Release minimap2-2.12 (r827)

MANIFEST.in

@@ -4,6 +4,7 @@ include ksw2_dispatch.c
 include getopt.c
 include main.c
 include README.md
+include sse2neon/emmintrin.h
 include python/mappy.c
 include python/cmappy.h
 include python/cmappy.pxd

NEWS.md

@@ -1,3 +1,27 @@
+Release 2.12-r827 (6 August 2018)
+---------------------------------
+
+Changes to minimap2:
+
+ * Added option --split-prefix to write proper alignments (correct mapping
+   quality and clustered query sequences) given a multi-part index (#141 and
+   #189; mostly by @hasindu2008).
+
+ * Fixed a memory leak when option -y is in use.
+
+Changes to mappy:
+
+ * Support the MD/cs tag (#183 and #203).
+
+ * Allow mappy to index a single sequence, to add extra flags and to change the
+   scoring system.
+
+Minimap2 should produce alignments identical to v2.11.
+
+(2.12: 6 August 2018, r827)
+
+
+
 Release 2.11-r797 (20 June 2018)
 --------------------------------
 

README.md

@@ -15,8 +15,9 @@ cd minimap2 && make
 ./minimap2 -ax map-pb ref.fa pacbio.fq.gz > aln.sam       # PacBio genomic reads
 ./minimap2 -ax map-ont ref.fa ont.fq.gz > aln.sam         # Oxford Nanopore genomic reads
 ./minimap2 -ax sr ref.fa read1.fa read2.fa > aln.sam      # short genomic paired-end reads
-./minimap2 -ax splice ref.fa rna-reads.fa > aln.sam       # spliced long reads
-./minimap2 -ax splice -k14 -uf ref.fa reads.fa > aln.sam  # Nanopore Direct RNA-seq
+./minimap2 -ax splice ref.fa rna-reads.fa > aln.sam       # spliced long reads (strand unknown)
+./minimap2 -ax splice -uf -k14 ref.fa reads.fa > aln.sam  # noisy Nanopore Direct RNA-seq
+./minimap2 -ax splice -uf -C5 ref.fa query.fa > aln.sam   # Final PacBio Iso-seq or traditional cDNA
 ./minimap2 -cx asm5 asm1.fa asm2.fa > aln.paf             # intra-species asm-to-asm alignment
 ./minimap2 -x ava-pb reads.fa reads.fa > overlaps.paf     # PacBio read overlap
 ./minimap2 -x ava-ont reads.fa reads.fa > overlaps.paf    # Nanopore read overlap
@@ -69,8 +70,8 @@ Detailed evaluations are available from the [minimap2 paper][doi] or the
 Minimap2 is optimized for x86-64 CPUs. You can acquire precompiled binaries from
 the [release page][release] with:
 ```sh
-curl -L https://github.com/lh3/minimap2/releases/download/v2.11/minimap2-2.11_x64-linux.tar.bz2 | tar -jxvf -
-./minimap2-2.11_x64-linux/minimap2
+curl -L https://github.com/lh3/minimap2/releases/download/v2.12/minimap2-2.12_x64-linux.tar.bz2 | tar -jxvf -
+./minimap2-2.12_x64-linux/minimap2
 ```
 If you want to compile from the source, you need to have a C compiler, GNU make
 and zlib development files installed. Then type `make` in the source code

align.c

@@ -563,11 +563,12 @@ static void mm_align1(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int
 		re = (int32_t)a[as1+cnt1-1].x + 1;
 		qe = (int32_t)a[as1+cnt1-1].y + 1;
 	} else {
-		if (is_splice) {
-			mm_fix_bad_ends_splice(km, opt, mi, r, mat, qlen, qseq0, a, &as1, &cnt1);
-		} else {
-			mm_fix_bad_ends(r, a, opt->bw, opt->min_chain_score * 2, &as1, &cnt1);
-		}
+		if (!(opt->flag & MM_F_NO_END_FLT)) {
+			if (is_splice)
+				mm_fix_bad_ends_splice(km, opt, mi, r, mat, qlen, qseq0, a, &as1, &cnt1);
+			else
+				mm_fix_bad_ends(r, a, opt->bw, opt->min_chain_score * 2, &as1, &cnt1);
+		} else as1 = r->as, cnt1 = r->cnt;
 		mm_filter_bad_seeds(km, as1, cnt1, a, 10, 40, opt->max_gap>>1, 10);
 		mm_filter_bad_seeds_alt(km, as1, cnt1, a, 30, opt->max_gap>>1);
 		mm_adjust_minier(mi, qseq0, &a[as1], &rs, &qs);

cookbook.md

@@ -24,18 +24,18 @@
 
 This cookbook walks you through a variety of applications of minimap2 and its
 companion script `paftools.js`. All data here are freely available from the
-minimap2 release page at version tag [v2.11][v2.11]. Some examples only work
-with v2.11 or later.
+minimap2 release page at version tag [v2.12][v2.12]. Some examples only work
+with v2.10 or later.
 
 To acquire the data used in this cookbook and to install minimap2 and paftools,
 please follow the command lines below:
 ```sh
 # install minimap2 executables
-curl -L https://github.com/lh3/minimap2/releases/download/v2.11/minimap2-2.11_x64-linux.tar.bz2 | tar jxf -
-cp minimap2-2.11_x64-linux/{minimap2,k8,paftools.js} .  # copy executables
+curl -L https://github.com/lh3/minimap2/releases/download/v2.12/minimap2-2.12_x64-linux.tar.bz2 | tar jxf -
+cp minimap2-2.12_x64-linux/{minimap2,k8,paftools.js} .  # copy executables
 export PATH="$PATH:"`pwd`                               # put the current directory on PATH
 # download example datasets
-curl -L https://github.com/lh3/minimap2/releases/download/v2.11/cookbook-data.tgz | tar zxf -
+curl -L https://github.com/lh3/minimap2/releases/download/v2.10/cookbook-data.tgz | tar zxf -
 ```
 
 ## <a name="map-reads"></a>Mapping Genomic Reads
@@ -240,4 +240,4 @@ with `-x ava-pb` (99% vs 93% with `-x ava-ont`).
 [pbsim]: https://github.com/pfaucon/PBSIM-PacBio-Simulator
 [mason2]: https://github.com/seqan/seqan/tree/master/apps/mason2
 [paf]: https://github.com/lh3/miniasm/blob/master/PAF.md
-[v2.11]: https://github.com/lh3/minimap2/releases/tag/v2.11
+[v2.12]: https://github.com/lh3/minimap2/releases/tag/v2.12

format.c

@@ -133,10 +133,10 @@ void mm_write_sam_hdr(const mm_idx_t *idx, const char *rg, const char *ver, int
 	free(str.s);
 }
 
-static void write_cs_core(kstring_t *s, const uint8_t *tseq, const uint8_t *qseq, const mm_reg1_t *r, char *tmp, int no_iden)
+static void write_cs_core(kstring_t *s, const uint8_t *tseq, const uint8_t *qseq, const mm_reg1_t *r, char *tmp, int no_iden, int write_tag)
 {
 	int i, q_off, t_off;
-	mm_sprintf_lite(s, "\tcs:Z:");
+	if (write_tag) mm_sprintf_lite(s, "\tcs:Z:");
 	for (i = q_off = t_off = 0; i < (int)r->p->n_cigar; ++i) {
 		int j, op = r->p->cigar[i]&0xf, len = r->p->cigar[i]>>4;
 		assert(op >= 0 && op <= 3);
@@ -181,10 +181,10 @@ static void write_cs_core(kstring_t *s, const uint8_t *tseq, const uint8_t *qseq
 	assert(t_off == r->re - r->rs && q_off == r->qe - r->qs);
 }
 
-static void write_MD_core(kstring_t *s, const uint8_t *tseq, const uint8_t *qseq, const mm_reg1_t *r, char *tmp)
+static void write_MD_core(kstring_t *s, const uint8_t *tseq, const uint8_t *qseq, const mm_reg1_t *r, char *tmp, int write_tag)
 {
 	int i, q_off, t_off, l_MD = 0;
-	mm_sprintf_lite(s, "\tMD:Z:");
+	if (write_tag) mm_sprintf_lite(s, "\tMD:Z:");
 	for (i = q_off = t_off = 0; i < (int)r->p->n_cigar; ++i) {
 		int j, op = r->p->cigar[i]&0xf, len = r->p->cigar[i]>>4;
 		assert(op >= 0 && op <= 2); // introns (aka reference skips) are not supported
@@ -210,7 +210,7 @@ static void write_MD_core(kstring_t *s, const uint8_t *tseq, const uint8_t *qseq
 	assert(t_off == r->re - r->rs && q_off == r->qe - r->qs);
 }
 
-static void write_cs_or_MD(void *km, kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, const mm_reg1_t *r, int no_iden, int is_MD)
+static void write_cs_or_MD(void *km, kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, const mm_reg1_t *r, int no_iden, int is_MD, int write_tag)
 {
 	extern unsigned char seq_nt4_table[256];
 	int i;
@@ -230,11 +230,34 @@ static void write_cs_or_MD(void *km, kstring_t *s, const mm_idx_t *mi, const mm_
 			qseq[r->qe - i - 1] = c >= 4? 4 : 3 - c;
 		}
 	}
-	if (is_MD) write_MD_core(s, tseq, qseq, r, tmp);
-	else write_cs_core(s, tseq, qseq, r, tmp, no_iden);
+	if (is_MD) write_MD_core(s, tseq, qseq, r, tmp, write_tag);
+	else write_cs_core(s, tseq, qseq, r, tmp, no_iden, write_tag);
 	kfree(km, qseq); kfree(km, tseq); kfree(km, tmp);
 }
 
+int mm_gen_cs_or_MD(void *km, char **buf, int *max_len, const mm_idx_t *mi, const mm_reg1_t *r, const char *seq, int is_MD, int no_iden)
+{
+	mm_bseq1_t t;
+	kstring_t str;
+	str.s = *buf, str.l = 0, str.m = *max_len;
+	t.l_seq = strlen(seq);
+	t.seq = (char*)seq;
+	write_cs_or_MD(km, &str, mi, &t, r, no_iden, is_MD, 0);
+	*max_len = str.m;
+	*buf = str.s;
+	return str.l;
+}
+
+int mm_gen_cs(void *km, char **buf, int *max_len, const mm_idx_t *mi, const mm_reg1_t *r, const char *seq, int no_iden)
+{
+	return mm_gen_cs_or_MD(km, buf, max_len, mi, r, seq, 0, no_iden);
+}
+
+int mm_gen_MD(void *km, char **buf, int *max_len, const mm_idx_t *mi, const mm_reg1_t *r, const char *seq)
+{
+	return mm_gen_cs_or_MD(km, buf, max_len, mi, r, seq, 1, 0);
+}
+
 static inline void write_tags(kstring_t *s, const mm_reg1_t *r)
 {
 	int type;
@@ -277,7 +300,7 @@ void mm_write_paf(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, const m
 			mm_sprintf_lite(s, "%d%c", r->p->cigar[k]>>4, "MIDNSHP=XB"[r->p->cigar[k]&0xf]);
 	}
 	if (r->p && (opt_flag & (MM_F_OUT_CS|MM_F_OUT_MD)))
-		write_cs_or_MD(km, s, mi, t, r, !(opt_flag&MM_F_OUT_CS_LONG), opt_flag&MM_F_OUT_MD);
+		write_cs_or_MD(km, s, mi, t, r, !(opt_flag&MM_F_OUT_CS_LONG), opt_flag&MM_F_OUT_MD, 1);
 	if ((opt_flag & MM_F_COPY_COMMENT) && t->comment)
 		mm_sprintf_lite(s, "\t%s", t->comment);
 }
@@ -476,7 +499,7 @@ void mm_write_sam2(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, int se
 			}
 		}
 		if (r->p && (opt_flag & (MM_F_OUT_CS|MM_F_OUT_MD)))
-			write_cs_or_MD(km, s, mi, t, r, !(opt_flag&MM_F_OUT_CS_LONG), opt_flag&MM_F_OUT_MD);
+			write_cs_or_MD(km, s, mi, t, r, !(opt_flag&MM_F_OUT_CS_LONG), opt_flag&MM_F_OUT_MD, 1);
 		if (cigar_in_tag)
 			write_sam_cigar(s, flag, 1, t->l_seq, r, opt_flag);
 	}

index.c

@@ -372,7 +372,9 @@ mm_idx_t *mm_idx_str(int w, int k, int is_hpc, int bucket_bits, int n, const cha
 	uint64_t sum_len = 0;
 	mm128_v a = {0,0,0};
 	mm_idx_t *mi;
+	khash_t(str) *h;
 	int i, flag = 0;
+
 	if (n <= 0) return 0;
 	for (i = 0; i < n; ++i) // get the total length
 		sum_len += strlen(seq[i]);
@@ -383,13 +385,17 @@ mm_idx_t *mm_idx_str(int w, int k, int is_hpc, int bucket_bits, int n, const cha
 	mi->n_seq = n;
 	mi->seq = (mm_idx_seq_t*)kcalloc(mi->km, n, sizeof(mm_idx_seq_t)); // ->seq is allocated from km
 	mi->S = (uint32_t*)calloc((sum_len + 7) / 8, 4);
+	mi->h = h = kh_init(str);
 	for (i = 0, sum_len = 0; i < n; ++i) {
 		const char *s = seq[i];
 		mm_idx_seq_t *p = &mi->seq[i];
 		uint32_t j;
 		if (name && name[i]) {
+			int absent;
 			p->name = (char*)kmalloc(mi->km, strlen(name[i]) + 1);
 			strcpy(p->name, name[i]);
+			kh_put(str, h, p->name, &absent);
+			assert(absent);
 		}
 		p->offset = sum_len;
 		p->len = strlen(s);
@@ -512,14 +518,19 @@ mm_idx_t *mm_idx_load(FILE *fp)
 int64_t mm_idx_is_idx(const char *fn)
 {
 	int fd, is_idx = 0;
-	off_t ret, off_end;
+	int64_t ret, off_end;
 	char magic[4];
 
 	if (strcmp(fn, "-") == 0) return 0; // read from pipe; not an index
 	fd = open(fn, O_RDONLY);
 	if (fd < 0) return -1; // error
+#ifdef WIN32
+	if ((off_end = _lseeki64(fd, 0, SEEK_END)) >= 4) {
+		_lseeki64(fd, 0, SEEK_SET);
+#else
 	if ((off_end = lseek(fd, 0, SEEK_END)) >= 4) {
 		lseek(fd, 0, SEEK_SET);
+#endif // WIN32
 		ret = read(fd, magic, 4);
 		if (ret == 4 && strncmp(magic, MM_IDX_MAGIC, 4) == 0)
 			is_idx = 1;

main.c

@@ -10,7 +10,7 @@
 #include "getopt.h"
 #endif
 
-#define MM_VERSION "2.11-r815-dirty"
+#define MM_VERSION "2.12-r827"
 
 #ifdef __linux__
 #include <sys/resource.h>
@@ -62,6 +62,7 @@ static struct option long_options[] = {
 	{ "eqx",            no_argument,       0, 0 },   // 32
 	{ "paf-no-hit",     no_argument,       0, 0 },   // 33
 	{ "split-prefix",   required_argument, 0, 0 },   // 34
+	{ "no-end-flt",     no_argument,       0, 0 },   // 35
 	{ "help",           no_argument,       0, 'h' },
 	{ "max-intron-len", required_argument, 0, 'G' },
 	{ "version",        no_argument,       0, 'V' },
@@ -183,6 +184,7 @@ int main(int argc, char *argv[])
 		else if (c == 0 && long_idx ==32) opt.flag |= MM_F_EQX; // --eqx
 		else if (c == 0 && long_idx ==33) opt.flag |= MM_F_PAF_NO_HIT; // --paf-no-hit
 		else if (c == 0 && long_idx ==34) opt.split_prefix = optarg; // --split-prefix
+		else if (c == 0 && long_idx ==35) opt.flag |= MM_F_NO_END_FLT; // --no-end-flt
 		else if (c == 0 && long_idx == 14) { // --frag
 			yes_or_no(&opt, MM_F_FRAG_MODE, long_idx, optarg, 1);
 		} else if (c == 0 && long_idx == 15) { // --secondary

map.c

@@ -29,6 +29,11 @@ void mm_tbuf_destroy(mm_tbuf_t *b)
 	free(b);
 }
 
+void *mm_tbuf_get_km(mm_tbuf_t *b)
+{
+	return b->km;
+}
+
 static int mm_dust_minier(void *km, int n, mm128_t *a, int l_seq, const char *seq, int sdust_thres)
 {
 	int n_dreg, j, k, u = 0;
@@ -596,6 +601,7 @@ static void *worker_pipeline(void *shared, int step, void *in)
 				free(s->reg[i]);
 				free(s->seq[i].seq); free(s->seq[i].name);
 				if (s->seq[i].qual) free(s->seq[i].qual);
+				if (s->seq[i].comment) free(s->seq[i].comment);
 			}
 		}
 		free(s->reg); free(s->n_reg); free(s->seq); // seg_off, n_seg, rep_len and frag_gap were allocated with reg; no memory leak here
@@ -682,7 +688,7 @@ int mm_split_merge(int n_segs, const char **fn, const mm_mapopt_t *opt, int n_sp
 	for (pl.rid_shift[0] = 0, i = 1; i < n_split_idx; ++i)
 		pl.rid_shift[i] += pl.rid_shift[i - 1];
 	if (opt->flag & MM_F_OUT_SAM)
-		for (i = 0; i < pl.mi->n_seq; ++i)
+		for (i = 0; i < (int32_t)pl.mi->n_seq; ++i)
 			printf("@SQ\tSN:%s\tLN:%d\n", pl.mi->seq[i].name, pl.mi->seq[i].len);
 
 	kt_pipeline(2, worker_pipeline, &pl, 3);

minimap.h

@@ -33,6 +33,7 @@
 #define MM_F_COPY_COMMENT  0x2000000
 #define MM_F_EQX           0x4000000 // use =/X instead of M
 #define MM_F_PAF_NO_HIT    0x8000000 // output unmapped reads to PAF
+#define MM_F_NO_END_FLT    0x10000000
 
 #define MM_I_HPC          0x1
 #define MM_I_NO_SEQ       0x2
@@ -301,6 +302,8 @@ mm_tbuf_t *mm_tbuf_init(void);
  */
 void mm_tbuf_destroy(mm_tbuf_t *b);
 
+void *mm_tbuf_get_km(mm_tbuf_t *b);
+
 /**
  * Align a query sequence against an index
  *
@@ -337,6 +340,22 @@ int mm_map_file(const mm_idx_t *idx, const char *fn, const mm_mapopt_t *opt, int
 
 int mm_map_file_frag(const mm_idx_t *idx, int n_segs, const char **fn, const mm_mapopt_t *opt, int n_threads);
 
+/**
+ * Generate the cs tag (new in 2.12)
+ *
+ * @param km         memory blocks; set to NULL if unsure
+ * @param buf        buffer to write the cs/MD tag; typicall NULL on the first call
+ * @param max_len    max length of the buffer; typically set to 0 on the first call
+ * @param mi         index
+ * @param r          alignment
+ * @param seq        query sequence
+ * @param no_iden    true to use : instead of =
+ *
+ * @return the length of cs
+ */
+int mm_gen_cs(void *km, char **buf, int *max_len, const mm_idx_t *mi, const mm_reg1_t *r, const char *seq, int no_iden);
+int mm_gen_MD(void *km, char **buf, int *max_len, const mm_idx_t *mi, const mm_reg1_t *r, const char *seq);
+
 // query sequence name and sequence in the minimap2 index
 int mm_idx_index_name(mm_idx_t *mi);
 int mm_idx_name2id(const mm_idx_t *mi, const char *name);

minimap2.1

@@ -1,4 +1,4 @@
-.TH minimap2 1 "20 June 2018" "minimap2-2.11 (r797)" "Bioinformatics tools"
+.TH minimap2 1 "6 August 2018" "minimap2-2.12 (r827)" "Bioinformatics tools"
 .SH NAME
 .PP
 minimap2 - mapping and alignment between collections of DNA sequences
@@ -360,6 +360,10 @@ the length of the terminal gap in the chain. This option is only effective
 with
 .BR --splice .
 It helps to avoid tiny terminal exons. [6]
+.TP
+.B --no-end-flt
+Don't filter seeds towards the ends of chains before performing base-level
+alignment.
 .SS Input/output options
 .TP 10
 .B -a

options.c

@@ -132,6 +132,11 @@ int mm_set_opt(const char *preset, mm_idxopt_t *io, mm_mapopt_t *mo)
 
 int mm_check_opt(const mm_idxopt_t *io, const mm_mapopt_t *mo)
 {
+	if (mo->split_prefix && (mo->flag & (MM_F_OUT_CS|MM_F_OUT_MD))) {
+		if (mm_verbose >= 1)
+			fprintf(stderr, "[ERROR]\033[1;31m --cs or --MD doesn't work with --split-prefix\033[0m\n");
+		return -6;
+	}
 	if (io->k <= 0 || io->w <= 0) {
 		if (mm_verbose >= 1)
 			fprintf(stderr, "[ERROR]\033[1;31m -k and -w must be positive\033[0m\n");

python/README.rst

@@ -43,20 +43,24 @@ The following Python script demonstrates the key functionality of mappy:
 APIs
 ----
 
-Mappy implements two classes and one global function.
+Mappy implements two classes and two global function.
 
 Class mappy.Aligner
 ~~~~~~~~~~~~~~~~~~~
 
 .. code:: python
 
-	mappy.Aligner(fn_idx_in, preset=None, ...)
+	mappy.Aligner(fn_idx_in=None, preset=None, ...)
 
 This constructor accepts the following arguments:
 
 * **fn_idx_in**: index or sequence file name. Minimap2 automatically tests the
   file type. If a sequence file is provided, minimap2 builds an index. The
-  sequence file can be optionally gzip'd.
+  sequence file can be optionally gzip'd. This option has no effect if **seq**
+  is set.
+
+* **seq**: a single sequence to index. The sequence name will be set to
+  :code:`N/A`.
 
 * **preset**: minimap2 preset. Currently, minimap2 supports the following
   presets: **sr** for single-end short reads; **map-pb** for PacBio
@@ -79,17 +83,28 @@ This constructor accepts the following arguments:
 
 * **n_threads**: number of indexing threads; 3 by default
 
-* **fn_idx_out**: name of file to which the index is written
+* **extra_flags**: additional flags defined in minimap.h
+
+* **fn_idx_out**: name of file to which the index is written. This parameter
+  has no effect if **seq** is set.
+
+* **scoring**: scoring system. It is a tuple/list consisting of 4, 6 or 7
+  positive integers. The first 4 elements specify match scoring, mismatch
+  penalty, gap open and gap extension penalty. The 5th and 6th elements, if
+  present, set long-gap open and long-gap extension penalty. The 7th sets a
+  mismatch penalty involving ambiguous bases.
 
 .. code:: python
 
-	mappy.Aligner.map(seq, seq2=None)
+	mappy.Aligner.map(seq, seq2=None, cs=False, MD=False)
 
 This method aligns :code:`seq` against the index. It is a generator, *yielding*
 a series of :code:`mappy.Alignment` objects. If :code:`seq2` is present, mappy
 performs paired-end alignment, assuming the two ends are in the FR orientation.
 Alignments of the two ends can be distinguished by the :code:`read_num` field
-(see Class mappy.Alignment below).
+(see Class mappy.Alignment below). Argument :code:`cs` asks mappy to generate
+the :code:`cs` tag; :code:`MD` is similar. These two arguments might slightly
+degrade performance and are not enabled by default.
 
 .. code:: python
 
@@ -139,6 +154,11 @@ properties:
 * **cigar**: CIGAR returned as an array of shape :code:`(n_cigar,2)`. The two
   numbers give the length and the operator of each CIGAR operation.
 
+* **MD**: the :code:`MD` tag as in the SAM format. It is an empty string unless
+  the :code:`MD` argument is applied when calling :code:`mappy.Aligner.map()`.
+
+* **cs**: the :code:`cs` tag.
+
 An :code:`Alignment` object can be converted to a string with :code:`str()` in
 the following format:
 

python/cmappy.h

@@ -126,8 +126,8 @@ static char *mappy_fetch_seq(const mm_idx_t *mi, const char *name, int st, int e
 	*len = 0;
 	rid = mm_idx_name2id(mi, name);
 	if (rid < 0) return 0;
-	if (st >= mi->seq[rid].len || st >= en) return 0;
-	if (en < 0 || en > mi->seq[rid].len)
+	if ((uint32_t)st >= mi->seq[rid].len || st >= en) return 0;
+	if (en < 0 || (uint32_t)en > mi->seq[rid].len)
 		en = mi->seq[rid].len;
 	s = (char*)malloc(en - st + 1);
 	*len = mm_idx_getseq(mi, rid, st, en, (uint8_t*)s);
@@ -137,4 +137,16 @@ static char *mappy_fetch_seq(const mm_idx_t *mi, const char *name, int st, int e
 	return s;
 }
 
+static mm_idx_t *mappy_idx_seq(int w, int k, int is_hpc, int bucket_bits, const char *seq, int len)
+{
+	const char *fake_name = "N/A";
+	char *s;
+	mm_idx_t *mi;
+	s = (char*)calloc(len + 1, 1);
+	memcpy(s, seq, len);
+	mi = mm_idx_str(w, k, is_hpc, bucket_bits, 1, (const char**)&s, (const char**)&fake_name);
+	free(s);
+	return mi;
+}
+
 #endif

python/cmappy.pxd

@@ -40,6 +40,7 @@ cdef extern from "minimap.h":
 		int32_t mid_occ
 		int32_t max_occ
 		int mini_batch_size
+		const char *split_prefix
 
 	int mm_set_opt(char *preset, mm_idxopt_t *io, mm_mapopt_t *mo)
 	int mm_verbose
@@ -86,6 +87,9 @@ cdef extern from "minimap.h":
 
 	mm_tbuf_t *mm_tbuf_init()
 	void mm_tbuf_destroy(mm_tbuf_t *b)
+	void *mm_tbuf_get_km(mm_tbuf_t *b)
+	int mm_gen_cs(void *km, char **buf, int *max_len, const mm_idx_t *mi, const mm_reg1_t *r, const char *seq, int no_iden)
+	int mm_gen_MD(void *km, char **buf, int *max_len, const mm_idx_t *mi, const mm_reg1_t *r, const char *seq)
 
 #
 # Helper header (because it is hard to expose mm_reg1_t with Cython)
@@ -106,6 +110,7 @@ cdef extern from "cmappy.h":
 	void mm_free_reg1(mm_reg1_t *r)
 	mm_reg1_t *mm_map_aux(const mm_idx_t *mi, const char *seq1, const char *seq2, int *n_regs, mm_tbuf_t *b, const mm_mapopt_t *opt)
 	char *mappy_fetch_seq(const mm_idx_t *mi, const char *name, int st, int en, int *l)
+	mm_idx_t *mappy_idx_seq(int w, int k, int is_hpc, int bucket_bits, const char *seq, int l)
 
 	ctypedef struct kstring_t:
 		unsigned l, m

python/mappy.pyx

@@ -14,9 +14,9 @@ cdef class Alignment:
 	cdef int8_t _strand, _trans_strand
 	cdef uint8_t _mapq, _is_primary
 	cdef int _seg_id
-	cdef _ctg, _cigar # these are python objects
+	cdef _ctg, _cigar, _cs, _MD # these are python objects
 
-	def __cinit__(self, ctg, cl, cs, ce, strand, qs, qe, mapq, cigar, is_primary, mlen, blen, NM, trans_strand, seg_id):
+	def __cinit__(self, ctg, cl, cs, ce, strand, qs, qe, mapq, cigar, is_primary, mlen, blen, NM, trans_strand, seg_id, cs_str, MD_str):
 		self._ctg = ctg if isinstance(ctg, str) else ctg.decode()
 		self._ctg_len, self._r_st, self._r_en = cl, cs, ce
 		self._strand, self._q_st, self._q_en = strand, qs, qe
@@ -26,6 +26,8 @@ cdef class Alignment:
 		self._is_primary = is_primary
 		self._trans_strand = trans_strand
 		self._seg_id = seg_id
+		self._cs = cs_str
+		self._MD = MD_str
 
 	@property
 	def ctg(self): return self._ctg
@@ -72,6 +74,12 @@ cdef class Alignment:
 	@property
 	def read_num(self): return self._seg_id + 1
 
+	@property
+	def cs(self): return self._cs
+
+	@property
+	def MD(self): return self._MD
+
 	@property
 	def cigar_str(self):
 		return "".join(map(lambda x: str(x[0]) + 'MIDNSH'[x[1]], self._cigar))
@@ -85,8 +93,10 @@ cdef class Alignment:
 		if self._trans_strand > 0: ts = 'ts:A:+'
 		elif self._trans_strand < 0: ts = 'ts:A:-'
 		else: ts = 'ts:A:.'
-		return "\t".join([str(self._q_st), str(self._q_en), strand, self._ctg, str(self._ctg_len), str(self._r_st), str(self._r_en),
-				str(self._mlen), str(self._blen), str(self._mapq), tp, ts, "cg:Z:" + self.cigar_str])
+		a = [str(self._q_st), str(self._q_en), strand, self._ctg, str(self._ctg_len), str(self._r_st), str(self._r_en),
+			str(self._mlen), str(self._blen), str(self._mapq), tp, ts, "cg:Z:" + self.cigar_str]
+		if self._cs != "": a.append("cs:Z:" + self._cs)
+		return "\t".join(a)
 
 cdef class ThreadBuffer:
 	cdef cmappy.mm_tbuf_t *_b
@@ -102,7 +112,7 @@ cdef class Aligner:
 	cdef cmappy.mm_idxopt_t idx_opt
 	cdef cmappy.mm_mapopt_t map_opt
 
-	def __cinit__(self, fn_idx_in, preset=None, k=None, w=None, min_cnt=None, min_chain_score=None, min_dp_score=None, bw=None, best_n=None, n_threads=3, fn_idx_out=None, max_frag_len=None):
+	def __cinit__(self, fn_idx_in=None, preset=None, k=None, w=None, min_cnt=None, min_chain_score=None, min_dp_score=None, bw=None, best_n=None, n_threads=3, fn_idx_out=None, max_frag_len=None, extra_flags=None, seq=None, scoring=None):
 		cmappy.mm_set_opt(NULL, &self.idx_opt, &self.map_opt) # set the default options
 		if preset is not None:
 			cmappy.mm_set_opt(str.encode(preset), &self.idx_opt, &self.map_opt) # apply preset
@@ -116,17 +126,32 @@ cdef class Aligner:
 		if bw is not None: self.map_opt.bw = bw
 		if best_n is not None: self.map_opt.best_n = best_n
 		if max_frag_len is not None: self.map_opt.max_frag_len = max_frag_len
+		if extra_flags is not None: self.map_opt.flag |= extra_flags
+		if scoring is not None and len(scoring) >= 4:
+			self.map_opt.a, self.map_opt.b = scoring[0], scoring[1]
+			self.map_opt.q, self.map_opt.e = scoring[2], scoring[3]
+			self.map_opt.q2, self.map_opt.e2 = self.map_opt.q, self.map_opt.e
+			if len(scoring) >= 6:
+				self.map_opt.q2, self.map_opt.e2 = scoring[4], scoring[5]
+				if len(scoring) >= 7:
+					self.map_opt.sc_ambi = scoring[6]
 
 		cdef cmappy.mm_idx_reader_t *r;
-		if fn_idx_out is None:
-			r = cmappy.mm_idx_reader_open(str.encode(fn_idx_in), &self.idx_opt, NULL)
+
+		if seq is None:
+			if fn_idx_out is None:
+				r = cmappy.mm_idx_reader_open(str.encode(fn_idx_in), &self.idx_opt, NULL)
+			else:
+				r = cmappy.mm_idx_reader_open(str.encode(fn_idx_in), &self.idx_opt, fn_idx_out)
+			if r is not NULL:
+				self._idx = cmappy.mm_idx_reader_read(r, n_threads) # NB: ONLY read the first part
+				cmappy.mm_idx_reader_close(r)
+				cmappy.mm_mapopt_update(&self.map_opt, self._idx)
+				cmappy.mm_idx_index_name(self._idx)
 		else:
-			r = cmappy.mm_idx_reader_open(str.encode(fn_idx_in), &self.idx_opt, fn_idx_out)
-		if r is not NULL:
-			self._idx = cmappy.mm_idx_reader_read(r, n_threads) # NB: ONLY read the first part
-			cmappy.mm_idx_reader_close(r)
+			self._idx = cmappy.mappy_idx_seq(self.idx_opt.w, self.idx_opt.k, self.idx_opt.flag&1, self.idx_opt.bucket_bits, str.encode(seq), len(seq))
 			cmappy.mm_mapopt_update(&self.map_opt, self._idx)
-			cmappy.mm_idx_index_name(self._idx)
+			self.map_opt.mid_occ = 1000 # don't filter high-occ seeds
 
 	def __dealloc__(self):
 		if self._idx is not NULL:
@@ -135,18 +160,24 @@ cdef class Aligner:
 	def __bool__(self):
 		return (self._idx != NULL)
 
-	def map(self, seq, seq2=None, buf=None, max_frag_len=None):
+	def map(self, seq, seq2=None, buf=None, cs=False, MD=False, max_frag_len=None, extra_flags=None):
 		cdef cmappy.mm_reg1_t *regs
 		cdef cmappy.mm_hitpy_t h
 		cdef ThreadBuffer b
 		cdef int n_regs
+		cdef char *cs_str = NULL
+		cdef int l_cs_str, m_cs_str = 0
+		cdef void *km
 		cdef cmappy.mm_mapopt_t map_opt
+
 		map_opt = self.map_opt
 		if max_frag_len is not None: map_opt.max_frag_len = max_frag_len
+		if extra_flags is not None: map_opt.flag |= extra_flags
 
 		if self._idx is NULL: return None
 		if buf is None: b = ThreadBuffer()
 		else: b = buf
+		km = cmappy.mm_tbuf_get_km(b._b)
 
 		_seq = seq if isinstance(seq, bytes) else seq.encode()
 		if seq2 is None:
@@ -157,13 +188,21 @@ cdef class Aligner:
 
 		for i in range(n_regs):
 			cmappy.mm_reg2hitpy(self._idx, &regs[i], &h)
-			cigar = []
-			for k in range(h.n_cigar32):
+			cigar, _cs, _MD = [], '', ''
+			for k in range(h.n_cigar32): # convert the 32-bit CIGAR encoding to Python array
 				c = h.cigar32[k]
 				cigar.append([c>>4, c&0xf])
-			yield Alignment(h.ctg, h.ctg_len, h.ctg_start, h.ctg_end, h.strand, h.qry_start, h.qry_end, h.mapq, cigar, h.is_primary, h.mlen, h.blen, h.NM, h.trans_strand, h.seg_id)
+			if cs or MD: # generate the cs and/or the MD tag, if requested
+				if cs:
+					l_cs_str = cmappy.mm_gen_cs(km, &cs_str, &m_cs_str, self._idx, &regs[i], _seq, 1)
+					_cs = cs_str[:l_cs_str] if isinstance(cs_str, str) else cs_str[:l_cs_str].decode()
+				if MD:
+					l_cs_str = cmappy.mm_gen_MD(km, &cs_str, &m_cs_str, self._idx, &regs[i], _seq)
+					_MD = cs_str[:l_cs_str] if isinstance(cs_str, str) else cs_str[:l_cs_str].decode()
+			yield Alignment(h.ctg, h.ctg_len, h.ctg_start, h.ctg_end, h.strand, h.qry_start, h.qry_end, h.mapq, cigar, h.is_primary, h.mlen, h.blen, h.NM, h.trans_strand, h.seg_id, _cs, _MD)
 			cmappy.mm_free_reg1(&regs[i])
 		free(regs)
+		free(cs_str)
 
 	def seq(self, str name, int start=0, int end=0x7fffffff):
 		cdef int l

python/minimap2.py

@@ -4,7 +4,7 @@ import sys, getopt
 import mappy as mp
 
 def main(argv):
-	opts, args = getopt.getopt(argv[1:], "x:n:m:k:w:r:")
+	opts, args = getopt.getopt(argv[1:], "x:n:m:k:w:r:c")
 	if len(args) < 2:
 		print("Usage: minimap2.py [options] <ref.fa>|<ref.mmi> <query.fq>")
 		print("Options:")
@@ -14,9 +14,10 @@ def main(argv):
 		print("  -k INT      k-mer length")
 		print("  -w INT      minimizer window length")
 		print("  -r INT      band width")
+		print("  -c          output the cs tag")
 		sys.exit(1)
 
-	preset, min_cnt, min_sc, k, w, bw = None, None, None, None, None, None
+	preset, min_cnt, min_sc, k, w, bw, out_cs = None, None, None, None, None, None, False
 	for opt, arg in opts:
 		if opt == '-x': preset = arg
 		elif opt == '-n': min_cnt = int(arg)
@@ -24,11 +25,12 @@ def main(argv):
 		elif opt == '-r': bw = int(arg)
 		elif opt == '-k': k = int(arg)
 		elif opt == '-w': w = int(arg)
+		elif opt == '-c': out_cs = True
 
 	a = mp.Aligner(args[0], preset=preset, min_cnt=min_cnt, min_chain_score=min_sc, k=k, w=w, bw=bw)
 	if not a: raise Exception("ERROR: failed to load/build index file '{}'".format(args[0]))
 	for name, seq, qual in mp.fastx_read(args[1]): # read one sequence
-		for h in a.map(seq): # traverse hits
+		for h in a.map(seq, cs=out_cs): # traverse hits
 			print('{}\t{}\t{}'.format(name, len(seq), h))
 
 if __name__ == "__main__":

setup.py

@@ -14,16 +14,26 @@ else: # with Cython
 	module_src = 'python/mappy.pyx'
 	cmdclass['build_ext'] = build_ext
 
-import sys
+import sys, platform
+
 sys.path.append('python')
 
+extra_compile_args = ['-DHAVE_KALLOC']
+include_dirs = ["."]
+
+if platform.machine() in ["aarch64", "arm64"]:
+	include_dirs.append("sse2neon/")
+	extra_compile_args.extend(['-ftree-vectorize', '-DKSW_SSE2_ONLY', '-D__SSE2__'])
+else:
+	extra_compile_args.append('-msse4.1') # WARNING: ancient x86_64 CPUs don't have SSE4
+
 def readme():
-    with open('python/README.rst') as f:
-        return f.read()
+	with open('python/README.rst') as f:
+		return f.read()
 
 setup(
 	name = 'mappy',
-	version = '2.11',
+	version = '2.12',
 	url = 'https://github.com/lh3/minimap2',
 	description = 'Minimap2 python binding',
 	long_description = readme(),
@@ -35,12 +45,12 @@ setup(
     ext_modules = [Extension('mappy',
 		sources = [module_src, 'align.c', 'bseq.c', 'chain.c', 'format.c', 'hit.c', 'index.c', 'pe.c', 'options.c',
 				   'ksw2_extd2_sse.c', 'ksw2_exts2_sse.c', 'ksw2_extz2_sse.c', 'ksw2_ll_sse.c',
-				   'kalloc.c', 'kthread.c', 'map.c', 'misc.c', 'sdust.c', 'sketch.c', 'esterr.c'],
+				   'kalloc.c', 'kthread.c', 'map.c', 'misc.c', 'sdust.c', 'sketch.c', 'esterr.c', 'splitidx.c'],
 		depends = ['minimap.h', 'bseq.h', 'kalloc.h', 'kdq.h', 'khash.h', 'kseq.h', 'ksort.h',
 				   'ksw2.h', 'kthread.h', 'kvec.h', 'mmpriv.h', 'sdust.h',
 				   'python/cmappy.h', 'python/cmappy.pxd'],
-		extra_compile_args = ['-DHAVE_KALLOC', '-msse4.1'], # WARNING: ancient x86_64 CPUs don't have SSE4
-		include_dirs = ['.'],
+		extra_compile_args = extra_compile_args,
+		include_dirs = include_dirs,
 		libraries = ['z', 'm', 'pthread'])],
 	classifiers = [
 		'Development Status :: 5 - Production/Stable',