]> git.donarmstrong.com Git - samtools.git/blob - bam_import.c
* samtools-0.1.5-17 (r416)
[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 unsigned short bam_char2flag_table[256] = {
40         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
41         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
42         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
43         0,BAM_FREAD1,BAM_FREAD2,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
44         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
45         BAM_FPROPER_PAIR,0,BAM_FMREVERSE,0, 0,BAM_FMUNMAP,0,0, 0,0,0,0, 0,0,0,0,
46         0,0,0,0, BAM_FDUP,0,BAM_FQCFAIL,0, 0,0,0,0, 0,0,0,0,
47         BAM_FPAIRED,0,BAM_FREVERSE,BAM_FSECONDARY, 0,BAM_FUNMAP,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         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
50         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
51         0,0,0,0, 0,0,0,0, 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 };
57
58 char *bam_nt16_rev_table = "=ACMGRSVTWYHKDBN";
59
60 struct __tamFile_t {
61         gzFile fp;
62         kstream_t *ks;
63         kstring_t *str;
64         uint64_t n_lines;
65         int is_first;
66 };
67
68 char **__bam_get_lines(const char *fn, int *_n) // for bam_plcmd.c only
69 {
70         char **list = 0, *s;
71         int n = 0, dret, m = 0;
72         gzFile fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
73         kstream_t *ks;
74         kstring_t *str;
75         str = (kstring_t*)calloc(1, sizeof(kstring_t));
76         ks = ks_init(fp);
77         while (ks_getuntil(ks, '\n', str, &dret) > 0) {
78                 if (n == m) {
79                         m = m? m << 1 : 16;
80                         list = (char**)realloc(list, m * sizeof(char*));
81                 }
82                 if (str->s[str->l-1] == '\r')
83                         str->s[--str->l] = '\0';
84                 s = list[n++] = (char*)calloc(str->l + 1, 1);
85                 strcpy(s, str->s);
86         }
87         ks_destroy(ks);
88         gzclose(fp);
89         free(str->s); free(str);
90         *_n = n;
91         return list;
92 }
93
94 static bam_header_t *hash2header(const kh_ref_t *hash)
95 {
96         bam_header_t *header;
97         khiter_t k;
98         header = bam_header_init();
99         header->n_targets = kh_size(hash);
100         header->target_name = (char**)calloc(kh_size(hash), sizeof(char*));
101         header->target_len = (uint32_t*)calloc(kh_size(hash), 4);
102         for (k = kh_begin(hash); k != kh_end(hash); ++k) {
103                 if (kh_exist(hash, k)) {
104                         int i = (int)kh_value(hash, k);
105                         header->target_name[i] = (char*)kh_key(hash, k);
106                         header->target_len[i] = kh_value(hash, k)>>32;
107                 }
108         }
109         bam_init_header_hash(header);
110         return header;
111 }
112 bam_header_t *sam_header_read2(const char *fn)
113 {
114         bam_header_t *header;
115         int c, dret, ret;
116         gzFile fp;
117         kstream_t *ks;
118         kstring_t *str;
119         kh_ref_t *hash;
120         khiter_t k;
121         hash = kh_init(ref);
122         fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
123         assert(fp);
124         ks = ks_init(fp);
125         str = (kstring_t*)calloc(1, sizeof(kstring_t));
126         while (ks_getuntil(ks, 0, str, &dret) > 0) {
127                 char *s = strdup(str->s);
128                 int len, i;
129                 i = kh_size(hash);
130                 ks_getuntil(ks, 0, str, &dret);
131                 len = atoi(str->s);
132                 k = kh_put(ref, hash, s, &ret);
133                 kh_value(hash, k) = (uint64_t)len<<32 | i;
134                 if (dret != '\n')
135                         while ((c = ks_getc(ks)) != '\n' && c != -1);
136         }
137         ks_destroy(ks);
138         gzclose(fp);
139         free(str->s); free(str);
140         fprintf(stderr, "[sam_header_read2] %d sequences loaded.\n", kh_size(hash));
141         header = hash2header(hash);
142         kh_destroy(ref, hash);
143         return header;
144 }
145 static inline uint8_t *alloc_data(bam1_t *b, int size)
146 {
147         if (b->m_data < size) {
148                 b->m_data = size;
149                 kroundup32(b->m_data);
150                 b->data = (uint8_t*)realloc(b->data, b->m_data);
151         }
152         return b->data;
153 }
154 static inline void parse_error(int64_t n_lines, const char * __restrict msg)
155 {
156         fprintf(stderr, "Parse error at line %lld: %s\n", (long long)n_lines, msg);
157         abort();
158 }
159 static inline void append_text(bam_header_t *header, kstring_t *str)
160 {
161         int x = header->l_text, y = header->l_text + str->l + 2; // 2 = 1 byte dret + 1 byte null
162         kroundup32(x); kroundup32(y);
163         if (x < y) header->text = (char*)realloc(header->text, y);
164         strncpy(header->text + header->l_text, str->s, str->l+1); // we cannot use strcpy() here.
165         header->l_text += str->l + 1;
166         header->text[header->l_text] = 0;
167 }
168
169 int sam_header_parse_rg(bam_header_t *h)
170 {
171         kstring_t *rgid, *rglib;
172         char *p, *q, *s, *r;
173         int n = 0;
174
175         // free
176         if (h == 0) return 0;
177         bam_strmap_destroy(h->rg2lib); h->rg2lib = 0;
178         if (h->l_text < 3) return 0;
179         // parse @RG lines
180         h->rg2lib = bam_strmap_init();
181         rgid = calloc(1, sizeof(kstring_t));
182         rglib = calloc(1, sizeof(kstring_t));
183         s = h->text;
184         while ((s = strstr(s, "@RG")) != 0) {
185                 if (rgid->l && rglib->l) {
186                         bam_strmap_put(h->rg2lib, rgid->s, rglib->s);
187                         ++n;
188                 }
189                 rgid->l = rglib->l = 0;
190                 s += 3;
191                 r = s;
192                 if ((p = strstr(s, "ID:")) != 0) {
193                         q = p + 3;
194                         for (p = q; *p && *p != '\t' && *p != '\r' && *p != '\n'; ++p);
195                         kputsn(q, p - q, rgid);
196                 } else {
197                         fprintf(stderr, "[bam_header_parse] missing ID tag in @RG lines.\n");
198                         break;
199                 }
200                 if (r < p) r = p;
201                 if ((p = strstr(s, "LB:")) != 0) {
202                         q = p + 3;
203                         for (p = q; *p && *p != '\t' && *p != '\r' && *p != '\n'; ++p);
204                         kputsn(q, p - q, rglib);
205                 } else {
206                         fprintf(stderr, "[bam_header_parse] missing LB tag in @RG lines.\n");
207                         break;
208                 }
209                 if (r < p) r = p;
210                 s = r + 3;
211         }
212         if (rgid->l && rglib->l) {
213                 bam_strmap_put(h->rg2lib, rgid->s, rglib->s);
214                 ++n;
215         }
216         free(rgid->s); free(rgid);
217         free(rglib->s); free(rglib);
218         if (n == 0) {
219                 bam_strmap_destroy(h->rg2lib);
220                 h->rg2lib = 0;
221         }
222         return n;
223 }
224
225 int sam_header_parse(bam_header_t *h)
226 {
227         int i;
228         char *s, *p, *q, *r;
229
230         // free
231         free(h->target_len); free(h->target_name);
232         h->n_targets = 0; h->target_len = 0; h->target_name = 0;
233         if (h->l_text < 3) return 0;
234         // count number of @SQ
235         s = h->text;
236         while ((s = strstr(s, "@SQ")) != 0) {
237                 ++h->n_targets;
238                 s += 3;
239         }
240         if (h->n_targets == 0) return 0;
241         h->target_len = (uint32_t*)calloc(h->n_targets, 4);
242         h->target_name = (char**)calloc(h->n_targets, sizeof(void*));
243         // parse @SQ lines
244         i = 0;
245         s = h->text;
246         while ((s = strstr(s, "@SQ")) != 0) {
247                 s += 3;
248                 r = s;
249                 if ((p = strstr(s, "SN:")) != 0) {
250                         q = p + 3;
251                         for (p = q; *p && *p != '\t' && *p != '\r' && *p != '\n'; ++p);
252                         h->target_name[i] = (char*)calloc(p - q + 1, 1);
253                         strncpy(h->target_name[i], q, p - q);
254                 } else goto header_err_ret;
255                 if (r < p) r = p;
256                 if ((p = strstr(s, "LN:")) != 0) h->target_len[i] = strtol(p + 3, 0, 10);
257                 else goto header_err_ret;
258                 if (r < p) r = p;
259                 s = r + 3;
260                 ++i;
261         }
262         sam_header_parse_rg(h);
263         return h->n_targets;
264
265 header_err_ret:
266         fprintf(stderr, "[bam_header_parse] missing SN or LN tag in @SQ lines.\n");
267         free(h->target_len); free(h->target_name);
268         h->n_targets = 0; h->target_len = 0; h->target_name = 0;
269         return 0;
270 }
271
272 bam_header_t *sam_header_read(tamFile fp)
273 {
274         int ret, dret;
275         bam_header_t *header = bam_header_init();
276         kstring_t *str = fp->str;
277         while ((ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret)) >= 0 && str->s[0] == '@') { // skip header
278                 str->s[str->l] = dret; // note that str->s is NOT null terminated!!
279                 append_text(header, str);
280                 if (dret != '\n') {
281                         ret = ks_getuntil(fp->ks, '\n', str, &dret);
282                         str->s[str->l] = '\n'; // NOT null terminated!!
283                         append_text(header, str);
284                 }
285                 ++fp->n_lines;
286         }
287         sam_header_parse(header);
288         bam_init_header_hash(header);
289         fp->is_first = 1;
290         return header;
291 }
292
293 int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b)
294 {
295         int ret, doff, doff0, dret, z = 0;
296         bam1_core_t *c = &b->core;
297         kstring_t *str = fp->str;
298         kstream_t *ks = fp->ks;
299
300         if (fp->is_first) {
301                 fp->is_first = 0;
302                 ret = str->l;
303         } else {
304                 do { // special consideration for empty lines
305                         ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret);
306                         if (ret >= 0) z += str->l + 1;
307                 } while (ret == 0);
308         }
309         if (ret < 0) return -1;
310         ++fp->n_lines;
311         doff = 0;
312
313         { // name
314                 c->l_qname = strlen(str->s) + 1;
315                 memcpy(alloc_data(b, doff + c->l_qname) + doff, str->s, c->l_qname);
316                 doff += c->l_qname;
317         }
318         { // flag
319                 long flag;
320                 char *s;
321                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
322                 flag = strtol((char*)str->s, &s, 0);
323                 if (*s) { // not the end of the string
324                         flag = 0;
325                         for (s = str->s; *s; ++s)
326                                 flag |= bam_char2flag_table[(int)*s];
327                 }
328                 c->flag = flag;
329         }
330         { // tid, pos, qual
331                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->tid = bam_get_tid(header, str->s);
332                 if (c->tid < 0 && strcmp(str->s, "*")) {
333                         if (header->n_targets == 0) {
334                                 fprintf(stderr, "[sam_read1] missing header? Abort!\n");
335                                 exit(1);
336                         } else fprintf(stderr, "[sam_read1] reference '%s' is recognized as '*'.\n", str->s);
337                 }
338                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->pos = isdigit(str->s[0])? atoi(str->s) - 1 : -1;
339                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->qual = isdigit(str->s[0])? atoi(str->s) : 0;
340                 if (ret < 0) return -2;
341         }
342         { // cigar
343                 char *s, *t;
344                 int i, op;
345                 long x;
346                 c->n_cigar = 0;
347                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -3;
348                 z += str->l + 1;
349                 if (str->s[0] != '*') {
350                         for (s = str->s; *s; ++s) {
351                                 if (isalpha(*s)) ++c->n_cigar;
352                                 else if (!isdigit(*s)) parse_error(fp->n_lines, "invalid CIGAR character");
353                         }
354                         b->data = alloc_data(b, doff + c->n_cigar * 4);
355                         for (i = 0, s = str->s; i != c->n_cigar; ++i) {
356                                 x = strtol(s, &t, 10);
357                                 op = toupper(*t);
358                                 if (op == 'M') op = BAM_CMATCH;
359                                 else if (op == 'I') op = BAM_CINS;
360                                 else if (op == 'D') op = BAM_CDEL;
361                                 else if (op == 'N') op = BAM_CREF_SKIP;
362                                 else if (op == 'S') op = BAM_CSOFT_CLIP;
363                                 else if (op == 'H') op = BAM_CHARD_CLIP;
364                                 else if (op == 'P') op = BAM_CPAD;
365                                 else parse_error(fp->n_lines, "invalid CIGAR operation");
366                                 s = t + 1;
367                                 bam1_cigar(b)[i] = x << BAM_CIGAR_SHIFT | op;
368                         }
369                         if (*s) parse_error(fp->n_lines, "unmatched CIGAR operation");
370                         c->bin = bam_reg2bin(c->pos, bam_calend(c, bam1_cigar(b)));
371                         doff += c->n_cigar * 4;
372                 } else c->bin = bam_reg2bin(c->pos, c->pos + 1);
373         }
374         { // mtid, mpos, isize
375                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
376                 c->mtid = strcmp(str->s, "=")? bam_get_tid(header, str->s) : c->tid;
377                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
378                 c->mpos = isdigit(str->s[0])? atoi(str->s) - 1 : -1;
379                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
380                 c->isize = (str->s[0] == '-' || isdigit(str->s[0]))? atoi(str->s) : 0;
381                 if (ret < 0) return -4;
382         }
383         { // seq and qual
384                 int i;
385                 uint8_t *p;
386                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -5; // seq
387                 z += str->l + 1;
388                 c->l_qseq = strlen(str->s);
389                 if (c->n_cigar && c->l_qseq != (int32_t)bam_cigar2qlen(c, bam1_cigar(b)))
390                         parse_error(fp->n_lines, "CIGAR and sequence length are inconsistent");
391                 p = (uint8_t*)alloc_data(b, doff + c->l_qseq + (c->l_qseq+1)/2) + doff;
392                 memset(p, 0, (c->l_qseq+1)/2);
393                 for (i = 0; i < c->l_qseq; ++i)
394                         p[i/2] |= bam_nt16_table[(int)str->s[i]] << 4*(1-i%2);
395                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -6; // qual
396                 z += str->l + 1;
397                 if (strcmp(str->s, "*") && c->l_qseq != strlen(str->s))
398                         parse_error(fp->n_lines, "sequence and quality are inconsistent");
399                 p += (c->l_qseq+1)/2;
400                 if (strcmp(str->s, "*") == 0) for (i = 0; i < c->l_qseq; ++i) p[i] = 0xff;
401                 else for (i = 0; i < c->l_qseq; ++i) p[i] = str->s[i] - 33;
402                 doff += c->l_qseq + (c->l_qseq+1)/2;
403         }
404         doff0 = doff;
405         if (dret != '\n' && dret != '\r') { // aux
406                 while (ks_getuntil(ks, KS_SEP_TAB, str, &dret) >= 0) {
407                         uint8_t *s, type, key[2];
408                         z += str->l + 1;
409                         if (str->l < 6 || str->s[2] != ':' || str->s[4] != ':')
410                                 parse_error(fp->n_lines, "missing colon in auxiliary data");
411                         key[0] = str->s[0]; key[1] = str->s[1];
412                         type = str->s[3];
413                         s = alloc_data(b, doff + 3) + doff;
414                         s[0] = key[0]; s[1] = key[1]; s += 2; doff += 2;
415                         if (type == 'A' || type == 'a' || type == 'c' || type == 'C') { // c and C for backward compatibility
416                                 s = alloc_data(b, doff + 2) + doff;
417                                 *s++ = 'A'; *s = str->s[5];
418                                 doff += 2;
419                         } else if (type == 'I' || type == 'i') {
420                                 long long x;
421                                 s = alloc_data(b, doff + 5) + doff;
422                                 x = (long long)atoll(str->s + 5);
423                                 if (x < 0) {
424                                         if (x >= -127) {
425                                                 *s++ = 'c'; *(int8_t*)s = (int8_t)x;
426                                                 s += 1; doff += 2;
427                                         } else if (x >= -32767) {
428                                                 *s++ = 's'; *(int16_t*)s = (int16_t)x;
429                                                 s += 2; doff += 3;
430                                         } else {
431                                                 *s++ = 'i'; *(int32_t*)s = (int32_t)x;
432                                                 s += 4; doff += 5;
433                                                 if (x < -2147483648ll)
434                                                         fprintf(stderr, "Parse warning at line %lld: integer %lld is out of range.",
435                                                                         (long long)fp->n_lines, x);
436                                         }
437                                 } else {
438                                         if (x <= 255) {
439                                                 *s++ = 'C'; *s++ = (uint8_t)x;
440                                                 doff += 2;
441                                         } else if (x <= 65535) {
442                                                 *s++ = 'S'; *(uint16_t*)s = (uint16_t)x;
443                                                 s += 2; doff += 3;
444                                         } else {
445                                                 *s++ = 'I'; *(uint32_t*)s = (uint32_t)x;
446                                                 s += 4; doff += 5;
447                                                 if (x > 4294967295ll)
448                                                         fprintf(stderr, "Parse warning at line %lld: integer %lld is out of range.",
449                                                                         (long long)fp->n_lines, x);
450                                         }
451                                 }
452                         } else if (type == 'f') {
453                                 s = alloc_data(b, doff + 5) + doff;
454                                 *s++ = 'f';
455                                 *(float*)s = (float)atof(str->s + 5);
456                                 s += 4; doff += 5;
457                         } else if (type == 'd') {
458                                 s = alloc_data(b, doff + 9) + doff;
459                                 *s++ = 'd';
460                                 *(float*)s = (float)atof(str->s + 9);
461                                 s += 8; doff += 9;
462                         } else if (type == 'Z' || type == 'H') {
463                                 int size = 1 + (str->l - 5) + 1;
464                                 if (type == 'H') { // check whether the hex string is valid
465                                         int i;
466                                         if ((str->l - 5) % 2 == 1) parse_error(fp->n_lines, "length of the hex string not even");
467                                         for (i = 0; i < str->l - 5; ++i) {
468                                                 int c = toupper(str->s[5 + i]);
469                                                 if (!((c >= '0' && c <= '9') || (c >= 'A' && c <= 'F')))
470                                                         parse_error(fp->n_lines, "invalid hex character");
471                                         }
472                                 }
473                                 s = alloc_data(b, doff + size) + doff;
474                                 *s++ = type;
475                                 memcpy(s, str->s + 5, str->l - 5);
476                                 s[str->l - 5] = 0;
477                                 doff += size;
478                         } else parse_error(fp->n_lines, "unrecognized type");
479                         if (dret == '\n' || dret == '\r') break;
480                 }
481         }
482         b->l_aux = doff - doff0;
483         b->data_len = doff;
484         return z;
485 }
486
487 tamFile sam_open(const char *fn)
488 {
489         tamFile fp;
490         fp = (tamFile)calloc(1, sizeof(struct __tamFile_t));
491         fp->str = (kstring_t*)calloc(1, sizeof(kstring_t));
492         fp->fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
493         fp->ks = ks_init(fp->fp);
494         return fp;
495 }
496
497 void sam_close(tamFile fp)
498 {
499         if (fp) {
500                 ks_destroy(fp->ks);
501                 gzclose(fp->fp);
502                 free(fp->str->s); free(fp->str);
503                 free(fp);
504         }
505 }