13 #include "sam_header.h"
17 KSTREAM_INIT(gzFile, gzread, 16384)
18 KHASH_MAP_INIT_STR(ref, uint64_t)
20 void bam_init_header_hash(bam_header_t *header);
21 void bam_destroy_header_hash(bam_header_t *header);
22 int32_t bam_get_tid(const bam_header_t *header, const char *seq_name);
24 unsigned char bam_nt16_table[256] = {
25 15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
26 15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
27 15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
28 1, 2, 4, 8, 15,15,15,15, 15,15,15,15, 15, 0 /*=*/,15,15,
29 15, 1,14, 2, 13,15,15, 4, 11,15,15,12, 15, 3,15,15,
30 15,15, 5, 6, 8,15, 7, 9, 15,10,15,15, 15,15,15,15,
31 15, 1,14, 2, 13,15,15, 4, 11,15,15,12, 15, 3,15,15,
32 15,15, 5, 6, 8,15, 7, 9, 15,10,15,15, 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,
37 15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
38 15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
39 15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
40 15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15
43 unsigned short bam_char2flag_table[256] = {
44 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
45 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
46 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
47 0,BAM_FREAD1,BAM_FREAD2,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
48 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
49 BAM_FPROPER_PAIR,0,BAM_FMREVERSE,0, 0,BAM_FMUNMAP,0,0, 0,0,0,0, 0,0,0,0,
50 0,0,0,0, BAM_FDUP,0,BAM_FQCFAIL,0, 0,0,0,0, 0,0,0,0,
51 BAM_FPAIRED,0,BAM_FREVERSE,BAM_FSECONDARY, 0,BAM_FUNMAP,0,0, 0,0,0,0, 0,0,0,0,
52 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
53 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
54 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
55 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
56 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
57 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
58 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
59 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0
62 char *bam_nt16_rev_table = "=ACMGRSVTWYHKDBN";
72 char **__bam_get_lines(const char *fn, int *_n) // for bam_plcmd.c only
75 int n = 0, dret, m = 0;
76 gzFile fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
79 str = (kstring_t*)calloc(1, sizeof(kstring_t));
81 while (ks_getuntil(ks, '\n', str, &dret) > 0) {
84 list = (char**)realloc(list, m * sizeof(char*));
86 if (str->s[str->l-1] == '\r')
87 str->s[--str->l] = '\0';
88 s = list[n++] = (char*)calloc(str->l + 1, 1);
93 free(str->s); free(str);
98 static bam_header_t *hash2header(const kh_ref_t *hash)
100 bam_header_t *header;
102 header = bam_header_init();
103 header->n_targets = kh_size(hash);
104 header->target_name = (char**)calloc(kh_size(hash), sizeof(char*));
105 header->target_len = (uint32_t*)calloc(kh_size(hash), 4);
106 for (k = kh_begin(hash); k != kh_end(hash); ++k) {
107 if (kh_exist(hash, k)) {
108 int i = (int)kh_value(hash, k);
109 header->target_name[i] = (char*)kh_key(hash, k);
110 header->target_len[i] = kh_value(hash, k)>>32;
113 bam_init_header_hash(header);
116 bam_header_t *sam_header_read2(const char *fn)
118 bam_header_t *header;
119 int c, dret, ret, error = 0;
125 if (fn == 0) return 0;
126 fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
127 if (fp == 0) return 0;
130 str = (kstring_t*)calloc(1, sizeof(kstring_t));
131 while (ks_getuntil(ks, 0, str, &dret) > 0) {
132 char *s = strdup(str->s);
135 ks_getuntil(ks, 0, str, &dret);
137 k = kh_put(ref, hash, s, &ret);
139 fprintf(stderr, "[sam_header_read2] duplicated sequence name: %s\n", s);
142 kh_value(hash, k) = (uint64_t)len<<32 | i;
144 while ((c = ks_getc(ks)) != '\n' && c != -1);
148 free(str->s); free(str);
149 fprintf(stderr, "[sam_header_read2] %d sequences loaded.\n", kh_size(hash));
151 header = hash2header(hash);
152 kh_destroy(ref, hash);
155 static inline uint8_t *alloc_data(bam1_t *b, int size)
157 if (b->m_data < size) {
159 kroundup32(b->m_data);
160 b->data = (uint8_t*)realloc(b->data, b->m_data);
164 static inline void parse_error(int64_t n_lines, const char * __restrict msg)
166 fprintf(stderr, "Parse error at line %lld: %s\n", (long long)n_lines, msg);
169 static inline void append_text(bam_header_t *header, kstring_t *str)
171 size_t x = header->l_text, y = header->l_text + str->l + 2; // 2 = 1 byte dret + 1 byte null
172 kroundup32(x); kroundup32(y);
176 header->text = (char*)realloc(header->text, y);
179 fprintf(stderr,"realloc failed to alloc %ld bytes\n", y);
184 if ( header->l_text+str->l+1 >= header->n_text )
186 fprintf(stderr,"append_text FIXME: %ld>=%ld, x=%ld,y=%ld\n", header->l_text+str->l+1,header->n_text,x,y);
189 strncpy(header->text + header->l_text, str->s, str->l+1); // we cannot use strcpy() here.
190 header->l_text += str->l + 1;
191 header->text[header->l_text] = 0;
194 int sam_header_parse(bam_header_t *h)
198 free(h->target_len); free(h->target_name);
199 h->n_targets = 0; h->target_len = 0; h->target_name = 0;
200 if (h->l_text < 3) return 0;
201 if (h->dict == 0) h->dict = sam_header_parse2(h->text);
202 tmp = sam_header2list(h->dict, "SQ", "SN", &h->n_targets);
203 if (h->n_targets == 0) return 0;
204 h->target_name = calloc(h->n_targets, sizeof(void*));
205 for (i = 0; i < h->n_targets; ++i)
206 h->target_name[i] = strdup(tmp[i]);
208 tmp = sam_header2list(h->dict, "SQ", "LN", &h->n_targets);
209 h->target_len = calloc(h->n_targets, 4);
210 for (i = 0; i < h->n_targets; ++i)
211 h->target_len[i] = atoi(tmp[i]);
216 bam_header_t *sam_header_read(tamFile fp)
219 bam_header_t *header = bam_header_init();
220 kstring_t *str = fp->str;
221 while ((ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret)) >= 0 && str->s[0] == '@') { // skip header
222 str->s[str->l] = dret; // note that str->s is NOT null terminated!!
223 append_text(header, str);
225 ret = ks_getuntil(fp->ks, '\n', str, &dret);
226 str->s[str->l] = '\n'; // NOT null terminated!!
227 append_text(header, str);
231 sam_header_parse(header);
232 bam_init_header_hash(header);
237 int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b)
239 int ret, doff, doff0, dret, z = 0;
240 bam1_core_t *c = &b->core;
241 kstring_t *str = fp->str;
242 kstream_t *ks = fp->ks;
248 do { // special consideration for empty lines
249 ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret);
250 if (ret >= 0) z += str->l + 1;
253 if (ret < 0) return -1;
258 c->l_qname = strlen(str->s) + 1;
259 memcpy(alloc_data(b, doff + c->l_qname) + doff, str->s, c->l_qname);
265 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
266 flag = strtol((char*)str->s, &s, 0);
267 if (*s) { // not the end of the string
269 for (s = str->s; *s; ++s)
270 flag |= bam_char2flag_table[(int)*s];
275 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->tid = bam_get_tid(header, str->s);
276 if (c->tid < 0 && strcmp(str->s, "*")) {
277 if (header->n_targets == 0) {
278 fprintf(stderr, "[sam_read1] missing header? Abort!\n");
280 } else fprintf(stderr, "[sam_read1] reference '%s' is recognized as '*'.\n", str->s);
282 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->pos = isdigit(str->s[0])? atoi(str->s) - 1 : -1;
283 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->qual = isdigit(str->s[0])? atoi(str->s) : 0;
284 if (ret < 0) return -2;
291 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -3;
293 if (str->s[0] != '*') {
294 for (s = str->s; *s; ++s) {
295 if ((isalpha(*s)) || (*s=='=')) ++c->n_cigar;
296 else if (!isdigit(*s)) parse_error(fp->n_lines, "invalid CIGAR character");
298 b->data = alloc_data(b, doff + c->n_cigar * 4);
299 for (i = 0, s = str->s; i != c->n_cigar; ++i) {
300 x = strtol(s, &t, 10);
302 if (op == 'M') op = BAM_CMATCH;
303 else if (op == 'I') op = BAM_CINS;
304 else if (op == 'D') op = BAM_CDEL;
305 else if (op == 'N') op = BAM_CREF_SKIP;
306 else if (op == 'S') op = BAM_CSOFT_CLIP;
307 else if (op == 'H') op = BAM_CHARD_CLIP;
308 else if (op == 'P') op = BAM_CPAD;
309 else if (op == '=') op = BAM_CEQUAL;
310 else if (op == 'X') op = BAM_CDIFF;
311 else parse_error(fp->n_lines, "invalid CIGAR operation");
313 bam1_cigar(b)[i] = x << BAM_CIGAR_SHIFT | op;
315 if (*s) parse_error(fp->n_lines, "unmatched CIGAR operation");
316 c->bin = bam_reg2bin(c->pos, bam_calend(c, bam1_cigar(b)));
317 doff += c->n_cigar * 4;
319 if (!(c->flag&BAM_FUNMAP)) {
320 fprintf(stderr, "Parse warning at line %lld: mapped sequence without CIGAR\n", (long long)fp->n_lines);
321 c->flag |= BAM_FUNMAP;
323 c->bin = bam_reg2bin(c->pos, c->pos + 1);
326 { // mtid, mpos, isize
327 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
328 c->mtid = strcmp(str->s, "=")? bam_get_tid(header, str->s) : c->tid;
329 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
330 c->mpos = isdigit(str->s[0])? atoi(str->s) - 1 : -1;
331 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
332 c->isize = (str->s[0] == '-' || isdigit(str->s[0]))? atoi(str->s) : 0;
333 if (ret < 0) return -4;
338 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -5; // seq
340 if (strcmp(str->s, "*")) {
341 c->l_qseq = strlen(str->s);
342 if (c->n_cigar && c->l_qseq != (int32_t)bam_cigar2qlen(c, bam1_cigar(b))) {
343 fprintf(stderr, "Line %ld, sequence length %i vs %i from CIGAR\n",
344 (long)fp->n_lines, c->l_qseq, (int32_t)bam_cigar2qlen(c, bam1_cigar(b)));
345 parse_error(fp->n_lines, "CIGAR and sequence length are inconsistent");
347 p = (uint8_t*)alloc_data(b, doff + c->l_qseq + (c->l_qseq+1)/2) + doff;
348 memset(p, 0, (c->l_qseq+1)/2);
349 for (i = 0; i < c->l_qseq; ++i)
350 p[i/2] |= bam_nt16_table[(int)str->s[i]] << 4*(1-i%2);
351 } else c->l_qseq = 0;
352 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -6; // qual
354 if (strcmp(str->s, "*") && c->l_qseq != strlen(str->s))
355 parse_error(fp->n_lines, "sequence and quality are inconsistent");
356 p += (c->l_qseq+1)/2;
357 if (strcmp(str->s, "*") == 0) for (i = 0; i < c->l_qseq; ++i) p[i] = 0xff;
358 else for (i = 0; i < c->l_qseq; ++i) p[i] = str->s[i] - 33;
359 doff += c->l_qseq + (c->l_qseq+1)/2;
362 if (dret != '\n' && dret != '\r') { // aux
363 while (ks_getuntil(ks, KS_SEP_TAB, str, &dret) >= 0) {
364 uint8_t *s, type, key[2];
366 if (str->l < 6 || str->s[2] != ':' || str->s[4] != ':')
367 parse_error(fp->n_lines, "missing colon in auxiliary data");
368 key[0] = str->s[0]; key[1] = str->s[1];
370 s = alloc_data(b, doff + 3) + doff;
371 s[0] = key[0]; s[1] = key[1]; s += 2; doff += 2;
372 if (type == 'A' || type == 'a' || type == 'c' || type == 'C') { // c and C for backward compatibility
373 s = alloc_data(b, doff + 2) + doff;
374 *s++ = 'A'; *s = str->s[5];
376 } else if (type == 'I' || type == 'i') {
378 s = alloc_data(b, doff + 5) + doff;
379 x = (long long)atoll(str->s + 5);
382 *s++ = 'c'; *(int8_t*)s = (int8_t)x;
384 } else if (x >= -32767) {
385 *s++ = 's'; *(int16_t*)s = (int16_t)x;
388 *s++ = 'i'; *(int32_t*)s = (int32_t)x;
390 if (x < -2147483648ll)
391 fprintf(stderr, "Parse warning at line %lld: integer %lld is out of range.",
392 (long long)fp->n_lines, x);
396 *s++ = 'C'; *s++ = (uint8_t)x;
398 } else if (x <= 65535) {
399 *s++ = 'S'; *(uint16_t*)s = (uint16_t)x;
402 *s++ = 'I'; *(uint32_t*)s = (uint32_t)x;
404 if (x > 4294967295ll)
405 fprintf(stderr, "Parse warning at line %lld: integer %lld is out of range.",
406 (long long)fp->n_lines, x);
409 } else if (type == 'f') {
410 s = alloc_data(b, doff + 5) + doff;
412 *(float*)s = (float)atof(str->s + 5);
414 } else if (type == 'd') {
415 s = alloc_data(b, doff + 9) + doff;
417 *(float*)s = (float)atof(str->s + 9);
419 } else if (type == 'Z' || type == 'H') {
420 int size = 1 + (str->l - 5) + 1;
421 if (type == 'H') { // check whether the hex string is valid
423 if ((str->l - 5) % 2 == 1) parse_error(fp->n_lines, "length of the hex string not even");
424 for (i = 0; i < str->l - 5; ++i) {
425 int c = toupper(str->s[5 + i]);
426 if (!((c >= '0' && c <= '9') || (c >= 'A' && c <= 'F')))
427 parse_error(fp->n_lines, "invalid hex character");
430 s = alloc_data(b, doff + size) + doff;
432 memcpy(s, str->s + 5, str->l - 5);
435 } else if (type == 'B') {
436 int32_t n = 0, Bsize, k = 0, size;
438 if (str->l < 8) parse_error(fp->n_lines, "too few values in aux type B");
439 Bsize = bam_aux_type2size(str->s[5]); // the size of each element
440 for (p = (char*)str->s + 6; *p; ++p) // count the number of elements in the array
442 p = str->s + 7; // now p points to the first number in the array
443 size = 6 + Bsize * n; // total number of bytes allocated to this tag
444 s = alloc_data(b, doff + 6 * Bsize * n) + doff; // allocate memory
445 *s++ = 'B'; *s++ = str->s[5];
446 memcpy(s, &n, 4); s += 4; // write the number of elements
447 if (str->s[5] == 'c') while (p < str->s + str->l) ((int8_t*)s)[k++] = (int8_t)strtol(p, &p, 0), ++p;
448 else if (str->s[5] == 'C') while (p < str->s + str->l) ((uint8_t*)s)[k++] = (uint8_t)strtol(p, &p, 0), ++p;
449 else if (str->s[5] == 's') while (p < str->s + str->l) ((int16_t*)s)[k++] = (int16_t)strtol(p, &p, 0), ++p; // FIXME: avoid unaligned memory
450 else if (str->s[5] == 'S') while (p < str->s + str->l) ((uint16_t*)s)[k++] = (uint16_t)strtol(p, &p, 0), ++p;
451 else if (str->s[5] == 'i') while (p < str->s + str->l) ((int32_t*)s)[k++] = (int32_t)strtol(p, &p, 0), ++p;
452 else if (str->s[5] == 'I') while (p < str->s + str->l) ((uint32_t*)s)[k++] = (uint32_t)strtol(p, &p, 0), ++p;
453 else if (str->s[5] == 'f') while (p < str->s + str->l) ((float*)s)[k++] = (float)strtod(p, &p), ++p;
454 else parse_error(fp->n_lines, "unrecognized array type");
455 s += Bsize * n; doff += size;
456 } else parse_error(fp->n_lines, "unrecognized type");
457 if (dret == '\n' || dret == '\r') break;
460 b->l_aux = doff - doff0;
465 tamFile sam_open(const char *fn)
468 gzFile gzfp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "rb") : gzopen(fn, "rb");
469 if (gzfp == 0) return 0;
470 fp = (tamFile)calloc(1, sizeof(struct __tamFile_t));
471 fp->str = (kstring_t*)calloc(1, sizeof(kstring_t));
473 fp->ks = ks_init(fp->fp);
477 void sam_close(tamFile fp)
482 free(fp->str->s); free(fp->str);