r1108: fixed missing inversions for #816 and #806

hit.c

@@ -252,7 +252,7 @@ void mm_sync_regs(void *km, int n_regs, mm_reg1_t *regs) // keep mm_reg1_t::{id,
 	mm_set_sam_pri(n_regs, regs);
 }
 
-void mm_select_sub(void *km, float pri_ratio, int min_diff, int best_n, int *n_, mm_reg1_t *r)
+void mm_select_sub(void *km, float pri_ratio, int min_diff, int best_n, int check_strand, int min_strand_sc, int *n_, mm_reg1_t *r)
 {
 	if (pri_ratio > 0.0f && *n_ > 0) {
 		int i, k, n = *n_, n_2nd = 0;
@@ -264,6 +264,9 @@ void mm_select_sub(void *km, float pri_ratio, int min_diff, int best_n, int *n_,
 				if (!(r[i].qs == r[p].qs && r[i].qe == r[p].qe && r[i].rid == r[p].rid && r[i].rs == r[p].rs && r[i].re == r[p].re)) // not identical hits
 					r[k++] = r[i], ++n_2nd;
 				else if (r[i].p) free(r[i].p);
+			} else if (check_strand && n_2nd < best_n && r[i].score > min_strand_sc && r[i].rev != r[p].rev) {
+				r[i].strand_retained = 1;
+				r[k++] = r[i], ++n_2nd;
 			} else if (r[i].p) free(r[i].p);
 		}
 		if (k != n) mm_sync_regs(km, k, r); // removing hits requires sync()
@@ -271,6 +274,19 @@ void mm_select_sub(void *km, float pri_ratio, int min_diff, int best_n, int *n_,
 	}
 }
 
+int mm_filter_strand_retained(int n_regs, mm_reg1_t *r)
+{
+	int i, k;
+	for (i = k = 0; i < n_regs; ++i) {
+		int p = r[i].parent;
+		if (!r[i].strand_retained || r[i].div < r[p].div * 5.0f) {
+			if (k < i) r[k++] = r[i];
+			else ++k;
+		}
+	}
+	return k;
+}
+
 void mm_filter_regs(const mm_mapopt_t *opt, int qlen, int *n_regs, mm_reg1_t *regs)
 { // NB: after this call, mm_reg1_t::parent can be -1 if its parent filtered out
 	int i, k;

main.c

@@ -7,7 +7,7 @@
 #include "mmpriv.h"
 #include "ketopt.h"
 
-#define MM_VERSION "2.22-r1105-dirty"
+#define MM_VERSION "2.22-r1108-dirty"
 
 #ifdef __linux__
 #include <sys/resource.h>

map.c

@@ -212,7 +212,7 @@ static void chain_post(const mm_mapopt_t *opt, int max_chain_gap_ref, const mm_i
 {
 	if (!(opt->flag & MM_F_ALL_CHAINS)) { // don't choose primary mapping(s)
 		mm_set_parent(km, opt->mask_level, opt->mask_len, *n_regs, regs, opt->a * 2 + opt->b, opt->flag&MM_F_HARD_MLEVEL, opt->alt_drop);
-		if (n_segs <= 1) mm_select_sub(km, opt->pri_ratio, mi->k*2, opt->best_n, n_regs, regs);
+		if (n_segs <= 1) mm_select_sub(km, opt->pri_ratio, mi->k*2, opt->best_n, 1, opt->max_gap, n_regs, regs);
 		else mm_select_sub_multi(km, opt->pri_ratio, 0.2f, 0.7f, max_chain_gap_ref, mi->k*2, opt->best_n, n_segs, qlens, n_regs, regs);
 	}
 }
@@ -223,7 +223,7 @@ static mm_reg1_t *align_regs(const mm_mapopt_t *opt, const mm_idx_t *mi, void *k
 	regs = mm_align_skeleton(km, opt, mi, qlen, seq, n_regs, regs, a); // this calls mm_filter_regs()
 	if (!(opt->flag & MM_F_ALL_CHAINS)) { // don't choose primary mapping(s)
 		mm_set_parent(km, opt->mask_level, opt->mask_len, *n_regs, regs, opt->a * 2 + opt->b, opt->flag&MM_F_HARD_MLEVEL, opt->alt_drop);
-		mm_select_sub(km, opt->pri_ratio, mi->k*2, opt->best_n, n_regs, regs);
+		mm_select_sub(km, opt->pri_ratio, mi->k*2, opt->best_n, 0, opt->max_gap, n_regs, regs);
 		mm_set_sam_pri(*n_regs, regs);
 	}
 	return regs;
@@ -335,8 +335,10 @@ void mm_map_frag(const mm_idx_t *mi, int n_segs, const int *qlens, const char **
 						i == regs0[j].as? 0 : ((int32_t)a[i].y - (int32_t)a[i-1].y) - ((int32_t)a[i].x - (int32_t)a[i-1].x));
 
 	chain_post(opt, max_chain_gap_ref, mi, b->km, qlen_sum, n_segs, qlens, &n_regs0, regs0, a);
-	if (!is_sr && !(opt->flag&MM_F_QSTRAND))
+	if (!is_sr && !(opt->flag&MM_F_QSTRAND)) {
 		mm_est_err(mi, qlen_sum, n_regs0, regs0, a, n_mini_pos, mini_pos);
+		n_regs0 = mm_filter_strand_retained(n_regs0, regs0);
+	}
 
 	if (n_segs == 1) { // uni-segment
 		regs0 = align_regs(opt, mi, b->km, qlens[0], seqs[0], &n_regs0, regs0, a);
@@ -509,7 +511,7 @@ static void merge_hits(step_t *s)
 			mm_hit_sort(km, &s->n_reg[k], s->reg[k], opt->alt_drop);
 			mm_set_parent(km, opt->mask_level, opt->mask_len, s->n_reg[k], s->reg[k], opt->a * 2 + opt->b, opt->flag&MM_F_HARD_MLEVEL, opt->alt_drop);
 			if (!(opt->flag & MM_F_ALL_CHAINS)) {
-				mm_select_sub(km, opt->pri_ratio, s->p->mi->k*2, opt->best_n, &s->n_reg[k], s->reg[k]);
+				mm_select_sub(km, opt->pri_ratio, s->p->mi->k*2, opt->best_n, 0, opt->max_gap, &s->n_reg[k], s->reg[k]);
 				mm_set_sam_pri(s->n_reg[k], s->reg[k]);
 			}
 			mm_set_mapq(km, s->n_reg[k], s->reg[k], opt->min_chain_score, opt->a, rep_len, !!(opt->flag & MM_F_SR));

minimap.h

@@ -108,7 +108,7 @@ typedef struct {
 	int32_t mlen, blen;     // seeded exact match length; seeded alignment block length
 	int32_t n_sub;          // number of suboptimal mappings
 	int32_t score0;         // initial chaining score (before chain merging/spliting)
-	uint32_t mapq:8, split:2, rev:1, inv:1, sam_pri:1, proper_frag:1, pe_thru:1, seg_split:1, seg_id:8, split_inv:1, is_alt:1, dummy:6;
+	uint32_t mapq:8, split:2, rev:1, inv:1, sam_pri:1, proper_frag:1, pe_thru:1, seg_split:1, seg_id:8, split_inv:1, is_alt:1, strand_retained:1, dummy:5;
 	uint32_t hash;
 	float div;
 	mm_extra_t *p;

mmpriv.h

@@ -91,8 +91,9 @@ void mm_sync_regs(void *km, int n_regs, mm_reg1_t *regs);
 int mm_squeeze_a(void *km, int n_regs, mm_reg1_t *regs, mm128_t *a);
 int mm_set_sam_pri(int n, mm_reg1_t *r);
 void mm_set_parent(void *km, float mask_level, int mask_len, int n, mm_reg1_t *r, int sub_diff, int hard_mask_level, float alt_diff_frac);
-void mm_select_sub(void *km, float pri_ratio, int min_diff, int best_n, int *n_, mm_reg1_t *r);
+void mm_select_sub(void *km, float pri_ratio, int min_diff, int best_n, int check_strand, int min_strand_sc, int *n_, mm_reg1_t *r);
 void mm_select_sub_multi(void *km, float pri_ratio, float pri1, float pri2, int max_gap_ref, int min_diff, int best_n, int n_segs, const int *qlens, int *n_, mm_reg1_t *r);
+int mm_filter_strand_retained(int n_regs, mm_reg1_t *r);
 void mm_filter_regs(const mm_mapopt_t *opt, int qlen, int *n_regs, mm_reg1_t *regs);
 void mm_hit_sort(void *km, int *n_regs, mm_reg1_t *r, float alt_diff_frac);
 void mm_set_mapq(void *km, int n_regs, mm_reg1_t *regs, int min_chain_sc, int match_sc, int rep_len, int is_sr);