]> git.donarmstrong.com Git - samtools.git/blob - bam_import.c
* samtools-0.1.4-8 (r338)
[samtools.git] / bam_import.c
1 #include <zlib.h>
2 #include <stdio.h>
3 #include <ctype.h>
4 #include <string.h>
5 #include <stdlib.h>
6 #include <unistd.h>
7 #include <assert.h>
8 #include "kstring.h"
9 #include "bam.h"
10 #include "kseq.h"
11 #include "khash.h"
12
13 KSTREAM_INIT(gzFile, gzread, 8192)
14 KHASH_MAP_INIT_STR(ref, uint64_t)
15
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);
19
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
37 };
38
39 char *bam_nt16_rev_table = "=ACMGRSVTWYHKDBN";
40
41 struct __tamFile_t {
42         gzFile fp;
43         kstream_t *ks;
44         kstring_t *str;
45         uint64_t n_lines;
46         int is_first;
47 };
48
49 char **__bam_get_lines(const char *fn, int *_n) // for bam_plcmd.c only
50 {
51         char **list = 0, *s;
52         int n = 0, dret, m = 0;
53         gzFile fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
54         kstream_t *ks;
55         kstring_t *str;
56         str = (kstring_t*)calloc(1, sizeof(kstring_t));
57         ks = ks_init(fp);
58         while (ks_getuntil(ks, '\n', str, &dret) > 0) {
59                 if (n == m) {
60                         m = m? m << 1 : 16;
61                         list = (char**)realloc(list, m * sizeof(char*));
62                 }
63                 if (str->s[str->l-1] == '\r')
64                         str->s[--str->l] = '\0';
65                 s = list[n++] = (char*)calloc(str->l + 1, 1);
66                 strcpy(s, str->s);
67         }
68         ks_destroy(ks);
69         gzclose(fp);
70         free(str->s); free(str);
71         *_n = n;
72         return list;
73 }
74
75 static bam_header_t *hash2header(const kh_ref_t *hash)
76 {
77         bam_header_t *header;
78         khiter_t k;
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;
88                 }
89         }
90         bam_init_header_hash(header);
91         return header;
92 }
93 bam_header_t *sam_header_read2(const char *fn)
94 {
95         bam_header_t *header;
96         int c, dret, ret;
97         gzFile fp;
98         kstream_t *ks;
99         kstring_t *str;
100         kh_ref_t *hash;
101         khiter_t k;
102         hash = kh_init(ref);
103         fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
104         assert(fp);
105         ks = ks_init(fp);
106         str = (kstring_t*)calloc(1, sizeof(kstring_t));
107         while (ks_getuntil(ks, 0, str, &dret) > 0) {
108                 char *s = strdup(str->s);
109                 int len, i;
110                 i = kh_size(hash);
111                 ks_getuntil(ks, 0, str, &dret);
112                 len = atoi(str->s);
113                 k = kh_put(ref, hash, s, &ret);
114                 kh_value(hash, k) = (uint64_t)len<<32 | i;
115                 if (dret != '\n')
116                         while ((c = ks_getc(ks)) != '\n' && c != -1);
117         }
118         ks_destroy(ks);
119         gzclose(fp);
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);
124         return header;
125 }
126 static inline uint8_t *alloc_data(bam1_t *b, int size)
127 {
128         if (b->m_data < size) {
129                 b->m_data = size;
130                 kroundup32(b->m_data);
131                 b->data = (uint8_t*)realloc(b->data, b->m_data);
132         }
133         return b->data;
134 }
135 static inline void parse_error(int64_t n_lines, const char * __restrict msg)
136 {
137         fprintf(stderr, "Parse error at line %lld: %s\n", (long long)n_lines, msg);
138         abort();
139 }
140 static inline void append_text(bam_header_t *header, kstring_t *str)
141 {
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;
148 }
149
150 int sam_header_parse_rg(bam_header_t *h)
151 {
152         kstring_t *rgid, *rglib;
153         char *p, *q, *s, *r;
154         int n = 0;
155
156         bam_strmap_destroy(h->rg2lib);
157         // parse @RG lines
158         h->rg2lib = bam_strmap_init();
159         rgid = calloc(1, sizeof(kstring_t));
160         rglib = calloc(1, sizeof(kstring_t));
161         s = h->text;
162         while ((s = strstr(s, "@RG")) != 0) {
163                 if (rgid->l && rglib->l) {
164                         bam_strmap_put(h->rg2lib, rgid->s, rglib->s);
165                         ++n;
166                 }
167                 rgid->l = rglib->l = 0;
168                 s += 3;
169                 r = s;
170                 if ((p = strstr(s, "ID:")) != 0) {
171                         q = p + 3;
172                         for (p = q; *p && *p != '\t' && *p != '\r' && *p != '\n'; ++p);
173                         kputsn(q, p - q, rgid);
174                 } else {
175                         fprintf(stderr, "[bam_header_parse] missing ID tag in @RG lines.\n");
176                         break;
177                 }
178                 if (r < p) r = p;
179                 if ((p = strstr(s, "LB:")) != 0) {
180                         q = p + 3;
181                         for (p = q; *p && *p != '\t' && *p != '\r' && *p != '\n'; ++p);
182                         kputsn(q, p - q, rglib);
183                 } else {
184                         fprintf(stderr, "[bam_header_parse] missing LB tag in @RG lines.\n");
185                         break;
186                 }
187                 if (r < p) r = p;
188                 s = r + 3;
189         }
190         if (rgid->l && rglib->l) {
191                 bam_strmap_put(h->rg2lib, rgid->s, rglib->s);
192                 ++n;
193         }
194         free(rgid->s); free(rgid);
195         free(rglib->s); free(rglib);
196         return n;
197 }
198
199 int sam_header_parse(bam_header_t *h)
200 {
201         int i;
202         char *s, *p, *q, *r;
203
204         // free
205         free(h->target_len); free(h->target_name);
206         h->n_targets = 0; h->target_len = 0; h->target_name = 0;
207         if (h->l_text < 3) return 0;
208         // count number of @SQ
209         s = h->text;
210         while ((s = strstr(s, "@SQ")) != 0) {
211                 ++h->n_targets;
212                 s += 3;
213         }
214         if (h->n_targets == 0) return 0;
215         h->target_len = (uint32_t*)calloc(h->n_targets, 4);
216         h->target_name = (char**)calloc(h->n_targets, sizeof(void*));
217         // parse @SQ lines
218         i = 0;
219         s = h->text;
220         while ((s = strstr(s, "@SQ")) != 0) {
221                 s += 3;
222                 r = s;
223                 if ((p = strstr(s, "SN:")) != 0) {
224                         q = p + 3;
225                         for (p = q; *p && *p != '\t' && *p != '\r' && *p != '\n'; ++p);
226                         h->target_name[i] = (char*)calloc(p - q + 1, 1);
227                         strncpy(h->target_name[i], q, p - q);
228                 } else goto header_err_ret;
229                 if (r < p) r = p;
230                 if ((p = strstr(s, "LN:")) != 0) h->target_len[i] = strtol(p + 3, 0, 10);
231                 else goto header_err_ret;
232                 if (r < p) r = p;
233                 s = r + 3;
234                 ++i;
235         }
236         sam_header_parse_rg(h);
237         return h->n_targets;
238
239 header_err_ret:
240         fprintf(stderr, "[bam_header_parse] missing SN or LN tag in @SQ lines.\n");
241         free(h->target_len); free(h->target_name);
242         h->n_targets = 0; h->target_len = 0; h->target_name = 0;
243         return 0;
244 }
245
246 bam_header_t *sam_header_read(tamFile fp)
247 {
248         int ret, dret;
249         bam_header_t *header = bam_header_init();
250         kstring_t *str = fp->str;
251         while ((ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret)) >= 0 && str->s[0] == '@') { // skip header
252                 str->s[str->l] = dret; // note that str->s is NOT null terminated!!
253                 append_text(header, str);
254                 if (dret != '\n') {
255                         ret = ks_getuntil(fp->ks, '\n', str, &dret);
256                         str->s[str->l] = '\n'; // NOT null terminated!!
257                         append_text(header, str);
258                 }
259                 ++fp->n_lines;
260         }
261         sam_header_parse(header);
262         bam_init_header_hash(header);
263         fp->is_first = 1;
264         return header;
265 }
266
267 int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b)
268 {
269         int ret, doff, doff0, dret, z = 0;
270         bam1_core_t *c = &b->core;
271         kstring_t *str = fp->str;
272         kstream_t *ks = fp->ks;
273
274         if (fp->is_first) {
275                 fp->is_first = 0;
276                 ret = str->l;
277         } else {
278                 do { // special consideration for empty lines
279                         ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret);
280                         if (ret >= 0) z += str->l + 1;
281                 } while (ret == 0);
282         }
283         if (ret < 0) return -1;
284         ++fp->n_lines;
285         doff = 0;
286
287         { // name
288                 c->l_qname = strlen(str->s) + 1;
289                 memcpy(alloc_data(b, doff + c->l_qname) + doff, str->s, c->l_qname);
290                 doff += c->l_qname;
291         }
292         { // flag, tid, pos, qual
293                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->flag = atoi(str->s);
294                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->tid = bam_get_tid(header, str->s);
295                 if (c->tid < 0 && strcmp(str->s, "*")) {
296                         if (header->n_targets == 0) {
297                                 fprintf(stderr, "[sam_read1] missing header? Abort!\n");
298                                 exit(1);
299                         } else fprintf(stderr, "[sam_read1] reference '%s' is recognized as '*'.\n", str->s);
300                 }
301                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->pos = isdigit(str->s[0])? atoi(str->s) - 1 : -1;
302                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->qual = isdigit(str->s[0])? atoi(str->s) : 0;
303                 if (ret < 0) return -2;
304         }
305         { // cigar
306                 char *s, *t;
307                 int i, op;
308                 long x;
309                 c->n_cigar = 0;
310                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -3;
311                 z += str->l + 1;
312                 if (str->s[0] != '*') {
313                         for (s = str->s; *s; ++s) {
314                                 if (isalpha(*s)) ++c->n_cigar;
315                                 else if (!isdigit(*s)) parse_error(fp->n_lines, "invalid CIGAR character");
316                         }
317                         b->data = alloc_data(b, doff + c->n_cigar * 4);
318                         for (i = 0, s = str->s; i != c->n_cigar; ++i) {
319                                 x = strtol(s, &t, 10);
320                                 op = toupper(*t);
321                                 if (op == 'M') op = BAM_CMATCH;
322                                 else if (op == 'I') op = BAM_CINS;
323                                 else if (op == 'D') op = BAM_CDEL;
324                                 else if (op == 'N') op = BAM_CREF_SKIP;
325                                 else if (op == 'S') op = BAM_CSOFT_CLIP;
326                                 else if (op == 'H') op = BAM_CHARD_CLIP;
327                                 else if (op == 'P') op = BAM_CPAD;
328                                 else parse_error(fp->n_lines, "invalid CIGAR operation");
329                                 s = t + 1;
330                                 bam1_cigar(b)[i] = x << BAM_CIGAR_SHIFT | op;
331                         }
332                         if (*s) parse_error(fp->n_lines, "unmatched CIGAR operation");
333                         c->bin = bam_reg2bin(c->pos, bam_calend(c, bam1_cigar(b)));
334                         doff += c->n_cigar * 4;
335                 } else c->bin = bam_reg2bin(c->pos, c->pos + 1);
336         }
337         { // mtid, mpos, isize
338                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
339                 c->mtid = strcmp(str->s, "=")? bam_get_tid(header, str->s) : c->tid;
340                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
341                 c->mpos = isdigit(str->s[0])? atoi(str->s) - 1 : -1;
342                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
343                 c->isize = (str->s[0] == '-' || isdigit(str->s[0]))? atoi(str->s) : 0;
344                 if (ret < 0) return -4;
345         }
346         { // seq and qual
347                 int i;
348                 uint8_t *p;
349                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -5; // seq
350                 z += str->l + 1;
351                 c->l_qseq = strlen(str->s);
352                 if (c->n_cigar && c->l_qseq != (int32_t)bam_cigar2qlen(c, bam1_cigar(b)))
353                         parse_error(fp->n_lines, "CIGAR and sequence length are inconsistent");
354                 p = (uint8_t*)alloc_data(b, doff + c->l_qseq + (c->l_qseq+1)/2) + doff;
355                 bzero(p, (c->l_qseq+1)/2);
356                 for (i = 0; i < c->l_qseq; ++i)
357                         p[i/2] |= bam_nt16_table[(int)str->s[i]] << 4*(1-i%2);
358                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -6; // qual
359                 z += str->l + 1;
360                 if (strcmp(str->s, "*") && c->l_qseq != strlen(str->s))
361                         parse_error(fp->n_lines, "sequence and quality are inconsistent");
362                 p += (c->l_qseq+1)/2;
363                 if (strcmp(str->s, "*") == 0) for (i = 0; i < c->l_qseq; ++i) p[i] = 0xff;
364                 else for (i = 0; i < c->l_qseq; ++i) p[i] = str->s[i] - 33;
365                 doff += c->l_qseq + (c->l_qseq+1)/2;
366         }
367         doff0 = doff;
368         if (dret != '\n' && dret != '\r') { // aux
369                 while (ks_getuntil(ks, KS_SEP_TAB, str, &dret) >= 0) {
370                         uint8_t *s, type, key[2];
371                         z += str->l + 1;
372                         if (str->l < 6 || str->s[2] != ':' || str->s[4] != ':')
373                                 parse_error(fp->n_lines, "missing colon in auxiliary data");
374                         key[0] = str->s[0]; key[1] = str->s[1];
375                         type = str->s[3];
376                         s = alloc_data(b, doff + 3) + doff;
377                         s[0] = key[0]; s[1] = key[1]; s += 2; doff += 2;
378                         if (type == 'A' || type == 'a' || type == 'c' || type == 'C') { // c and C for backward compatibility
379                                 s = alloc_data(b, doff + 2) + doff;
380                                 *s++ = 'A'; *s = str->s[5];
381                                 doff += 2;
382                         } else if (type == 'I' || type == 'i') {
383                                 long long x;
384                                 s = alloc_data(b, doff + 5) + doff;
385                                 x = (long long)atoll(str->s + 5);
386                                 if (x < 0) {
387                                         if (x >= -127) {
388                                                 *s++ = 'c'; *(int8_t*)s = (int8_t)x;
389                                                 s += 1; doff += 2;
390                                         } else if (x >= -32767) {
391                                                 *s++ = 's'; *(int16_t*)s = (int16_t)x;
392                                                 s += 2; doff += 3;
393                                         } else {
394                                                 *s++ = 'i'; *(int32_t*)s = (int32_t)x;
395                                                 s += 4; doff += 5;
396                                                 if (x < -2147483648ll)
397                                                         fprintf(stderr, "Parse warning at line %lld: integer %lld is out of range.",
398                                                                         (long long)fp->n_lines, x);
399                                         }
400                                 } else {
401                                         if (x <= 255) {
402                                                 *s++ = 'C'; *s++ = (uint8_t)x;
403                                                 doff += 2;
404                                         } else if (x <= 65535) {
405                                                 *s++ = 'S'; *(uint16_t*)s = (uint16_t)x;
406                                                 s += 2; doff += 3;
407                                         } else {
408                                                 *s++ = 'I'; *(uint32_t*)s = (uint32_t)x;
409                                                 s += 4; doff += 5;
410                                                 if (x > 4294967295ll)
411                                                         fprintf(stderr, "Parse warning at line %lld: integer %lld is out of range.",
412                                                                         (long long)fp->n_lines, x);
413                                         }
414                                 }
415                         } else if (type == 'f') {
416                                 s = alloc_data(b, doff + 5) + doff;
417                                 *s++ = 'f';
418                                 *(float*)s = (float)atof(str->s + 5);
419                                 s += 4; doff += 5;
420                         } else if (type == 'd') {
421                                 s = alloc_data(b, doff + 9) + doff;
422                                 *s++ = 'd';
423                                 *(float*)s = (float)atof(str->s + 9);
424                                 s += 8; doff += 9;
425                         } else if (type == 'Z' || type == 'H') {
426                                 int size = 1 + (str->l - 5) + 1;
427                                 if (type == 'H') { // check whether the hex string is valid
428                                         int i;
429                                         if ((str->l - 5) % 2 == 1) parse_error(fp->n_lines, "length of the hex string not even");
430                                         for (i = 0; i < str->l - 5; ++i) {
431                                                 int c = toupper(str->s[5 + i]);
432                                                 if (!((c >= '0' && c <= '9') || (c >= 'A' && c <= 'F')))
433                                                         parse_error(fp->n_lines, "invalid hex character");
434                                         }
435                                 }
436                                 s = alloc_data(b, doff + size) + doff;
437                                 *s++ = type;
438                                 memcpy(s, str->s + 5, str->l - 5);
439                                 s[str->l - 5] = 0;
440                                 doff += size;
441                         } else parse_error(fp->n_lines, "unrecognized type");
442                         if (dret == '\n' || dret == '\r') break;
443                 }
444         }
445         b->l_aux = doff - doff0;
446         b->data_len = doff;
447         return z;
448 }
449
450 tamFile sam_open(const char *fn)
451 {
452         tamFile fp;
453         fp = (tamFile)calloc(1, sizeof(struct __tamFile_t));
454         fp->str = (kstring_t*)calloc(1, sizeof(kstring_t));
455         fp->fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
456         fp->ks = ks_init(fp->fp);
457         return fp;
458 }
459
460 void sam_close(tamFile fp)
461 {
462         if (fp) {
463                 ks_destroy(fp->ks);
464                 gzclose(fp->fp);
465                 free(fp->str->s); free(fp->str);
466                 free(fp);
467         }
468 }