7 typedef bam1_t *bam1_p;
9 KHASH_SET_INIT_STR(name)
10 KHASH_MAP_INIT_INT64(pos, bam1_p)
12 #define BUFFER_SIZE 0x40000
19 static inline void stack_insert(tmp_stack_t *stack, bam1_t *b)
21 if (stack->n == stack->max) {
22 stack->max = stack->max? stack->max<<1 : 0x10000;
23 stack->a = (bam1_t**)realloc(stack->a, sizeof(bam1_t*) * stack->max);
25 stack->a[stack->n++] = b;
28 static inline void dump_best(tmp_stack_t *stack, khash_t(pos) *best_hash, bamFile out)
31 for (i = 0; i != stack->n; ++i) {
32 bam_write1(out, stack->a[i]);
33 bam_destroy1(stack->a[i]);
36 if (kh_size(best_hash) > BUFFER_SIZE) kh_clear(pos, best_hash);
39 static void clear_del_set(khash_t(name) *del_set)
42 for (k = kh_begin(del_set); k < kh_end(del_set); ++k)
43 if (kh_exist(del_set, k))
44 free((char*)kh_key(del_set, k));
45 kh_clear(name, del_set);
48 void bam_rmdup_core(bamFile in, bamFile out)
52 int last_tid = -1, last_pos = -1;
53 uint64_t n_checked = 0, n_removed = 0;
56 khash_t(pos) *best_hash;
57 khash_t(name) *del_set;
59 best_hash = kh_init(pos);
60 del_set = kh_init(name);
62 memset(&stack, 0, sizeof(tmp_stack_t));
63 header = bam_header_read(in);
64 bam_header_write(out, header);
66 kh_resize(name, del_set, 4 * BUFFER_SIZE);
67 kh_resize(pos, best_hash, 3 * BUFFER_SIZE);
68 while (bam_read1(in, b) >= 0) {
69 bam1_core_t *c = &b->core;
70 if (c->tid != last_tid || last_pos != c->pos) {
71 dump_best(&stack, best_hash, out); // write the result
72 if (c->tid != last_tid) {
73 kh_clear(pos, best_hash);
74 if (kh_size(del_set)) { // check
75 fprintf(stderr, "[bam_rmdup_core] %llu unmatched pairs\n", (long long)kh_size(del_set));
76 clear_del_set(del_set);
78 if ((int)c->tid == -1) { // append unmapped reads
80 while (bam_read1(in, b) >= 0) bam_write1(out, b);
84 fprintf(stderr, "[bam_rmdup_core] processing reference %s...\n", header->target_name[c->tid]);
87 if (!(c->flag&BAM_FPAIRED) || (c->flag&(BAM_FUNMAP|BAM_FMUNMAP)) || (c->mtid >= 0 && c->tid != c->mtid)) {
89 } else if (c->isize > 0) { // paired, head
90 uint64_t key = (uint64_t)c->pos<<32 | c->isize;
93 k = kh_put(pos, best_hash, key, &ret);
94 if (ret == 0) { // found in best_hash
95 bam1_t *p = kh_val(best_hash, k);
97 if (p->core.qual < c->qual) { // the current alignment is better
98 kh_put(name, del_set, strdup(bam1_qname(p)), &ret); // p will be removed
99 bam_copy1(p, b); // replaced as b
100 } else kh_put(name, del_set, strdup(bam1_qname(b)), &ret); // b will be removed
102 fprintf(stderr, "[bam_rmdup_core] inconsistent BAM file for pair '%s'. Continue anyway.\n", bam1_qname(b));
103 } else { // not found in best_hash
104 kh_val(best_hash, k) = bam_dup1(b);
105 stack_insert(&stack, kh_val(best_hash, k));
107 } else { // paired, tail
108 k = kh_get(name, del_set, bam1_qname(b));
109 if (k != kh_end(del_set)) {
110 free((char*)kh_key(del_set, k));
111 kh_del(name, del_set, k);
112 } else bam_write1(out, b);
116 dump_best(&stack, best_hash, out);
118 bam_header_destroy(header);
119 clear_del_set(del_set);
120 kh_destroy(name, del_set);
121 kh_destroy(pos, best_hash);
124 fprintf(stderr, "[bam_rmdup_core] %lld / %lld = %.4lf\n", (long long)n_removed, (long long)n_checked,
125 (double)n_removed/n_checked);
127 int bam_rmdup(int argc, char *argv[])
131 fprintf(stderr, "Usage: samtools rmdup <input.srt.bam> <output.bam>\n\n");
132 fprintf(stderr, "Note: Picard is recommended for this task.\n");
135 in = (strcmp(argv[1], "-") == 0)? bam_dopen(fileno(stdin), "r") : bam_open(argv[1], "r");
136 out = (strcmp(argv[2], "-") == 0)? bam_dopen(fileno(stdout), "w") : bam_open(argv[2], "w");
137 if (in == 0 || out == 0) {
138 fprintf(stderr, "[bam_rmdup] fail to read/write input files\n");
141 bam_rmdup_core(in, out);