		p = q;
+	bam_plp_t iter;
+	iter = calloc(1, sizeof(struct __bam_plp_t));
+	iter->mp = mp_init();
+	iter->head = iter->tail = mp_alloc(iter->mp);
+	iter->dummy = mp_alloc(iter->mp);
+	iter->max_tid = iter->max_pos = -1;
+	iter->flag_mask = BAM_DEF_MASK;
+	if (func) {
+		iter->func = func;
+		iter->data = data;
+		iter->b = bam_init1();
 	}
-	buf->head = buf->tail;
+	return iter;
 }
 
-void bam_plbuf_set_mask(bam_plbuf_t *buf, int mask)
-{
-	if (mask < 0) buf->flag_mask = BAM_DEF_MASK;
-	else buf->flag_mask = BAM_FUNMAP | mask;
-}
-
-bam_plbuf_t *bam_plbuf_init(bam_pileup_f func, void *data)
+void bam_plp_destroy(bam_plp_t iter)
 {
-	bam_plbuf_t *buf;
-	buf = (bam_plbuf_t*)calloc(1, sizeof(bam_plbuf_t));
-	buf->func = func; buf->func_data = data;
-	buf->mp = mp_init();
-	buf->head = buf->tail = mp_alloc(buf->mp);
-	buf->dummy = mp_alloc(buf->mp);
-	buf->max_tid = buf->max_pos = -1;
-	buf->flag_mask = BAM_DEF_MASK;
-	return buf;
+	mp_free(iter->mp, iter->dummy);
+	mp_free(iter->mp, iter->head);
+	if (iter->mp->cnt != 0)
+		fprintf(stderr, "[bam_plp_destroy] memory leak: %d. Continue anyway.\n", iter->mp->cnt);
+	mp_destroy(iter->mp);
+	if (iter->b) bam_destroy1(iter->b);
+	free(iter->plp);
+	free(iter);
 }
 
-void bam_plbuf_destroy(bam_plbuf_t *buf)
+const bam_pileup1_t *bam_plp_next(bam_plp_t iter, int *_tid, int *_pos, int *_n_plp)
 {
-	mp_free(buf->mp, buf->dummy);
-	mp_free(buf->mp, buf->head);
-	if (buf->mp->cnt != 0)
-		fprintf(stderr, "[bam_plbuf_destroy] memory leak: %d. Continue anyway.\n", buf->mp->cnt);
-	mp_destroy(buf->mp);
-	free(buf->pu);
-	free(buf);
+	if (iter->error) { *_n_plp = -1; return 0; }
+	*_n_plp = 0;
+	if (iter->is_eof && iter->head->next == 0) return 0;
+	while (iter->is_eof || iter->max_tid > iter->tid || (iter->max_tid == iter->tid && iter->max_pos > iter->pos)) {
+		int n_plp = 0;
+		lbnode_t *p, *q;
+		// write iter->plp at iter->pos
+		iter->dummy->next = iter->head;
+		for (p = iter->head, q = iter->dummy; p->next; q = p, p = p->next) {
+			if (p->b.core.tid < iter->tid || (p->b.core.tid == iter->tid && p->end <= iter->pos)) { // then remove
+				q->next = p->next; mp_free(iter->mp, p); p = q;
+			} else if (p->b.core.tid == iter->tid && p->beg <= iter->pos) { // here: p->end > pos; then add to pileup
+				if (n_plp == iter->max_plp) { // then double the capacity
+					iter->max_plp = iter->max_plp? iter->max_plp<<1 : 256;
+					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
+			}
+		}
+		iter->head = iter->dummy->next; // dummy->next may be changed
+		*_n_plp = n_plp; *_tid = iter->tid; *_pos = iter->pos;
+		// update iter->tid and iter->pos
+		if (iter->head->next) {
+			if (iter->tid > iter->head->b.core.tid) {
+				fprintf(stderr, "[%s] unsorted input. Pileup aborts.\n", __func__);
+				iter->error = 1;
+				*_n_plp = -1;
+				return 0;
+			}
+		}
+		if (iter->tid < iter->head->b.core.tid) { // come to a new reference sequence
+			iter->tid = iter->head->b.core.tid; iter->pos = iter->head->beg; // jump to the next reference
+		} else if (iter->pos < iter->head->beg) { // here: tid == head->b.core.tid
+			iter->pos = iter->head->beg; // jump to the next position
+		} else ++iter->pos; // scan contiguously
+		// return
+		if (n_plp) return iter->plp;
+		if (iter->is_eof && iter->head->next == 0) break;
+	}
+	return 0;
 }
 
-int bam_plbuf_push(const bam1_t *b, bam_plbuf_t *buf)
+int bam_plp_push(bam_plp_t iter, const bam1_t *b)
 {
-	if (b) { // fill buffer
+	if (iter->error) return -1;
+	if (b) {
 		if (b->core.tid < 0) return 0;
-		if (b->core.flag & buf->flag_mask) return 0;
-		bam_copy1(&buf->tail->b, b);
-		buf->tail->beg = b->core.pos; buf->tail->end = bam_calend(&b->core, bam1_cigar(b));
-		if (b->core.tid < buf->max_tid) {
+		if (b->core.flag & iter->flag_mask) return 0;
+		bam_copy1(&iter->tail->b, b);
+		iter->tail->beg = b->core.pos; iter->tail->end = bam_calend(&b->core, bam1_cigar(b));
+		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;
 			return -1;
 		}
-		if ((b->core.tid == buf->max_tid) && (buf->tail->beg < buf->max_pos)) {
+		if ((b->core.tid == iter->max_tid) && (iter->tail->beg < iter->max_pos)) {
 			fprintf(stderr, "[bam_pileup_core] the input is not sorted (reads out of order)\n");
+			iter->error = 1;
 			return -1;
 		}
-		buf->max_tid = b->core.tid; buf->max_pos = buf->tail->beg;
-		if (buf->tail->end > buf->pos || buf->tail->b.core.tid > buf->tid) {
-			buf->tail->next = mp_alloc(buf->mp);
-			buf->tail = buf->tail->next;
-		}
-	} else buf->is_eof = 1;
-	while (buf->is_eof || buf->max_tid > buf->tid || (buf->max_tid == buf->tid && buf->max_pos > buf->pos)) {
-		int n_pu = 0;
-		lbnode_t *p, *q;
-		buf->dummy->next = buf->head;
-		for (p = buf->head, q = buf->dummy; p->next; q = p, p = p->next) {
-			if (p->b.core.tid < buf->tid || (p->b.core.tid == buf->tid && p->end <= buf->pos)) { // then remove from the list
-				q->next = p->next; mp_free(buf->mp, p); p = q;
-			} else if (p->b.core.tid == buf->tid && p->beg <= buf->pos) { // here: p->end > pos; then add to pileup
-				if (n_pu == buf->max_pu) { // then double the capacity
-					buf->max_pu = buf->max_pu? buf->max_pu<<1 : 256;
-					buf->pu = (bam_pileup1_t*)realloc(buf->pu, sizeof(bam_pileup1_t) * buf->max_pu);
-				}
-				buf->pu[n_pu].b = &p->b;
-				if (resolve_cigar(buf->pu + n_pu, buf->pos)) ++n_pu; // skip the read if we are looking at BAM_CREF_SKIP
-			}
+		iter->max_tid = b->core.tid; iter->max_pos = iter->tail->beg;
+		if (iter->tail->end > iter->pos || iter->tail->b.core.tid > iter->tid) {
+			iter->tail->next = mp_alloc(iter->mp);
+			iter->tail = iter->tail->next;
 		}
-		buf->head = buf->dummy->next; // dummy->next may be changed
-		if (n_pu) { // then call user defined function
-			buf->func(buf->tid, buf->pos, n_pu, buf->pu, buf->func_data);
-		}
-		// update tid and pos
-		if (buf->head->next) {
-			if (buf->tid > buf->head->b.core.tid) {
-				fprintf(stderr, "[bam_plbuf_push] unsorted input. Pileup aborts.\n");
-				return 1;
+	} else iter->is_eof = 1;
+	return 0;
+}
+
+const bam_pileup1_t *bam_plp_auto(bam_plp_t iter, int *_tid, int *_pos, int *_n_plp)
+{
+	const bam_pileup1_t *plp;
+	if (iter->func == 0 || iter->error) { *_n_plp = -1; return 0; }
+	if ((plp = bam_plp_next(iter, _tid, _pos, _n_plp)) != 0) return plp;
+	else {
+		*_n_plp = 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, _tid, _pos, _n_plp)) != 0) return plp;
 		}
-		if (buf->tid < buf->head->b.core.tid) { // come to a new reference sequence
-			buf->tid = buf->head->b.core.tid; buf->pos = buf->head->beg; // jump to the next reference
-		} else if (buf->pos < buf->head->beg) { // here: tid == head->b.core.tid
-			buf->pos = buf->head->beg; // jump to the next position
-		} else ++buf->pos; // scan contiguously
-		if (buf->is_eof && buf->head->next == 0) break;
+		bam_plp_push(iter, 0);
+		if ((plp = bam_plp_next(iter, _tid, _pos, _n_plp)) != 0) return plp;
+		return 0;
 	}
-	return 0;
 }
 
+void bam_plp_reset(bam_plp_t iter)
+{
+	lbnode_t *p, *q;
+	iter->max_tid = iter->max_pos = -1;
+	iter->tid = iter->pos = 0;
+	iter->is_eof = 0;
+	for (p = iter->head; p->next;) {
+		q = p->next;
+		mp_free(iter->mp, p);
+		p = q;
+	}
+	iter->head = iter->tail;
+}
+
+void bam_plp_set_mask(bam_plp_t iter, int mask)
+{
+	iter->flag_mask = mask < 0? BAM_DEF_MASK : (BAM_FUNMAP | mask);
+}
+
+/*****************
+ * callback APIs *
+ *****************/
+
 int bam_pileup_file(bamFile fp, int mask, bam_pileup_f func, void *func_data)
 {
 	bam_plbuf_t *buf;
@@ -236,3 +295,102 @@ int bam_pileup_file(bamFile fp, int mask, bam_pileup_f func, void *func_data)
 	bam_destroy1(b);
 	return 0;
 }
+
+void bam_plbuf_set_mask(bam_plbuf_t *buf, int mask)
+{
+	bam_plp_set_mask(buf->iter, mask);
+}
+
+void bam_plbuf_reset(bam_plbuf_t *buf)
+{
+	bam_plp_reset(buf->iter);
+}
+
+bam_plbuf_t *bam_plbuf_init(bam_pileup_f func, void *data)
+{
+	bam_plbuf_t *buf;
+	buf = calloc(1, sizeof(bam_plbuf_t));
+	buf->iter = bam_plp_init(0, 0);
+	buf->func = func;
+	buf->data = data;
+	return buf;
+}
+
+void bam_plbuf_destroy(bam_plbuf_t *buf)
+{
+	bam_plp_destroy(buf->iter);
+	free(buf);
+}
+
+int bam_plbuf_push(const bam1_t *b, bam_plbuf_t *buf)
+{
+	int ret, n_plp, tid, pos;
+	const bam_pileup1_t *plp;
+	ret = bam_plp_push(buf->iter, b);
+	if (ret < 0) return ret;
+	while ((plp = bam_plp_next(buf->iter, &tid, &pos, &n_plp)) != 0)
+		buf->func(tid, pos, n_plp, plp, buf->data);
+	return 0;
+}
+
+/***********
+ * mpileup *
+ ***********/
+
+struct __bam_mplp_t {
+	int n;
+	uint64_t min, *pos;
+	bam_plp_t *iter;
+	int *n_plp;
+	const bam_pileup1_t **plp;
+};
+
+bam_mplp_t bam_mplp_init(int n, bam_plp_auto_f func, void **data)
+{
+	int i;
+	bam_mplp_t iter;
+	iter = calloc(1, sizeof(struct __bam_mplp_t));
+	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) {
+		iter->iter[i] = bam_plp_init(func, data[i]);
+		iter->pos[i] = iter->min;
+	}
+	return iter;
+}
+
+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->iter); free(iter->pos); free(iter->n_plp); free(iter->plp);
+	free(iter);
+}
+
+int bam_mplp_auto(bam_mplp_t iter, int *_tid, int *_pos, int *n_plp, const bam_pileup1_t **plp)
+{
+	int i, ret = 0;
+	uint64_t new_min = (uint64_t)-1;
+	for (i = 0; i < iter->n; ++i) {
+		if (iter->pos[i] == iter->min) {
+			int tid, pos;
+			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->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) {
+			n_plp[i] = iter->n_plp[i], plp[i] = iter->plp[i];
+			++ret;
+		} else n_plp[i] = 0, plp[i] = 0;
+	}
+	return ret;
+}