X-Git-Url: https://git.donarmstrong.com/?p=samtools.git;a=blobdiff_plain;f=bam_sort.c;h=7d00cd1a83321729c1384cf56f066d2c83f65548;hp=f089cdff396695c0cf43d21d09be695463e187c9;hb=307c147168f7154e3755712797078c513e0b242a;hpb=fb6c8f27788dc657f1dc20dae3f90c3e617279d7

diff --git a/bam_sort.c b/bam_sort.c
index f089cdf..7d00cd1 100644
--- a/bam_sort.c
+++ b/bam_sort.c
@@ -1,6 +1,7 @@
 #include <stdlib.h>
 #include <ctype.h>
 #include <assert.h>
+#include <errno.h>
 #include <stdio.h>
 #include <string.h>
 #include <unistd.h>
@@ -9,221 +10,557 @@
 
 static int g_is_by_qname = 0;
 
-static inline int strnum_cmp(const char *a, const char *b)
+static int strnum_cmp(const char *_a, const char *_b)
 {
-	char *pa, *pb;
-	pa = (char*)a; pb = (char*)b;
+	const unsigned char *a = (const unsigned char*)_a, *b = (const unsigned char*)_b;
+	const unsigned char *pa = a, *pb = b;
 	while (*pa && *pb) {
 		if (isdigit(*pa) && isdigit(*pb)) {
-			long ai, bi;
-			ai = strtol(pa, &pa, 10);
-			bi = strtol(pb, &pb, 10);
-			if (ai != bi) return ai<bi? -1 : ai>bi? 1 : 0;
+			while (*pa == '0') ++pa;
+			while (*pb == '0') ++pb;
+			while (isdigit(*pa) && isdigit(*pb) && *pa == *pb) ++pa, ++pb;
+			if (isdigit(*pa) && isdigit(*pb)) {
+				int i = 0;
+				while (isdigit(pa[i]) && isdigit(pb[i])) ++i;
+				return isdigit(pa[i])? 1 : isdigit(pb[i])? -1 : (int)*pa - (int)*pb;
+			} else if (isdigit(*pa)) return 1;
+			else if (isdigit(*pb)) return -1;
+			else if (pa - a != pb - b) return pa - a < pb - b? 1 : -1;
 		} else {
-			if (*pa != *pb) break;
+			if (*pa != *pb) return (int)*pa - (int)*pb;
 			++pa; ++pb;
 		}
 	}
-	if (*pa == *pb)
-		return (pa-a) < (pb-b)? -1 : (pa-a) > (pb-b)? 1 : 0;
-	return *pa<*pb? -1 : *pa>*pb? 1 : 0;
+	return *pa? 1 : *pb? -1 : 0;
 }
 
 #define HEAP_EMPTY 0xffffffffffffffffull
 
 typedef struct {
 	int i;
-	uint64_t pos;
+	uint64_t pos, idx;
 	bam1_t *b;
 } heap1_t;
 
+#define __pos_cmp(a, b) ((a).pos > (b).pos || ((a).pos == (b).pos && ((a).i > (b).i || ((a).i == (b).i && (a).idx > (b).idx))))
+
 static inline int heap_lt(const heap1_t a, const heap1_t b)
 {
 	if (g_is_by_qname) {
-		int t = strnum_cmp(bam1_qname(a.b), bam1_qname(b.b));
-		return (t > 0 || (t == 0 && a.pos > b.pos));
-	} else return (a.pos > b.pos);
+		int t;
+		if (a.b == 0 || b.b == 0) return a.b == 0? 1 : 0;
+		t = strnum_cmp(bam1_qname(a.b), bam1_qname(b.b));
+		return (t > 0 || (t == 0 && (a.b->core.flag&0xc0) > (b.b->core.flag&0xc0)));
+	} else return __pos_cmp(a, b);
 }
 
 KSORT_INIT(heap, heap1_t, heap_lt)
 
-void bam_merge_core(int by_qname, const char *out, int n, char * const *fn)
+static void swap_header_targets(bam_header_t *h1, bam_header_t *h2)
+{
+	bam_header_t t;
+	t.n_targets = h1->n_targets, h1->n_targets = h2->n_targets, h2->n_targets = t.n_targets;
+	t.target_name = h1->target_name, h1->target_name = h2->target_name, h2->target_name = t.target_name;
+	t.target_len = h1->target_len, h1->target_len = h2->target_len, h2->target_len = t.target_len;
+}
+
+static void swap_header_text(bam_header_t *h1, bam_header_t *h2)
+{
+	int tempi;
+	char *temps;
+	tempi = h1->l_text, h1->l_text = h2->l_text, h2->l_text = tempi;
+	temps = h1->text, h1->text = h2->text, h2->text = temps;
+}
+
+#define MERGE_RG     1
+#define MERGE_UNCOMP 2
+#define MERGE_LEVEL1 4
+#define MERGE_FORCE  8
+
+/*!
+  @abstract    Merge multiple sorted BAM.
+  @param  is_by_qname whether to sort by query name
+  @param  out  output BAM file name
+  @param  headers  name of SAM file from which to copy '@' header lines,
+                   or NULL to copy them from the first file to be merged
+  @param  n    number of files to be merged
+  @param  fn   names of files to be merged
+
+  @discussion Padding information may NOT correctly maintained. This
+  function is NOT thread safe.
+ */
+int bam_merge_core2(int by_qname, const char *out, const char *headers, int n, char * const *fn, int flag, const char *reg, int n_threads, int level)
 {
 	bamFile fpout, *fp;
 	heap1_t *heap;
 	bam_header_t *hout = 0;
-	int i, j;
+	bam_header_t *hheaders = NULL;
+	int i, j, *RG_len = 0;
+	uint64_t idx = 0;
+	char **RG = 0, mode[8];
+	bam_iter_t *iter = 0;
+
+	if (headers) {
+		tamFile fpheaders = sam_open(headers);
+		if (fpheaders == 0) {
+			const char *message = strerror(errno);
+			fprintf(stderr, "[bam_merge_core] cannot open '%s': %s\n", headers, message);
+			return -1;
+		}
+		hheaders = sam_header_read(fpheaders);
+		sam_close(fpheaders);
+	}
 
 	g_is_by_qname = by_qname;
 	fp = (bamFile*)calloc(n, sizeof(bamFile));
 	heap = (heap1_t*)calloc(n, sizeof(heap1_t));
+	iter = (bam_iter_t*)calloc(n, sizeof(bam_iter_t));
+	// prepare RG tag
+	if (flag & MERGE_RG) {
+		RG = (char**)calloc(n, sizeof(void*));
+		RG_len = (int*)calloc(n, sizeof(int));
+		for (i = 0; i != n; ++i) {
+			int l = strlen(fn[i]);
+			const char *s = fn[i];
+			if (l > 4 && strcmp(s + l - 4, ".bam") == 0) l -= 4;
+			for (j = l - 1; j >= 0; --j) if (s[j] == '/') break;
+			++j; l -= j;
+			RG[i] = calloc(l + 1, 1);
+			RG_len[i] = l;
+			strncpy(RG[i], s + j, l);
+		}
+	}
+	// read the first
 	for (i = 0; i != n; ++i) {
-		heap1_t *h;
 		bam_header_t *hin;
-		assert(fp[i] = bam_open(fn[i], "r"));
+		fp[i] = bam_open(fn[i], "r");
+		if (fp[i] == 0) {
+			int j;
+			fprintf(stderr, "[bam_merge_core] fail to open file %s\n", fn[i]);
+			for (j = 0; j < i; ++j) bam_close(fp[j]);
+			free(fp); free(heap);
+			// FIXME: possible memory leak
+			return -1;
+		}
 		hin = bam_header_read(fp[i]);
-		if (i == 0) hout = hin;
-		else { // validate multiple baf
-			if (hout->n_targets != hin->n_targets) {
-				fprintf(stderr, "[bam_merge_core] file '%s' has different number of target sequences. Abort!\n", fn[i]);
-				abort();
-			}
-			for (j = 0; j < hout->n_targets; ++j) {
-				if (strcmp(hout->target_name[j], hin->target_name[j]) || hout->target_len[j] != hin->target_len[j]) {
-					fprintf(stderr, "[bam_merge_core] file '%s' has a different target sequence. Abort!\n", fn[i]);
-					abort();
+		if (i == 0) { // the first BAM
+			hout = hin;
+		} else { // validate multiple baf
+			int min_n_targets = hout->n_targets;
+			if (hin->n_targets < min_n_targets) min_n_targets = hin->n_targets;
+
+			for (j = 0; j < min_n_targets; ++j)
+				if (strcmp(hout->target_name[j], hin->target_name[j]) != 0) {
+					fprintf(stderr, "[bam_merge_core] different target sequence name: '%s' != '%s' in file '%s'\n",
+							hout->target_name[j], hin->target_name[j], fn[i]);
+					return -1;
 				}
+
+			// If this input file has additional target reference sequences,
+			// add them to the headers to be output
+			if (hin->n_targets > hout->n_targets) {
+				swap_header_targets(hout, hin);
+				// FIXME Possibly we should also create @SQ text headers
+				// for the newly added reference sequences
 			}
+
 			bam_header_destroy(hin);
 		}
-		h = heap + i;
+	}
+
+	if (hheaders) {
+		// If the text headers to be swapped in include any @SQ headers,
+		// check that they are consistent with the existing binary list
+		// of reference information.
+		if (hheaders->n_targets > 0) {
+			if (hout->n_targets != hheaders->n_targets) {
+				fprintf(stderr, "[bam_merge_core] number of @SQ headers in '%s' differs from number of target sequences\n", headers);
+				if (!reg) return -1;
+			}
+			for (j = 0; j < hout->n_targets; ++j)
+				if (strcmp(hout->target_name[j], hheaders->target_name[j]) != 0) {
+					fprintf(stderr, "[bam_merge_core] @SQ header '%s' in '%s' differs from target sequence\n", hheaders->target_name[j], headers);
+					if (!reg) return -1;
+				}
+		}
+
+		swap_header_text(hout, hheaders);
+		bam_header_destroy(hheaders);
+	}
+
+	if (reg) {
+		int tid, beg, end;
+		if (bam_parse_region(hout, reg, &tid, &beg, &end) < 0) {
+			fprintf(stderr, "[%s] Malformated region string or undefined reference name\n", __func__);
+			return -1;
+		}
+		for (i = 0; i < n; ++i) {
+			bam_index_t *idx;
+			idx = bam_index_load(fn[i]);
+			iter[i] = bam_iter_query(idx, tid, beg, end);
+			bam_index_destroy(idx);
+		}
+	}
+
+	for (i = 0; i < n; ++i) {
+		heap1_t *h = heap + i;
 		h->i = i;
 		h->b = (bam1_t*)calloc(1, sizeof(bam1_t));
-		if (bam_read1(fp[i], h->b) >= 0)
-			h->pos = ((uint64_t)h->b->core.tid<<32) | (uint32_t)h->b->core.pos<<1 | bam1_strand(h->b);
+		if (bam_iter_read(fp[i], iter[i], h->b) >= 0) {
+			h->pos = ((uint64_t)h->b->core.tid<<32) | (uint32_t)((int32_t)h->b->core.pos+1)<<1 | bam1_strand(h->b);
+			h->idx = idx++;
+		}
 		else h->pos = HEAP_EMPTY;
 	}
-	fpout = strcmp(out, "-")? bam_open(out, "w") : bam_dopen(fileno(stdout), "w");
-	assert(fpout);
+	if (flag & MERGE_UNCOMP) level = 0;
+	else if (flag & MERGE_LEVEL1) level = 1;
+	strcpy(mode, "w");
+	if (level >= 0) sprintf(mode + 1, "%d", level < 9? level : 9);
+	if ((fpout = strcmp(out, "-")? bam_open(out, "w") : bam_dopen(fileno(stdout), "w")) == 0) {
+		fprintf(stderr, "[%s] fail to create the output file.\n", __func__);
+		return -1;
+	}
 	bam_header_write(fpout, hout);
 	bam_header_destroy(hout);
+	if (!(flag & MERGE_UNCOMP)) bgzf_mt(fpout, n_threads, 256);
 
 	ks_heapmake(heap, n, heap);
 	while (heap->pos != HEAP_EMPTY) {
 		bam1_t *b = heap->b;
+		if (flag & MERGE_RG) {
+			uint8_t *rg = bam_aux_get(b, "RG");
+			if (rg) bam_aux_del(b, rg);
+			bam_aux_append(b, "RG", 'Z', RG_len[heap->i] + 1, (uint8_t*)RG[heap->i]);
+		}
 		bam_write1_core(fpout, &b->core, b->data_len, b->data);
-		if ((j = bam_read1(fp[heap->i], b)) >= 0)
-			heap->pos = ((uint64_t)b->core.tid<<32) | (uint32_t)b->core.pos<<1 | bam1_strand(b);
-		else if (j == -1) heap->pos = HEAP_EMPTY;
-		else fprintf(stderr, "[bam_merge_core] '%s' is truncated. Continue anyway.\n", fn[heap->i]);
+		if ((j = bam_iter_read(fp[heap->i], iter[heap->i], b)) >= 0) {
+			heap->pos = ((uint64_t)b->core.tid<<32) | (uint32_t)((int)b->core.pos+1)<<1 | bam1_strand(b);
+			heap->idx = idx++;
+		} else if (j == -1) {
+			heap->pos = HEAP_EMPTY;
+			free(heap->b->data); free(heap->b);
+			heap->b = 0;
+		} else fprintf(stderr, "[bam_merge_core] '%s' is truncated. Continue anyway.\n", fn[heap->i]);
 		ks_heapadjust(heap, 0, n, heap);
 	}
 
+	if (flag & MERGE_RG) {
+		for (i = 0; i != n; ++i) free(RG[i]);
+		free(RG); free(RG_len);
+	}
 	for (i = 0; i != n; ++i) {
+		bam_iter_destroy(iter[i]);
 		bam_close(fp[i]);
-		free(heap[i].b->data);
-		free(heap[i].b);
 	}
 	bam_close(fpout);
-	free(fp); free(heap);
+	free(fp); free(heap); free(iter);
+	return 0;
+}
+
+int bam_merge_core(int by_qname, const char *out, const char *headers, int n, char * const *fn, int flag, const char *reg)
+{
+	return bam_merge_core2(by_qname, out, headers, n, fn, flag, reg, 0, -1);
 }
+
 int bam_merge(int argc, char *argv[])
 {
-	int c, is_by_qname = 0;
-	while ((c = getopt(argc, argv, "n")) >= 0) {
+	int c, is_by_qname = 0, flag = 0, ret = 0, n_threads = 0, level = -1;
+	char *fn_headers = NULL, *reg = 0;
+
+	while ((c = getopt(argc, argv, "h:nru1R:f@:l:")) >= 0) {
 		switch (c) {
+		case 'r': flag |= MERGE_RG; break;
+		case 'f': flag |= MERGE_FORCE; break;
+		case 'h': fn_headers = strdup(optarg); break;
 		case 'n': is_by_qname = 1; break;
+		case '1': flag |= MERGE_LEVEL1; break;
+		case 'u': flag |= MERGE_UNCOMP; break;
+		case 'R': reg = strdup(optarg); break;
+		case 'l': level = atoi(optarg); break;
+		case '@': n_threads = atoi(optarg); break;
 		}
 	}
 	if (optind + 2 >= argc) {
-		fprintf(stderr, "Usage: samtools merge [-n] <out.bam> <in1.bam> <in2.bam> [...]\n");
+		fprintf(stderr, "\n");
+		fprintf(stderr, "Usage:   samtools merge [-nr] [-h inh.sam] <out.bam> <in1.bam> <in2.bam> [...]\n\n");
+		fprintf(stderr, "Options: -n       sort by read names\n");
+		fprintf(stderr, "         -r       attach RG tag (inferred from file names)\n");
+		fprintf(stderr, "         -u       uncompressed BAM output\n");
+		fprintf(stderr, "         -f       overwrite the output BAM if exist\n");
+		fprintf(stderr, "         -1       compress level 1\n");
+		fprintf(stderr, "         -l INT   compression level, from 0 to 9 [-1]\n");
+		fprintf(stderr, "         -@ INT   number of BAM compression threads [0]\n");
+		fprintf(stderr, "         -R STR   merge file in the specified region STR [all]\n");
+		fprintf(stderr, "         -h FILE  copy the header in FILE to <out.bam> [in1.bam]\n\n");
+		fprintf(stderr, "Note: Samtools' merge does not reconstruct the @RG dictionary in the header. Users\n");
+		fprintf(stderr, "      must provide the correct header with -h, or uses Picard which properly maintains\n");
+		fprintf(stderr, "      the header dictionary in merging.\n\n");
 		return 1;
 	}
-	bam_merge_core(is_by_qname, argv[optind], argc - optind - 1, argv + optind + 1);
-	return 0;
+	if (!(flag & MERGE_FORCE) && strcmp(argv[optind], "-")) {
+		FILE *fp = fopen(argv[optind], "rb");
+		if (fp != NULL) {
+			fclose(fp);
+			fprintf(stderr, "[%s] File '%s' exists. Please apply '-f' to overwrite. Abort.\n", __func__, argv[optind]);
+			return 1;
+		}
+	}
+	if (bam_merge_core2(is_by_qname, argv[optind], fn_headers, argc - optind - 1, argv + optind + 1, flag, reg, n_threads, level) < 0) ret = 1;
+	free(reg);
+	free(fn_headers);
+	return ret;
 }
 
+/***************
+ * BAM sorting *
+ ***************/
+
+#include <pthread.h>
+
 typedef bam1_t *bam1_p;
 
+static int change_SO(bam_header_t *h, const char *so)
+{
+	char *p, *q, *beg = 0, *end = 0, *newtext;
+	if (h->l_text > 3) {
+		if (strncmp(h->text, "@HD", 3) == 0) {
+			if ((p = strchr(h->text, '\n')) == 0) return -1;
+			*p = '\0';
+			if ((q = strstr(h->text, "\tSO:")) != 0) {
+				*p = '\n'; // change back
+				if (strncmp(q + 4, so, p - q - 4) != 0) {
+					beg = q;
+					for (q += 4; *q != '\n' && *q != '\t'; ++q);
+					end = q;
+				} else return 0; // no need to change
+			} else beg = end = p, *p = '\n';
+		}
+	}
+	if (beg == 0) { // no @HD
+		h->l_text += strlen(so) + 15;
+		newtext = malloc(h->l_text + 1);
+		sprintf(newtext, "@HD\tVN:1.3\tSO:%s\n", so);
+		strcat(newtext, h->text);
+	} else { // has @HD but different or no SO
+		h->l_text = (beg - h->text) + (4 + strlen(so)) + (h->text + h->l_text - end);
+		newtext = malloc(h->l_text + 1);
+		strncpy(newtext, h->text, beg - h->text);
+		sprintf(newtext + (beg - h->text), "\tSO:%s", so);
+		strcat(newtext, end);
+	}
+	free(h->text);
+	h->text = newtext;
+	return 0;
+}
+
 static inline int bam1_lt(const bam1_p a, const bam1_p b)
 {
 	if (g_is_by_qname) {
 		int t = strnum_cmp(bam1_qname(a), bam1_qname(b));
-		return (t < 0 || (t == 0 && (((uint64_t)a->core.tid<<32|a->core.pos) < ((uint64_t)b->core.tid<<32|b->core.pos))));
-	} else return (((uint64_t)a->core.tid<<32|a->core.pos) < ((uint64_t)b->core.tid<<32|b->core.pos));
+		return (t < 0 || (t == 0 && (a->core.flag&0xc0) < (b->core.flag&0xc0)));
+	} else return (((uint64_t)a->core.tid<<32|(a->core.pos+1)<<1|bam1_strand(a)) < ((uint64_t)b->core.tid<<32|(b->core.pos+1)<<1|bam1_strand(b)));
 }
 KSORT_INIT(sort, bam1_p, bam1_lt)
 
-static void sort_blocks(int n, int k, bam1_p *buf, const char *prefix, const bam_header_t *h)
+typedef struct {
+	size_t buf_len;
+	const char *prefix;
+	bam1_p *buf;
+	const bam_header_t *h;
+	int index;
+} worker_t;
+
+static void write_buffer(const char *fn, const char *mode, size_t l, bam1_p *buf, const bam_header_t *h, int n_threads)
 {
-	char *name;
-	int i;
+	size_t i;
 	bamFile fp;
-	ks_mergesort(sort, k, buf, 0);
-	name = (char*)calloc(strlen(prefix) + 20, 1);
-	if (n >= 0) sprintf(name, "%s.%.4d.bam", prefix, n);
-	else sprintf(name, "%s.bam", prefix);
-	assert(fp = bam_open(name, "w"));
-	free(name);
+	fp = strcmp(fn, "-")? bam_open(fn, mode) : bam_dopen(fileno(stdout), mode);
+	if (fp == 0) return;
 	bam_header_write(fp, h);
-	for (i = 0; i < k; ++i)
+	if (n_threads > 1) bgzf_mt(fp, n_threads, 256);
+	for (i = 0; i < l; ++i)
 		bam_write1_core(fp, &buf[i]->core, buf[i]->data_len, buf[i]->data);
 	bam_close(fp);
 }
 
-void bam_sort_core(int is_by_qname, const char *fn, const char *prefix, size_t max_mem)
+static void *worker(void *data)
+{
+	worker_t *w = (worker_t*)data;
+	char *name;
+	ks_mergesort(sort, w->buf_len, w->buf, 0);
+	name = (char*)calloc(strlen(w->prefix) + 20, 1);
+	sprintf(name, "%s.%.4d.bam", w->prefix, w->index);
+	write_buffer(name, "w1", w->buf_len, w->buf, w->h, 0);
+	free(name);
+	return 0;
+}
+
+static int sort_blocks(int n_files, size_t k, bam1_p *buf, const char *prefix, const bam_header_t *h, int n_threads)
 {
-	int n, ret, k, i;
-	size_t mem;
+	int i;
+	size_t rest;
+	bam1_p *b;
+	pthread_t *tid;
+	pthread_attr_t attr;
+	worker_t *w;
+
+	if (n_threads < 1) n_threads = 1;
+	if (k < n_threads * 64) n_threads = 1; // use a single thread if we only sort a small batch of records
+	pthread_attr_init(&attr);
+	pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_JOINABLE);
+	w = calloc(n_threads, sizeof(worker_t));
+	tid = calloc(n_threads, sizeof(pthread_t));
+	b = buf; rest = k;
+	for (i = 0; i < n_threads; ++i) {
+		w[i].buf_len = rest / (n_threads - i);
+		w[i].buf = b;
+		w[i].prefix = prefix;
+		w[i].h = h;
+		w[i].index = n_files + i;
+		b += w[i].buf_len; rest -= w[i].buf_len;
+		pthread_create(&tid[i], &attr, worker, &w[i]);
+	}
+	for (i = 0; i < n_threads; ++i) pthread_join(tid[i], 0);
+	free(tid); free(w);
+	return n_files + n_threads;
+}
+
+/*!
+  @abstract Sort an unsorted BAM file based on the chromosome order
+  and the leftmost position of an alignment
+
+  @param  is_by_qname whether to sort by query name
+  @param  fn       name of the file to be sorted
+  @param  prefix   prefix of the output and the temporary files; upon
+	                   sucessess, prefix.bam will be written.
+  @param  max_mem  approxiate maximum memory (very inaccurate)
+
+  @discussion It may create multiple temporary subalignment files
+  and then merge them by calling bam_merge_core(). This function is
+  NOT thread safe.
+ */
+void bam_sort_core_ext(int is_by_qname, const char *fn, const char *prefix, size_t _max_mem, int is_stdout, int n_threads, int level)
+{
+	int ret, i, n_files = 0;
+	size_t mem, max_k, k, max_mem;
 	bam_header_t *header;
 	bamFile fp;
 	bam1_t *b, **buf;
+	char *fnout = 0;
 
+	if (n_threads < 2) n_threads = 1;
 	g_is_by_qname = is_by_qname;
-	n = k = 0; mem = 0;
+	max_k = k = 0; mem = 0;
+	max_mem = _max_mem * n_threads;
+	buf = 0;
 	fp = strcmp(fn, "-")? bam_open(fn, "r") : bam_dopen(fileno(stdin), "r");
-	assert(fp);
+	if (fp == 0) {
+		fprintf(stderr, "[bam_sort_core] fail to open file %s\n", fn);
+		return;
+	}
 	header = bam_header_read(fp);
-	buf = (bam1_t**)calloc(max_mem / BAM_CORE_SIZE, sizeof(bam1_t*));
+	if (is_by_qname) change_SO(header, "queryname");
+	else change_SO(header, "coordinate");
 	// write sub files
 	for (;;) {
+		if (k == max_k) {
+			size_t old_max = max_k;
+			max_k = max_k? max_k<<1 : 0x10000;
+			buf = realloc(buf, max_k * sizeof(void*));
+			memset(buf + old_max, 0, sizeof(void*) * (max_k - old_max));
+		}
 		if (buf[k] == 0) buf[k] = (bam1_t*)calloc(1, sizeof(bam1_t));
 		b = buf[k];
 		if ((ret = bam_read1(fp, b)) < 0) break;
-		mem += ret;
+		if (b->data_len < b->m_data>>2) { // shrink
+			b->m_data = b->data_len;
+			kroundup32(b->m_data);
+			b->data = realloc(b->data, b->m_data);
+		}
+		mem += sizeof(bam1_t) + b->m_data + sizeof(void*) + sizeof(void*); // two sizeof(void*) for the data allocated to pointer arrays
 		++k;
 		if (mem >= max_mem) {
-			sort_blocks(n++, k, buf, prefix, header);
-			mem = 0; k = 0;
+			n_files = sort_blocks(n_files, k, buf, prefix, header, n_threads);
+			mem = k = 0;
 		}
 	}
 	if (ret != -1)
 		fprintf(stderr, "[bam_sort_core] truncated file. Continue anyway.\n");
-	if (n == 0) sort_blocks(-1, k, buf, prefix, header);
-	else { // then merge
-		char **fns, *fnout;
-		fprintf(stderr, "[bam_sort_core] merging from %d files...\n", n+1);
-		sort_blocks(n++, k, buf, prefix, header);
-		fnout = (char*)calloc(strlen(prefix) + 20, 1);
-		sprintf(fnout, "%s.bam", prefix);
-		fns = (char**)calloc(n, sizeof(char*));
-		for (i = 0; i < n; ++i) {
+	// output file name
+	fnout = calloc(strlen(prefix) + 20, 1);
+	if (is_stdout) sprintf(fnout, "-");
+	else sprintf(fnout, "%s.bam", prefix);
+	// write the final output
+	if (n_files == 0) { // a single block
+		char mode[8];
+		strcpy(mode, "w");
+		if (level >= 0) sprintf(mode + 1, "%d", level < 9? level : 9);
+		ks_mergesort(sort, k, buf, 0);
+		write_buffer(fnout, mode, k, buf, header, n_threads);
+	} else { // then merge
+		char **fns;
+		n_files = sort_blocks(n_files, k, buf, prefix, header, n_threads);
+		fprintf(stderr, "[bam_sort_core] merging from %d files...\n", n_files);
+		fns = (char**)calloc(n_files, sizeof(char*));
+		for (i = 0; i < n_files; ++i) {
 			fns[i] = (char*)calloc(strlen(prefix) + 20, 1);
 			sprintf(fns[i], "%s.%.4d.bam", prefix, i);
 		}
-		bam_merge_core(is_by_qname, fnout, n, fns);
-		free(fnout);
-		for (i = 0; i < n; ++i) {
+		bam_merge_core2(is_by_qname, fnout, 0, n_files, fns, 0, 0, n_threads, level);
+		for (i = 0; i < n_files; ++i) {
 			unlink(fns[i]);
 			free(fns[i]);
 		}
 		free(fns);
 	}
-	for (k = 0; k < max_mem / BAM_CORE_SIZE; ++k) {
-		if (buf[k]) {
-			free(buf[k]->data);
-			free(buf[k]);
-		}
+	free(fnout);
+	// free
+	for (k = 0; k < max_k; ++k) {
+		if (!buf[k]) continue;
+		free(buf[k]->data);
+		free(buf[k]);
 	}
 	free(buf);
 	bam_header_destroy(header);
 	bam_close(fp);
 }
 
+void bam_sort_core(int is_by_qname, const char *fn, const char *prefix, size_t max_mem)
+{
+	bam_sort_core_ext(is_by_qname, fn, prefix, max_mem, 0, 0, -1);
+}
+
 int bam_sort(int argc, char *argv[])
 {
-	size_t max_mem = 500000000;
-	int c, is_by_qname = 0;
-	while ((c = getopt(argc, argv, "nm:")) >= 0) {
+	size_t max_mem = 768<<20; // 512MB
+	int c, is_by_qname = 0, is_stdout = 0, n_threads = 0, level = -1;
+	while ((c = getopt(argc, argv, "nom:@:l:")) >= 0) {
 		switch (c) {
+		case 'o': is_stdout = 1; break;
 		case 'n': is_by_qname = 1; break;
-		case 'm': max_mem = atol(optarg); break;
+		case 'm': {
+				char *q;
+				max_mem = strtol(optarg, &q, 0);
+				if (*q == 'k' || *q == 'K') max_mem <<= 10;
+				else if (*q == 'm' || *q == 'M') max_mem <<= 20;
+				else if (*q == 'g' || *q == 'G') max_mem <<= 30;
+				break;
+			}
+		case '@': n_threads = atoi(optarg); break;
+		case 'l': level = atoi(optarg); break;
 		}
 	}
 	if (optind + 2 > argc) {
-		fprintf(stderr, "Usage: samtools sort [-n] [-m <maxMem>] <in.baf> <out.prefix>\n");
+		fprintf(stderr, "\n");
+		fprintf(stderr, "Usage:   samtools sort [options] <in.bam> <out.prefix>\n\n");
+		fprintf(stderr, "Options: -n        sort by read name\n");
+		fprintf(stderr, "         -o        final output to stdout\n");
+		fprintf(stderr, "         -l INT    compression level, from 0 to 9 [-1]\n");
+		fprintf(stderr, "         -@ INT    number of sorting and compression threads [1]\n");
+		fprintf(stderr, "         -m INT    max memory per thread; suffix K/M/G recognized [768M]\n");
+		fprintf(stderr, "\n");
 		return 1;
 	}
-	bam_sort_core(is_by_qname, argv[optind], argv[optind+1], max_mem);
+	bam_sort_core_ext(is_by_qname, argv[optind], argv[optind+1], max_mem, is_stdout, n_threads, level);
 	return 0;
 }