]> git.donarmstrong.com Git - samtools.git/blob - bam_import.c
* samtools-0.1.5-16 (r421)
[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         if (fn == 0) return 0;
122         fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
123         if (fp == 0) return 0;
124         hash = kh_init(ref);
125         ks = ks_init(fp);
126         str = (kstring_t*)calloc(1, sizeof(kstring_t));
127         while (ks_getuntil(ks, 0, str, &dret) > 0) {
128                 char *s = strdup(str->s);
129                 int len, i;
130                 i = kh_size(hash);
131                 ks_getuntil(ks, 0, str, &dret);
132                 len = atoi(str->s);
133                 k = kh_put(ref, hash, s, &ret);
134                 kh_value(hash, k) = (uint64_t)len<<32 | i;
135                 if (dret != '\n')
136                         while ((c = ks_getc(ks)) != '\n' && c != -1);
137         }
138         ks_destroy(ks);
139         gzclose(fp);
140         free(str->s); free(str);
141         fprintf(stderr, "[sam_header_read2] %d sequences loaded.\n", kh_size(hash));
142         header = hash2header(hash);
143         kh_destroy(ref, hash);
144         return header;
145 }
146 static inline uint8_t *alloc_data(bam1_t *b, int size)
147 {
148         if (b->m_data < size) {
149                 b->m_data = size;
150                 kroundup32(b->m_data);
151                 b->data = (uint8_t*)realloc(b->data, b->m_data);
152         }
153         return b->data;
154 }
155 static inline void parse_error(int64_t n_lines, const char * __restrict msg)
156 {
157         fprintf(stderr, "Parse error at line %lld: %s\n", (long long)n_lines, msg);
158         abort();
159 }
160 static inline void append_text(bam_header_t *header, kstring_t *str)
161 {
162         int x = header->l_text, y = header->l_text + str->l + 2; // 2 = 1 byte dret + 1 byte null
163         kroundup32(x); kroundup32(y);
164         if (x < y) header->text = (char*)realloc(header->text, y);
165         strncpy(header->text + header->l_text, str->s, str->l+1); // we cannot use strcpy() here.
166         header->l_text += str->l + 1;
167         header->text[header->l_text] = 0;
168 }
169
170 int sam_header_parse_rg(bam_header_t *h)
171 {
172         kstring_t *rgid, *rglib;
173         char *p, *q, *s, *r;
174         int n = 0;
175
176         // free
177         if (h == 0) return 0;
178         bam_strmap_destroy(h->rg2lib); h->rg2lib = 0;
179         if (h->l_text < 3) return 0;
180         // parse @RG lines
181         h->rg2lib = bam_strmap_init();
182         rgid = calloc(1, sizeof(kstring_t));
183         rglib = calloc(1, sizeof(kstring_t));
184         s = h->text;
185         while ((s = strstr(s, "@RG")) != 0) {
186                 if (rgid->l && rglib->l) {
187                         bam_strmap_put(h->rg2lib, rgid->s, rglib->s);
188                         ++n;
189                 }
190                 rgid->l = rglib->l = 0;
191                 s += 3;
192                 r = s;
193                 if ((p = strstr(s, "ID:")) != 0) {
194                         q = p + 3;
195                         for (p = q; *p && *p != '\t' && *p != '\r' && *p != '\n'; ++p);
196                         kputsn(q, p - q, rgid);
197                 } else {
198                         fprintf(stderr, "[bam_header_parse] missing ID tag in @RG lines.\n");
199                         break;
200                 }
201                 if (r < p) r = p;
202                 if ((p = strstr(s, "LB:")) != 0) {
203                         q = p + 3;
204                         for (p = q; *p && *p != '\t' && *p != '\r' && *p != '\n'; ++p);
205                         kputsn(q, p - q, rglib);
206                 } else {
207                         fprintf(stderr, "[bam_header_parse] missing LB tag in @RG lines.\n");
208                         break;
209                 }
210                 if (r < p) r = p;
211                 s = r + 3;
212         }
213         if (rgid->l && rglib->l) {
214                 bam_strmap_put(h->rg2lib, rgid->s, rglib->s);
215                 ++n;
216         }
217         free(rgid->s); free(rgid);
218         free(rglib->s); free(rglib);
219         if (n == 0) {
220                 bam_strmap_destroy(h->rg2lib);
221                 h->rg2lib = 0;
222         }
223         return n;
224 }
225
226 int sam_header_parse(bam_header_t *h)
227 {
228         int i;
229         char *s, *p, *q, *r;
230
231         // free
232         free(h->target_len); free(h->target_name);
233         h->n_targets = 0; h->target_len = 0; h->target_name = 0;
234         if (h->l_text < 3) return 0;
235         // count number of @SQ
236         s = h->text;
237         while ((s = strstr(s, "@SQ")) != 0) {
238                 ++h->n_targets;
239                 s += 3;
240         }
241         if (h->n_targets == 0) return 0;
242         h->target_len = (uint32_t*)calloc(h->n_targets, 4);
243         h->target_name = (char**)calloc(h->n_targets, sizeof(void*));
244         // parse @SQ lines
245         i = 0;
246         s = h->text;
247         while ((s = strstr(s, "@SQ")) != 0) {
248                 s += 3;
249                 r = s;
250                 if ((p = strstr(s, "SN:")) != 0) {
251                         q = p + 3;
252                         for (p = q; *p && *p != '\t' && *p != '\r' && *p != '\n'; ++p);
253                         h->target_name[i] = (char*)calloc(p - q + 1, 1);
254                         strncpy(h->target_name[i], q, p - q);
255                 } else goto header_err_ret;
256                 if (r < p) r = p;
257                 if ((p = strstr(s, "LN:")) != 0) h->target_len[i] = strtol(p + 3, 0, 10);
258                 else goto header_err_ret;
259                 if (r < p) r = p;
260                 s = r + 3;
261                 ++i;
262         }
263         sam_header_parse_rg(h);
264         return h->n_targets;
265
266 header_err_ret:
267         fprintf(stderr, "[bam_header_parse] missing SN or LN tag in @SQ lines.\n");
268         free(h->target_len); free(h->target_name);
269         h->n_targets = 0; h->target_len = 0; h->target_name = 0;
270         return 0;
271 }
272
273 bam_header_t *sam_header_read(tamFile fp)
274 {
275         int ret, dret;
276         bam_header_t *header = bam_header_init();
277         kstring_t *str = fp->str;
278         while ((ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret)) >= 0 && str->s[0] == '@') { // skip header
279                 str->s[str->l] = dret; // note that str->s is NOT null terminated!!
280                 append_text(header, str);
281                 if (dret != '\n') {
282                         ret = ks_getuntil(fp->ks, '\n', str, &dret);
283                         str->s[str->l] = '\n'; // NOT null terminated!!
284                         append_text(header, str);
285                 }
286                 ++fp->n_lines;
287         }
288         sam_header_parse(header);
289         bam_init_header_hash(header);
290         fp->is_first = 1;
291         return header;
292 }
293
294 int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b)
295 {
296         int ret, doff, doff0, dret, z = 0;
297         bam1_core_t *c = &b->core;
298         kstring_t *str = fp->str;
299         kstream_t *ks = fp->ks;
300
301         if (fp->is_first) {
302                 fp->is_first = 0;
303                 ret = str->l;
304         } else {
305                 do { // special consideration for empty lines
306                         ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret);
307                         if (ret >= 0) z += str->l + 1;
308                 } while (ret == 0);
309         }
310         if (ret < 0) return -1;
311         ++fp->n_lines;
312         doff = 0;
313
314         { // name
315                 c->l_qname = strlen(str->s) + 1;
316                 memcpy(alloc_data(b, doff + c->l_qname) + doff, str->s, c->l_qname);
317                 doff += c->l_qname;
318         }
319         { // flag
320                 long flag;
321                 char *s;
322                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
323                 flag = strtol((char*)str->s, &s, 0);
324                 if (*s) { // not the end of the string
325                         flag = 0;
326                         for (s = str->s; *s; ++s)
327                                 flag |= bam_char2flag_table[(int)*s];
328                 }
329                 c->flag = flag;
330         }
331         { // tid, pos, qual
332                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->tid = bam_get_tid(header, str->s);
333                 if (c->tid < 0 && strcmp(str->s, "*")) {
334                         if (header->n_targets == 0) {
335                                 fprintf(stderr, "[sam_read1] missing header? Abort!\n");
336                                 exit(1);
337                         } else fprintf(stderr, "[sam_read1] reference '%s' is recognized as '*'.\n", str->s);
338                 }
339                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->pos = isdigit(str->s[0])? atoi(str->s) - 1 : -1;
340                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->qual = isdigit(str->s[0])? atoi(str->s) : 0;
341                 if (ret < 0) return -2;
342         }
343         { // cigar
344                 char *s, *t;
345                 int i, op;
346                 long x;
347                 c->n_cigar = 0;
348                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -3;
349                 z += str->l + 1;
350                 if (str->s[0] != '*') {
351                         for (s = str->s; *s; ++s) {
352                                 if (isalpha(*s)) ++c->n_cigar;
353                                 else if (!isdigit(*s)) parse_error(fp->n_lines, "invalid CIGAR character");
354                         }
355                         b->data = alloc_data(b, doff + c->n_cigar * 4);
356                         for (i = 0, s = str->s; i != c->n_cigar; ++i) {
357                                 x = strtol(s, &t, 10);
358                                 op = toupper(*t);
359                                 if (op == 'M') op = BAM_CMATCH;
360                                 else if (op == 'I') op = BAM_CINS;
361                                 else if (op == 'D') op = BAM_CDEL;
362                                 else if (op == 'N') op = BAM_CREF_SKIP;
363                                 else if (op == 'S') op = BAM_CSOFT_CLIP;
364                                 else if (op == 'H') op = BAM_CHARD_CLIP;
365                                 else if (op == 'P') op = BAM_CPAD;
366                                 else parse_error(fp->n_lines, "invalid CIGAR operation");
367                                 s = t + 1;
368                                 bam1_cigar(b)[i] = x << BAM_CIGAR_SHIFT | op;
369                         }
370                         if (*s) parse_error(fp->n_lines, "unmatched CIGAR operation");
371                         c->bin = bam_reg2bin(c->pos, bam_calend(c, bam1_cigar(b)));
372                         doff += c->n_cigar * 4;
373                 } else c->bin = bam_reg2bin(c->pos, c->pos + 1);
374         }
375         { // mtid, mpos, isize
376                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
377                 c->mtid = strcmp(str->s, "=")? bam_get_tid(header, str->s) : c->tid;
378                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
379                 c->mpos = isdigit(str->s[0])? atoi(str->s) - 1 : -1;
380                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
381                 c->isize = (str->s[0] == '-' || isdigit(str->s[0]))? atoi(str->s) : 0;
382                 if (ret < 0) return -4;
383         }
384         { // seq and qual
385                 int i;
386                 uint8_t *p;
387                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -5; // seq
388                 z += str->l + 1;
389                 c->l_qseq = strlen(str->s);
390                 if (c->n_cigar && c->l_qseq != (int32_t)bam_cigar2qlen(c, bam1_cigar(b)))
391                         parse_error(fp->n_lines, "CIGAR and sequence length are inconsistent");
392                 p = (uint8_t*)alloc_data(b, doff + c->l_qseq + (c->l_qseq+1)/2) + doff;
393                 memset(p, 0, (c->l_qseq+1)/2);
394                 for (i = 0; i < c->l_qseq; ++i)
395                         p[i/2] |= bam_nt16_table[(int)str->s[i]] << 4*(1-i%2);
396                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -6; // qual
397                 z += str->l + 1;
398                 if (strcmp(str->s, "*") && c->l_qseq != strlen(str->s))
399                         parse_error(fp->n_lines, "sequence and quality are inconsistent");
400                 p += (c->l_qseq+1)/2;
401                 if (strcmp(str->s, "*") == 0) for (i = 0; i < c->l_qseq; ++i) p[i] = 0xff;
402                 else for (i = 0; i < c->l_qseq; ++i) p[i] = str->s[i] - 33;
403                 doff += c->l_qseq + (c->l_qseq+1)/2;
404         }
405         doff0 = doff;
406         if (dret != '\n' && dret != '\r') { // aux
407                 while (ks_getuntil(ks, KS_SEP_TAB, str, &dret) >= 0) {
408                         uint8_t *s, type, key[2];
409                         z += str->l + 1;
410                         if (str->l < 6 || str->s[2] != ':' || str->s[4] != ':')
411                                 parse_error(fp->n_lines, "missing colon in auxiliary data");
412                         key[0] = str->s[0]; key[1] = str->s[1];
413                         type = str->s[3];
414                         s = alloc_data(b, doff + 3) + doff;
415                         s[0] = key[0]; s[1] = key[1]; s += 2; doff += 2;
416                         if (type == 'A' || type == 'a' || type == 'c' || type == 'C') { // c and C for backward compatibility
417                                 s = alloc_data(b, doff + 2) + doff;
418                                 *s++ = 'A'; *s = str->s[5];
419                                 doff += 2;
420                         } else if (type == 'I' || type == 'i') {
421                                 long long x;
422                                 s = alloc_data(b, doff + 5) + doff;
423                                 x = (long long)atoll(str->s + 5);
424                                 if (x < 0) {
425                                         if (x >= -127) {
426                                                 *s++ = 'c'; *(int8_t*)s = (int8_t)x;
427                                                 s += 1; doff += 2;
428                                         } else if (x >= -32767) {
429                                                 *s++ = 's'; *(int16_t*)s = (int16_t)x;
430                                                 s += 2; doff += 3;
431                                         } else {
432                                                 *s++ = 'i'; *(int32_t*)s = (int32_t)x;
433                                                 s += 4; doff += 5;
434                                                 if (x < -2147483648ll)
435                                                         fprintf(stderr, "Parse warning at line %lld: integer %lld is out of range.",
436                                                                         (long long)fp->n_lines, x);
437                                         }
438                                 } else {
439                                         if (x <= 255) {
440                                                 *s++ = 'C'; *s++ = (uint8_t)x;
441                                                 doff += 2;
442                                         } else if (x <= 65535) {
443                                                 *s++ = 'S'; *(uint16_t*)s = (uint16_t)x;
444                                                 s += 2; doff += 3;
445                                         } else {
446                                                 *s++ = 'I'; *(uint32_t*)s = (uint32_t)x;
447                                                 s += 4; doff += 5;
448                                                 if (x > 4294967295ll)
449                                                         fprintf(stderr, "Parse warning at line %lld: integer %lld is out of range.",
450                                                                         (long long)fp->n_lines, x);
451                                         }
452                                 }
453                         } else if (type == 'f') {
454                                 s = alloc_data(b, doff + 5) + doff;
455                                 *s++ = 'f';
456                                 *(float*)s = (float)atof(str->s + 5);
457                                 s += 4; doff += 5;
458                         } else if (type == 'd') {
459                                 s = alloc_data(b, doff + 9) + doff;
460                                 *s++ = 'd';
461                                 *(float*)s = (float)atof(str->s + 9);
462                                 s += 8; doff += 9;
463                         } else if (type == 'Z' || type == 'H') {
464                                 int size = 1 + (str->l - 5) + 1;
465                                 if (type == 'H') { // check whether the hex string is valid
466                                         int i;
467                                         if ((str->l - 5) % 2 == 1) parse_error(fp->n_lines, "length of the hex string not even");
468                                         for (i = 0; i < str->l - 5; ++i) {
469                                                 int c = toupper(str->s[5 + i]);
470                                                 if (!((c >= '0' && c <= '9') || (c >= 'A' && c <= 'F')))
471                                                         parse_error(fp->n_lines, "invalid hex character");
472                                         }
473                                 }
474                                 s = alloc_data(b, doff + size) + doff;
475                                 *s++ = type;
476                                 memcpy(s, str->s + 5, str->l - 5);
477                                 s[str->l - 5] = 0;
478                                 doff += size;
479                         } else parse_error(fp->n_lines, "unrecognized type");
480                         if (dret == '\n' || dret == '\r') break;
481                 }
482         }
483         b->l_aux = doff - doff0;
484         b->data_len = doff;
485         return z;
486 }
487
488 tamFile sam_open(const char *fn)
489 {
490         tamFile fp;
491         gzFile gzfp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
492         if (gzfp == 0) return 0;
493         fp = (tamFile)calloc(1, sizeof(struct __tamFile_t));
494         fp->str = (kstring_t*)calloc(1, sizeof(kstring_t));
495         fp->fp = gzfp;
496         fp->ks = ks_init(fp->fp);
497         return fp;
498 }
499
500 void sam_close(tamFile fp)
501 {
502         if (fp) {
503                 ks_destroy(fp->ks);
504                 gzclose(fp->fp);
505                 free(fp->str->s); free(fp->str);
506                 free(fp);
507         }
508 }