r219: added --outc to filter out very short hits

main.c

@@ -16,6 +16,7 @@ static ko_longopt_t long_options[] = {
 	{ "no-pre-chain",    ko_no_argument,       310 },
 	{ "aln",             ko_no_argument,       311 },
 	{ "max-intron-out",  ko_required_argument, 312 },
+	{ "outc",            ko_required_argument, 313 },
 	{ "version",         ko_no_argument,       401 },
 	{ "no-kalloc",       ko_no_argument,       501 },
 	{ "dbg-qname",       ko_no_argument,       502 },
@@ -83,6 +84,7 @@ static void print_usage(FILE *fp, const mp_idxopt_t *io, const mp_mapopt_t *mo,
 	fprintf(fp, "    -u           print unmapped query proteins in PAF\n");
 	fprintf(fp, "    --outn=NUM   output up to min{NUM,-N} alignments per query [%d]\n", mo->out_n);
 	fprintf(fp, "    --outs=FLOAT output if score at least FLOAT*bestScore [%g]\n", mo->out_sim);
+	fprintf(fp, "    --outc=FLOAT output if at least FLOAT fraction of query is aligned [%g]\n", mo->out_cov);
 	fprintf(fp, "    -K NUM       query batch size [2M]\n");
 }
 
@@ -140,6 +142,7 @@ int main(int argc, char *argv[])
 		else if (c == 310) mo.flag |= MP_F_NO_PRE_CHAIN; // --no-pre-chain
 		else if (c == 311) mo.flag |= MP_F_SHOW_RESIDUE; // --aln
 		else if (c == 312) mo.max_intron_flank = (mp_parse_num(o.arg) + 1) / 2; // --max-intron-out
+		else if (c == 313) mo.out_cov = atof(o.arg); // --outc
 		else if (c == 501) mp_dbg_flag |= MP_DBG_NO_KALLOC; // --no-kalloc
 		else if (c == 502) mp_dbg_flag |= MP_DBG_QNAME; // --dbg-qname
 		else if (c == 503) mp_dbg_flag |= MP_DBG_NO_REFINE; // --dbg-no-refine

map.c

@@ -301,9 +301,11 @@ static void *worker_pipeline(void *shared, int step, void *in)
 				best_sc = s->reg[i][0].p? s->reg[i][0].p->dp_max : s->reg[i][0].chn_sc;
 			}
 			for (j = 0; j < s->n_reg[i] && j < p->opt->out_n; ++j) {
-				int32_t sc = s->reg[i][j].p? s->reg[i][j].p->dp_max : s->reg[i][j].chn_sc;
+				const mp_reg1_t *r = &s->reg[i][j];
+				int32_t sc = r->p? r->p->dp_max : r->chn_sc;
 				if (sc <= 0 || sc < (double)best_sc * p->opt->out_sim) continue;
-				mp_write_output(&p->str, 0, p->mi, &s->seq[i], &s->reg[i][j], p->opt, ++p->id, j + 1);
+				if (r->qe - r->qs < (double)s->seq[i].l_seq * p->opt->out_cov) continue;
+				mp_write_output(&p->str, 0, p->mi, &s->seq[i], r, p->opt, ++p->id, j + 1);
 				fwrite(p->str.s, 1, p->str.l, stdout);
 			}
 			for (j = 0; j < s->n_reg[i]; ++j) {

miniprot.1

@@ -191,6 +191,11 @@ Output an alignment only if its score is at least
 .IR FLOAT *bestScore,
 where bestScore is the best alignment score of the protein [0.99]
 .TP
+.BI --outc \ FLOAT
+Output an alignment only if
+.I FLOAT
+fraction of the query protein is aligned [0.1]
+.TP
 .BI -K \ NUM
 Query batch size [2M]
 .SH OUTPUT FORMAT

miniprot.h

@@ -3,7 +3,7 @@
 
 #include <stdint.h>
 
-#define MP_VERSION "0.7-r218-dirty"
+#define MP_VERSION "0.7-r219-dirty"
 
 #define MP_F_NO_SPLICE    0x1
 #define MP_F_NO_ALIGN     0x2
@@ -55,7 +55,7 @@ typedef struct {
 	float mask_level;
 	int32_t mask_len;
 	float pri_ratio;
-	float out_sim;
+	float out_sim, out_cov;
 	int32_t best_n, out_n;
 	int32_t kmer2;
 	int32_t go, ge, io, fs; // gap open, extension, intron open, and frame-shift/stop-codon

options.c

@@ -61,6 +61,7 @@ void mp_mapopt_init(mp_mapopt_t *mo)
 	mo->best_n = 30;
 	mo->out_n = 1000;
 	mo->out_sim = 0.99f;
+	mo->out_cov = 0.1f;
 	#if MP_BITS_PER_AA == 4
 	mo->kmer2 = 5;
 	#else