13 KSTREAM_INIT(gzFile, gzread, 8192)
14 KHASH_MAP_INIT_STR(ref, uint64_t)
16 void bam_init_header_hash(bam_header_t *header);
17 void bam_destroy_header_hash(bam_header_t *header);
18 int32_t bam_get_tid(const bam_header_t *header, const char *seq_name);
20 unsigned char bam_nt16_table[256] = {
21 15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
22 15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
23 15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
24 1, 2, 4, 8, 15,15,15,15, 15,15,15,15, 15, 0 /*=*/,15,15,
25 15, 1,14, 2, 13,15,15, 4, 11,15,15,12, 15, 3,15,15,
26 15,15, 5, 6, 8,15, 7, 9, 15,10,15,15, 15,15,15,15,
27 15, 1,14, 2, 13,15,15, 4, 11,15,15,12, 15, 3,15,15,
28 15,15, 5, 6, 8,15, 7, 9, 15,10,15,15, 15,15,15,15,
29 15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
30 15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
31 15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
32 15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
33 15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
34 15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
35 15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
36 15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15
39 char *bam_nt16_rev_table = "=ACMGRSVTWYHKDBN";
49 char **__bam_get_lines(const char *fn, int *_n) // for bam_plcmd.c only
52 int n = 0, dret, m = 0;
53 gzFile fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
56 str = (kstring_t*)calloc(1, sizeof(kstring_t));
58 while (ks_getuntil(ks, '\n', str, &dret) > 0) {
61 list = (char**)realloc(list, m * sizeof(char*));
63 if (str->s[str->l-1] == '\r')
64 str->s[--str->l] = '\0';
65 s = list[n++] = (char*)calloc(str->l + 1, 1);
70 free(str->s); free(str);
75 static bam_header_t *hash2header(const kh_ref_t *hash)
79 header = bam_header_init();
80 header->n_targets = kh_size(hash);
81 header->target_name = (char**)calloc(kh_size(hash), sizeof(char*));
82 header->target_len = (uint32_t*)calloc(kh_size(hash), 4);
83 for (k = kh_begin(hash); k != kh_end(hash); ++k) {
84 if (kh_exist(hash, k)) {
85 int i = (int)kh_value(hash, k);
86 header->target_name[i] = (char*)kh_key(hash, k);
87 header->target_len[i] = kh_value(hash, k)>>32;
90 bam_init_header_hash(header);
93 bam_header_t *sam_header_read2(const char *fn)
103 fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
106 str = (kstring_t*)calloc(1, sizeof(kstring_t));
107 while (ks_getuntil(ks, 0, str, &dret) > 0) {
108 char *s = strdup(str->s);
111 ks_getuntil(ks, 0, str, &dret);
113 k = kh_put(ref, hash, s, &ret);
114 kh_value(hash, k) = (uint64_t)len<<32 | i;
116 while ((c = ks_getc(ks)) != '\n' && c != -1);
120 free(str->s); free(str);
121 fprintf(stderr, "[sam_header_read2] %d sequences loaded.\n", kh_size(hash));
122 header = hash2header(hash);
123 kh_destroy(ref, hash);
126 static inline uint8_t *alloc_data(bam1_t *b, int size)
128 if (b->m_data < size) {
130 kroundup32(b->m_data);
131 b->data = (uint8_t*)realloc(b->data, b->m_data);
135 static inline void parse_error(int64_t n_lines, const char * __restrict msg)
137 fprintf(stderr, "Parse error at line %lld: %s\n", (long long)n_lines, msg);
140 static inline void append_text(bam_header_t *header, kstring_t *str)
142 int x = header->l_text, y = header->l_text + str->l + 2; // 2 = 1 byte dret + 1 byte null
143 kroundup32(x); kroundup32(y);
144 if (x < y) header->text = (char*)realloc(header->text, y);
145 strncpy(header->text + header->l_text, str->s, str->l+1); // we cannot use strcpy() here.
146 header->l_text += str->l + 1;
147 header->text[header->l_text] = 0;
150 int sam_header_parse_rg(bam_header_t *h)
152 kstring_t *rgid, *rglib;
157 if (h == 0) return 0;
158 bam_strmap_destroy(h->rg2lib); h->rg2lib = 0;
159 if (h->l_text < 3) return 0;
161 h->rg2lib = bam_strmap_init();
162 rgid = calloc(1, sizeof(kstring_t));
163 rglib = calloc(1, sizeof(kstring_t));
165 while ((s = strstr(s, "@RG")) != 0) {
166 if (rgid->l && rglib->l) {
167 bam_strmap_put(h->rg2lib, rgid->s, rglib->s);
170 rgid->l = rglib->l = 0;
173 if ((p = strstr(s, "ID:")) != 0) {
175 for (p = q; *p && *p != '\t' && *p != '\r' && *p != '\n'; ++p);
176 kputsn(q, p - q, rgid);
178 fprintf(stderr, "[bam_header_parse] missing ID tag in @RG lines.\n");
182 if ((p = strstr(s, "LB:")) != 0) {
184 for (p = q; *p && *p != '\t' && *p != '\r' && *p != '\n'; ++p);
185 kputsn(q, p - q, rglib);
187 fprintf(stderr, "[bam_header_parse] missing LB tag in @RG lines.\n");
193 if (rgid->l && rglib->l) {
194 bam_strmap_put(h->rg2lib, rgid->s, rglib->s);
197 free(rgid->s); free(rgid);
198 free(rglib->s); free(rglib);
200 bam_strmap_destroy(h->rg2lib);
206 int sam_header_parse(bam_header_t *h)
212 free(h->target_len); free(h->target_name);
213 h->n_targets = 0; h->target_len = 0; h->target_name = 0;
214 if (h->l_text < 3) return 0;
215 // count number of @SQ
217 while ((s = strstr(s, "@SQ")) != 0) {
221 if (h->n_targets == 0) return 0;
222 h->target_len = (uint32_t*)calloc(h->n_targets, 4);
223 h->target_name = (char**)calloc(h->n_targets, sizeof(void*));
227 while ((s = strstr(s, "@SQ")) != 0) {
230 if ((p = strstr(s, "SN:")) != 0) {
232 for (p = q; *p && *p != '\t' && *p != '\r' && *p != '\n'; ++p);
233 h->target_name[i] = (char*)calloc(p - q + 1, 1);
234 strncpy(h->target_name[i], q, p - q);
235 } else goto header_err_ret;
237 if ((p = strstr(s, "LN:")) != 0) h->target_len[i] = strtol(p + 3, 0, 10);
238 else goto header_err_ret;
243 sam_header_parse_rg(h);
247 fprintf(stderr, "[bam_header_parse] missing SN or LN tag in @SQ lines.\n");
248 free(h->target_len); free(h->target_name);
249 h->n_targets = 0; h->target_len = 0; h->target_name = 0;
253 bam_header_t *sam_header_read(tamFile fp)
256 bam_header_t *header = bam_header_init();
257 kstring_t *str = fp->str;
258 while ((ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret)) >= 0 && str->s[0] == '@') { // skip header
259 str->s[str->l] = dret; // note that str->s is NOT null terminated!!
260 append_text(header, str);
262 ret = ks_getuntil(fp->ks, '\n', str, &dret);
263 str->s[str->l] = '\n'; // NOT null terminated!!
264 append_text(header, str);
268 sam_header_parse(header);
269 bam_init_header_hash(header);
274 int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b)
276 int ret, doff, doff0, dret, z = 0;
277 bam1_core_t *c = &b->core;
278 kstring_t *str = fp->str;
279 kstream_t *ks = fp->ks;
285 do { // special consideration for empty lines
286 ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret);
287 if (ret >= 0) z += str->l + 1;
290 if (ret < 0) return -1;
295 c->l_qname = strlen(str->s) + 1;
296 memcpy(alloc_data(b, doff + c->l_qname) + doff, str->s, c->l_qname);
299 { // flag, tid, pos, qual
300 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->flag = atoi(str->s);
301 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->tid = bam_get_tid(header, str->s);
302 if (c->tid < 0 && strcmp(str->s, "*")) {
303 if (header->n_targets == 0) {
304 fprintf(stderr, "[sam_read1] missing header? Abort!\n");
306 } else fprintf(stderr, "[sam_read1] reference '%s' is recognized as '*'.\n", str->s);
308 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->pos = isdigit(str->s[0])? atoi(str->s) - 1 : -1;
309 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->qual = isdigit(str->s[0])? atoi(str->s) : 0;
310 if (ret < 0) return -2;
317 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -3;
319 if (str->s[0] != '*') {
320 for (s = str->s; *s; ++s) {
321 if (isalpha(*s)) ++c->n_cigar;
322 else if (!isdigit(*s)) parse_error(fp->n_lines, "invalid CIGAR character");
324 b->data = alloc_data(b, doff + c->n_cigar * 4);
325 for (i = 0, s = str->s; i != c->n_cigar; ++i) {
326 x = strtol(s, &t, 10);
328 if (op == 'M') op = BAM_CMATCH;
329 else if (op == 'I') op = BAM_CINS;
330 else if (op == 'D') op = BAM_CDEL;
331 else if (op == 'N') op = BAM_CREF_SKIP;
332 else if (op == 'S') op = BAM_CSOFT_CLIP;
333 else if (op == 'H') op = BAM_CHARD_CLIP;
334 else if (op == 'P') op = BAM_CPAD;
335 else parse_error(fp->n_lines, "invalid CIGAR operation");
337 bam1_cigar(b)[i] = x << BAM_CIGAR_SHIFT | op;
339 if (*s) parse_error(fp->n_lines, "unmatched CIGAR operation");
340 c->bin = bam_reg2bin(c->pos, bam_calend(c, bam1_cigar(b)));
341 doff += c->n_cigar * 4;
342 } else c->bin = bam_reg2bin(c->pos, c->pos + 1);
344 { // mtid, mpos, isize
345 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
346 c->mtid = strcmp(str->s, "=")? bam_get_tid(header, str->s) : c->tid;
347 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
348 c->mpos = isdigit(str->s[0])? atoi(str->s) - 1 : -1;
349 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
350 c->isize = (str->s[0] == '-' || isdigit(str->s[0]))? atoi(str->s) : 0;
351 if (ret < 0) return -4;
356 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -5; // seq
358 c->l_qseq = strlen(str->s);
359 if (c->n_cigar && c->l_qseq != (int32_t)bam_cigar2qlen(c, bam1_cigar(b)))
360 parse_error(fp->n_lines, "CIGAR and sequence length are inconsistent");
361 p = (uint8_t*)alloc_data(b, doff + c->l_qseq + (c->l_qseq+1)/2) + doff;
362 bzero(p, (c->l_qseq+1)/2);
363 for (i = 0; i < c->l_qseq; ++i)
364 p[i/2] |= bam_nt16_table[(int)str->s[i]] << 4*(1-i%2);
365 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -6; // qual
367 if (strcmp(str->s, "*") && c->l_qseq != strlen(str->s))
368 parse_error(fp->n_lines, "sequence and quality are inconsistent");
369 p += (c->l_qseq+1)/2;
370 if (strcmp(str->s, "*") == 0) for (i = 0; i < c->l_qseq; ++i) p[i] = 0xff;
371 else for (i = 0; i < c->l_qseq; ++i) p[i] = str->s[i] - 33;
372 doff += c->l_qseq + (c->l_qseq+1)/2;
375 if (dret != '\n' && dret != '\r') { // aux
376 while (ks_getuntil(ks, KS_SEP_TAB, str, &dret) >= 0) {
377 uint8_t *s, type, key[2];
379 if (str->l < 6 || str->s[2] != ':' || str->s[4] != ':')
380 parse_error(fp->n_lines, "missing colon in auxiliary data");
381 key[0] = str->s[0]; key[1] = str->s[1];
383 s = alloc_data(b, doff + 3) + doff;
384 s[0] = key[0]; s[1] = key[1]; s += 2; doff += 2;
385 if (type == 'A' || type == 'a' || type == 'c' || type == 'C') { // c and C for backward compatibility
386 s = alloc_data(b, doff + 2) + doff;
387 *s++ = 'A'; *s = str->s[5];
389 } else if (type == 'I' || type == 'i') {
391 s = alloc_data(b, doff + 5) + doff;
392 x = (long long)atoll(str->s + 5);
395 *s++ = 'c'; *(int8_t*)s = (int8_t)x;
397 } else if (x >= -32767) {
398 *s++ = 's'; *(int16_t*)s = (int16_t)x;
401 *s++ = 'i'; *(int32_t*)s = (int32_t)x;
403 if (x < -2147483648ll)
404 fprintf(stderr, "Parse warning at line %lld: integer %lld is out of range.",
405 (long long)fp->n_lines, x);
409 *s++ = 'C'; *s++ = (uint8_t)x;
411 } else if (x <= 65535) {
412 *s++ = 'S'; *(uint16_t*)s = (uint16_t)x;
415 *s++ = 'I'; *(uint32_t*)s = (uint32_t)x;
417 if (x > 4294967295ll)
418 fprintf(stderr, "Parse warning at line %lld: integer %lld is out of range.",
419 (long long)fp->n_lines, x);
422 } else if (type == 'f') {
423 s = alloc_data(b, doff + 5) + doff;
425 *(float*)s = (float)atof(str->s + 5);
427 } else if (type == 'd') {
428 s = alloc_data(b, doff + 9) + doff;
430 *(float*)s = (float)atof(str->s + 9);
432 } else if (type == 'Z' || type == 'H') {
433 int size = 1 + (str->l - 5) + 1;
434 if (type == 'H') { // check whether the hex string is valid
436 if ((str->l - 5) % 2 == 1) parse_error(fp->n_lines, "length of the hex string not even");
437 for (i = 0; i < str->l - 5; ++i) {
438 int c = toupper(str->s[5 + i]);
439 if (!((c >= '0' && c <= '9') || (c >= 'A' && c <= 'F')))
440 parse_error(fp->n_lines, "invalid hex character");
443 s = alloc_data(b, doff + size) + doff;
445 memcpy(s, str->s + 5, str->l - 5);
448 } else parse_error(fp->n_lines, "unrecognized type");
449 if (dret == '\n' || dret == '\r') break;
452 b->l_aux = doff - doff0;
457 tamFile sam_open(const char *fn)
460 fp = (tamFile)calloc(1, sizeof(struct __tamFile_t));
461 fp->str = (kstring_t*)calloc(1, sizeof(kstring_t));
462 fp->fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
463 fp->ks = ks_init(fp->fp);
467 void sam_close(tamFile fp)
472 free(fp->str->s); free(fp->str);