6 #include "bam_endian.h"
8 #include "sam_header.h"
11 char *bam_flag2char_table = "pPuUrR12sfd\0\0\0\0\0";
13 /**************************
14 * CIGAR related routines *
15 **************************/
17 uint32_t bam_calend(const bam1_core_t *c, const uint32_t *cigar)
21 for (k = 0; k < c->n_cigar; ++k) {
22 int op = cigar[k] & BAM_CIGAR_MASK;
23 if (op == BAM_CMATCH || op == BAM_CDEL || op == BAM_CREF_SKIP)
24 end += cigar[k] >> BAM_CIGAR_SHIFT;
29 int32_t bam_cigar2qlen(const bam1_core_t *c, const uint32_t *cigar)
33 for (k = 0; k < c->n_cigar; ++k) {
34 int op = cigar[k] & BAM_CIGAR_MASK;
35 if (op == BAM_CMATCH || op == BAM_CINS || op == BAM_CSOFT_CLIP)
36 l += cigar[k] >> BAM_CIGAR_SHIFT;
45 bam_header_t *bam_header_init()
47 bam_is_be = bam_is_big_endian();
48 return (bam_header_t*)calloc(1, sizeof(bam_header_t));
51 void bam_header_destroy(bam_header_t *header)
54 extern void bam_destroy_header_hash(bam_header_t *header);
55 if (header == 0) return;
56 if (header->target_name) {
57 for (i = 0; i < header->n_targets; ++i)
58 free(header->target_name[i]);
59 free(header->target_name);
60 free(header->target_len);
63 if (header->dict) sam_header_free(header->dict);
64 if (header->rg2lib) sam_tbl_destroy(header->rg2lib);
65 bam_destroy_header_hash(header);
69 bam_header_t *bam_header_read(bamFile fp)
74 int32_t i = 1, name_len;
76 i = bgzf_check_EOF(fp);
78 // If the file is a pipe, checking the EOF marker will *always* fail
79 // with ESPIPE. Suppress the error message in this case.
80 if (errno != ESPIPE) perror("[bam_header_read] bgzf_check_EOF");
82 else if (i == 0) fprintf(stderr, "[bam_header_read] EOF marker is absent.\n");
84 magic_len = bam_read(fp, buf, 4);
85 if (magic_len != 4 || strncmp(buf, "BAM\001", 4) != 0) {
86 fprintf(stderr, "[bam_header_read] invalid BAM binary header (this is not a BAM file).\n");
89 header = bam_header_init();
90 // read plain text and the number of reference sequences
91 bam_read(fp, &header->l_text, 4);
92 if (bam_is_be) bam_swap_endian_4p(&header->l_text);
93 header->text = (char*)calloc(header->l_text + 1, 1);
94 bam_read(fp, header->text, header->l_text);
95 bam_read(fp, &header->n_targets, 4);
96 if (bam_is_be) bam_swap_endian_4p(&header->n_targets);
97 // read reference sequence names and lengths
98 header->target_name = (char**)calloc(header->n_targets, sizeof(char*));
99 header->target_len = (uint32_t*)calloc(header->n_targets, 4);
100 for (i = 0; i != header->n_targets; ++i) {
101 bam_read(fp, &name_len, 4);
102 if (bam_is_be) bam_swap_endian_4p(&name_len);
103 header->target_name[i] = (char*)calloc(name_len, 1);
104 bam_read(fp, header->target_name[i], name_len);
105 bam_read(fp, &header->target_len[i], 4);
106 if (bam_is_be) bam_swap_endian_4p(&header->target_len[i]);
111 int bam_header_write(bamFile fp, const bam_header_t *header)
114 int32_t i, name_len, x;
116 strncpy(buf, "BAM\001", 4);
117 bam_write(fp, buf, 4);
118 // write plain text and the number of reference sequences
120 x = bam_swap_endian_4(header->l_text);
121 bam_write(fp, &x, 4);
122 if (header->l_text) bam_write(fp, header->text, header->l_text);
123 x = bam_swap_endian_4(header->n_targets);
124 bam_write(fp, &x, 4);
126 bam_write(fp, &header->l_text, 4);
127 if (header->l_text) bam_write(fp, header->text, header->l_text);
128 bam_write(fp, &header->n_targets, 4);
130 // write sequence names and lengths
131 for (i = 0; i != header->n_targets; ++i) {
132 char *p = header->target_name[i];
133 name_len = strlen(p) + 1;
135 x = bam_swap_endian_4(name_len);
136 bam_write(fp, &x, 4);
137 } else bam_write(fp, &name_len, 4);
138 bam_write(fp, p, name_len);
140 x = bam_swap_endian_4(header->target_len[i]);
141 bam_write(fp, &x, 4);
142 } else bam_write(fp, &header->target_len[i], 4);
148 static void swap_endian_data(const bam1_core_t *c, int data_len, uint8_t *data)
151 uint32_t i, *cigar = (uint32_t*)(data + c->l_qname);
152 s = data + c->n_cigar*4 + c->l_qname + c->l_qseq + (c->l_qseq + 1)/2;
153 for (i = 0; i < c->n_cigar; ++i) bam_swap_endian_4p(&cigar[i]);
154 while (s < data + data_len) {
157 type = toupper(*s); ++s; // skip type
158 if (type == 'C' || type == 'A') ++s;
159 else if (type == 'S') { bam_swap_endian_2p(s); s += 2; }
160 else if (type == 'I' || type == 'F') { bam_swap_endian_4p(s); s += 4; }
161 else if (type == 'D') { bam_swap_endian_8p(s); s += 8; }
162 else if (type == 'Z' || type == 'H') { while (*s) ++s; ++s; }
166 int bam_read1(bamFile fp, bam1_t *b)
168 bam1_core_t *c = &b->core;
169 int32_t block_len, ret, i;
172 assert(BAM_CORE_SIZE == 32);
173 if ((ret = bam_read(fp, &block_len, 4)) != 4) {
174 if (ret == 0) return -1; // normal end-of-file
175 else return -2; // truncated
177 if (bam_read(fp, x, BAM_CORE_SIZE) != BAM_CORE_SIZE) return -3;
179 bam_swap_endian_4p(&block_len);
180 for (i = 0; i < 8; ++i) bam_swap_endian_4p(x + i);
182 c->tid = x[0]; c->pos = x[1];
183 c->bin = x[2]>>16; c->qual = x[2]>>8&0xff; c->l_qname = x[2]&0xff;
184 c->flag = x[3]>>16; c->n_cigar = x[3]&0xffff;
186 c->mtid = x[5]; c->mpos = x[6]; c->isize = x[7];
187 b->data_len = block_len - BAM_CORE_SIZE;
188 if (b->m_data < b->data_len) {
189 b->m_data = b->data_len;
190 kroundup32(b->m_data);
191 b->data = (uint8_t*)realloc(b->data, b->m_data);
193 if (bam_read(fp, b->data, b->data_len) != b->data_len) return -4;
194 b->l_aux = b->data_len - c->n_cigar * 4 - c->l_qname - c->l_qseq - (c->l_qseq+1)/2;
195 if (bam_is_be) swap_endian_data(c, b->data_len, b->data);
196 return 4 + block_len;
199 inline int bam_write1_core(bamFile fp, const bam1_core_t *c, int data_len, uint8_t *data)
201 uint32_t x[8], block_len = data_len + BAM_CORE_SIZE, y;
203 assert(BAM_CORE_SIZE == 32);
206 x[2] = (uint32_t)c->bin<<16 | c->qual<<8 | c->l_qname;
207 x[3] = (uint32_t)c->flag<<16 | c->n_cigar;
212 bgzf_flush_try(fp, 4 + block_len);
214 for (i = 0; i < 8; ++i) bam_swap_endian_4p(x + i);
216 bam_write(fp, bam_swap_endian_4p(&y), 4);
217 swap_endian_data(c, data_len, data);
218 } else bam_write(fp, &block_len, 4);
219 bam_write(fp, x, BAM_CORE_SIZE);
220 bam_write(fp, data, data_len);
221 if (bam_is_be) swap_endian_data(c, data_len, data);
222 return 4 + block_len;
225 int bam_write1(bamFile fp, const bam1_t *b)
227 return bam_write1_core(fp, &b->core, b->data_len, b->data);
230 char *bam_format1_core(const bam_header_t *header, const bam1_t *b, int of)
232 uint8_t *s = bam1_seq(b), *t = bam1_qual(b);
234 const bam1_core_t *c = &b->core;
236 str.l = str.m = 0; str.s = 0;
238 kputsn(bam1_qname(b), c->l_qname-1, &str); kputc('\t', &str);
239 if (of == BAM_OFDEC) { kputw(c->flag, &str); kputc('\t', &str); }
240 else if (of == BAM_OFHEX) ksprintf(&str, "0x%x\t", c->flag);
242 for (i = 0; i < 16; ++i)
243 if ((c->flag & 1<<i) && bam_flag2char_table[i])
244 kputc(bam_flag2char_table[i], &str);
247 if (c->tid < 0) kputsn("*\t", 2, &str);
248 else { kputs(header->target_name[c->tid], &str); kputc('\t', &str); }
249 kputw(c->pos + 1, &str); kputc('\t', &str); kputw(c->qual, &str); kputc('\t', &str);
250 if (c->n_cigar == 0) kputc('*', &str);
252 for (i = 0; i < c->n_cigar; ++i) {
253 kputw(bam1_cigar(b)[i]>>BAM_CIGAR_SHIFT, &str);
254 kputc("MIDNSHP"[bam1_cigar(b)[i]&BAM_CIGAR_MASK], &str);
258 if (c->mtid < 0) kputsn("*\t", 2, &str);
259 else if (c->mtid == c->tid) kputsn("=\t", 2, &str);
260 else { kputs(header->target_name[c->mtid], &str); kputc('\t', &str); }
261 kputw(c->mpos + 1, &str); kputc('\t', &str); kputw(c->isize, &str); kputc('\t', &str);
263 for (i = 0; i < c->l_qseq; ++i) kputc(bam_nt16_rev_table[bam1_seqi(s, i)], &str);
265 if (t[0] == 0xff) kputc('*', &str);
266 else for (i = 0; i < c->l_qseq; ++i) kputc(t[i] + 33, &str);
267 } else kputsn("*\t*", 3, &str);
269 while (s < b->data + b->data_len) {
270 uint8_t type, key[2];
271 key[0] = s[0]; key[1] = s[1];
272 s += 2; type = *s; ++s;
273 kputc('\t', &str); kputsn((char*)key, 2, &str); kputc(':', &str);
274 if (type == 'A') { kputsn("A:", 2, &str); kputc(*s, &str); ++s; }
275 else if (type == 'C') { kputsn("i:", 2, &str); kputw(*s, &str); ++s; }
276 else if (type == 'c') { kputsn("i:", 2, &str); kputw(*(int8_t*)s, &str); ++s; }
277 else if (type == 'S') { kputsn("i:", 2, &str); kputw(*(uint16_t*)s, &str); s += 2; }
278 else if (type == 's') { kputsn("i:", 2, &str); kputw(*(int16_t*)s, &str); s += 2; }
279 else if (type == 'I') { kputsn("i:", 2, &str); kputuw(*(uint32_t*)s, &str); s += 4; }
280 else if (type == 'i') { kputsn("i:", 2, &str); kputw(*(int32_t*)s, &str); s += 4; }
281 else if (type == 'f') { ksprintf(&str, "f:%g", *(float*)s); s += 4; }
282 else if (type == 'd') { ksprintf(&str, "d:%lg", *(double*)s); s += 8; }
283 else if (type == 'Z' || type == 'H') { kputc(type, &str); kputc(':', &str); while (*s) kputc(*s++, &str); ++s; }
288 char *bam_format1(const bam_header_t *header, const bam1_t *b)
290 return bam_format1_core(header, b, BAM_OFDEC);
293 void bam_view1(const bam_header_t *header, const bam1_t *b)
295 char *s = bam_format1(header, b);
300 // FIXME: we should also check the LB tag associated with each alignment
301 const char *bam_get_library(bam_header_t *h, const bam1_t *b)
304 if (h->dict == 0) h->dict = sam_header_parse2(h->text);
305 if (h->rg2lib == 0) h->rg2lib = sam_header2tbl(h->dict, "RG", "ID", "LB");
306 rg = bam_aux_get(b, "RG");
307 return (rg == 0)? 0 : sam_tbl_get(h->rg2lib, (const char*)(rg + 1));