Merge branch 'upstream'
[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 #ifdef _WIN32
9 #include <fcntl.h>
10 #endif
11 #include "kstring.h"
12 #include "bam.h"
13 #include "kseq.h"
14 #include "khash.h"
15
16 KSTREAM_INIT(gzFile, gzread, 8192)
17 KHASH_MAP_INIT_STR(ref, uint64_t)
18
19 void bam_init_header_hash(bam_header_t *header);
20 void bam_destroy_header_hash(bam_header_t *header);
21 int32_t bam_get_tid(const bam_header_t *header, const char *seq_name);
22
23 unsigned char bam_nt16_table[256] = {
24         15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
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          1, 2, 4, 8, 15,15,15,15, 15,15,15,15, 15, 0 /*=*/,15,15,
28         15, 1,14, 2, 13,15,15, 4, 11,15,15,12, 15, 3,15,15,
29         15,15, 5, 6,  8,15, 7, 9, 15,10,15,15, 15,15,15,15,
30         15, 1,14, 2, 13,15,15, 4, 11,15,15,12, 15, 3,15,15,
31         15,15, 5, 6,  8,15, 7, 9, 15,10,15,15, 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         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 };
41
42 unsigned short bam_char2flag_table[256] = {
43         0,0,0,0, 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         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
46         0,BAM_FREAD1,BAM_FREAD2,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
47         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
48         BAM_FPROPER_PAIR,0,BAM_FMREVERSE,0, 0,BAM_FMUNMAP,0,0, 0,0,0,0, 0,0,0,0,
49         0,0,0,0, BAM_FDUP,0,BAM_FQCFAIL,0, 0,0,0,0, 0,0,0,0,
50         BAM_FPAIRED,0,BAM_FREVERSE,BAM_FSECONDARY, 0,BAM_FUNMAP,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         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 };
60
61 char *bam_nt16_rev_table = "=ACMGRSVTWYHKDBN";
62
63 struct __tamFile_t {
64         gzFile fp;
65         kstream_t *ks;
66         kstring_t *str;
67         uint64_t n_lines;
68         int is_first;
69 };
70
71 char **__bam_get_lines(const char *fn, int *_n) // for bam_plcmd.c only
72 {
73         char **list = 0, *s;
74         int n = 0, dret, m = 0;
75         gzFile fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
76         kstream_t *ks;
77         kstring_t *str;
78         str = (kstring_t*)calloc(1, sizeof(kstring_t));
79         ks = ks_init(fp);
80         while (ks_getuntil(ks, '\n', str, &dret) > 0) {
81                 if (n == m) {
82                         m = m? m << 1 : 16;
83                         list = (char**)realloc(list, m * sizeof(char*));
84                 }
85                 if (str->s[str->l-1] == '\r')
86                         str->s[--str->l] = '\0';
87                 s = list[n++] = (char*)calloc(str->l + 1, 1);
88                 strcpy(s, str->s);
89         }
90         ks_destroy(ks);
91         gzclose(fp);
92         free(str->s); free(str);
93         *_n = n;
94         return list;
95 }
96
97 static bam_header_t *hash2header(const kh_ref_t *hash)
98 {
99         bam_header_t *header;
100         khiter_t k;
101         header = bam_header_init();
102         header->n_targets = kh_size(hash);
103         header->target_name = (char**)calloc(kh_size(hash), sizeof(char*));
104         header->target_len = (uint32_t*)calloc(kh_size(hash), 4);
105         for (k = kh_begin(hash); k != kh_end(hash); ++k) {
106                 if (kh_exist(hash, k)) {
107                         int i = (int)kh_value(hash, k);
108                         header->target_name[i] = (char*)kh_key(hash, k);
109                         header->target_len[i] = kh_value(hash, k)>>32;
110                 }
111         }
112         bam_init_header_hash(header);
113         return header;
114 }
115 bam_header_t *sam_header_read2(const char *fn)
116 {
117         bam_header_t *header;
118         int c, dret, ret;
119         gzFile fp;
120         kstream_t *ks;
121         kstring_t *str;
122         kh_ref_t *hash;
123         khiter_t k;
124         if (fn == 0) return 0;
125         fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
126         if (fp == 0) return 0;
127         hash = kh_init(ref);
128         ks = ks_init(fp);
129         str = (kstring_t*)calloc(1, sizeof(kstring_t));
130         while (ks_getuntil(ks, 0, str, &dret) > 0) {
131                 char *s = strdup(str->s);
132                 int len, i;
133                 i = kh_size(hash);
134                 ks_getuntil(ks, 0, str, &dret);
135                 len = atoi(str->s);
136                 k = kh_put(ref, hash, s, &ret);
137                 kh_value(hash, k) = (uint64_t)len<<32 | i;
138                 if (dret != '\n')
139                         while ((c = ks_getc(ks)) != '\n' && c != -1);
140         }
141         ks_destroy(ks);
142         gzclose(fp);
143         free(str->s); free(str);
144         fprintf(stderr, "[sam_header_read2] %d sequences loaded.\n", kh_size(hash));
145         header = hash2header(hash);
146         kh_destroy(ref, hash);
147         return header;
148 }
149 static inline uint8_t *alloc_data(bam1_t *b, int size)
150 {
151         if (b->m_data < size) {
152                 b->m_data = size;
153                 kroundup32(b->m_data);
154                 b->data = (uint8_t*)realloc(b->data, b->m_data);
155         }
156         return b->data;
157 }
158 static inline void parse_error(int64_t n_lines, const char * __restrict msg)
159 {
160         fprintf(stderr, "Parse error at line %lld: %s\n", (long long)n_lines, msg);
161         abort();
162 }
163 static inline void append_text(bam_header_t *header, kstring_t *str)
164 {
165         int x = header->l_text, y = header->l_text + str->l + 2; // 2 = 1 byte dret + 1 byte null
166         kroundup32(x); kroundup32(y);
167         if (x < y) header->text = (char*)realloc(header->text, y);
168         strncpy(header->text + header->l_text, str->s, str->l+1); // we cannot use strcpy() here.
169         header->l_text += str->l + 1;
170         header->text[header->l_text] = 0;
171 }
172
173 int sam_header_parse_rg(bam_header_t *h)
174 {
175         kstring_t *rgid, *rglib;
176         char *p, *q, *s, *r;
177         int n = 0;
178
179         // free
180         if (h == 0) return 0;
181         bam_strmap_destroy(h->rg2lib); h->rg2lib = 0;
182         if (h->l_text < 3) return 0;
183         // parse @RG lines
184         h->rg2lib = bam_strmap_init();
185         rgid = calloc(1, sizeof(kstring_t));
186         rglib = calloc(1, sizeof(kstring_t));
187         s = h->text;
188         while ((s = strstr(s, "@RG")) != 0) {
189                 if (rgid->l && rglib->l) {
190                         bam_strmap_put(h->rg2lib, rgid->s, rglib->s);
191                         ++n;
192                 }
193                 rgid->l = rglib->l = 0;
194                 s += 3;
195                 r = s;
196                 if ((p = strstr(s, "ID:")) != 0) {
197                         q = p + 3;
198                         for (p = q; *p && *p != '\t' && *p != '\r' && *p != '\n'; ++p);
199                         kputsn(q, p - q, rgid);
200                 } else {
201                         fprintf(stderr, "[bam_header_parse] missing ID tag in @RG lines.\n");
202                         break;
203                 }
204                 if (r < p) r = p;
205                 if ((p = strstr(s, "LB:")) != 0) {
206                         q = p + 3;
207                         for (p = q; *p && *p != '\t' && *p != '\r' && *p != '\n'; ++p);
208                         kputsn(q, p - q, rglib);
209                 } else {
210                         fprintf(stderr, "[bam_header_parse] missing LB tag in @RG lines.\n");
211                         break;
212                 }
213                 if (r < p) r = p;
214                 s = r + 3;
215         }
216         if (rgid->l && rglib->l) {
217                 bam_strmap_put(h->rg2lib, rgid->s, rglib->s);
218                 ++n;
219         }
220         free(rgid->s); free(rgid);
221         free(rglib->s); free(rglib);
222         if (n == 0) {
223                 bam_strmap_destroy(h->rg2lib);
224                 h->rg2lib = 0;
225         }
226         return n;
227 }
228
229 int sam_header_parse(bam_header_t *h)
230 {
231         int i;
232         char *s, *p, *q, *r;
233
234         // free
235         free(h->target_len); free(h->target_name);
236         h->n_targets = 0; h->target_len = 0; h->target_name = 0;
237         if (h->l_text < 3) return 0;
238         // count number of @SQ
239         s = h->text;
240         while ((s = strstr(s, "@SQ")) != 0) {
241                 ++h->n_targets;
242                 s += 3;
243         }
244         if (h->n_targets == 0) return 0;
245         h->target_len = (uint32_t*)calloc(h->n_targets, 4);
246         h->target_name = (char**)calloc(h->n_targets, sizeof(void*));
247         // parse @SQ lines
248         i = 0;
249         s = h->text;
250         while ((s = strstr(s, "@SQ")) != 0) {
251                 s += 3;
252                 r = s;
253                 if ((p = strstr(s, "SN:")) != 0) {
254                         q = p + 3;
255                         for (p = q; *p && *p != '\t' && *p != '\r' && *p != '\n'; ++p);
256                         h->target_name[i] = (char*)calloc(p - q + 1, 1);
257                         strncpy(h->target_name[i], q, p - q);
258                 } else goto header_err_ret;
259                 if (r < p) r = p;
260                 if ((p = strstr(s, "LN:")) != 0) h->target_len[i] = strtol(p + 3, 0, 10);
261                 else goto header_err_ret;
262                 if (r < p) r = p;
263                 s = r + 3;
264                 ++i;
265         }
266         sam_header_parse_rg(h);
267         return h->n_targets;
268
269 header_err_ret:
270         fprintf(stderr, "[bam_header_parse] missing SN or LN tag in @SQ lines.\n");
271         free(h->target_len); free(h->target_name);
272         h->n_targets = 0; h->target_len = 0; h->target_name = 0;
273         return 0;
274 }
275
276 bam_header_t *sam_header_read(tamFile fp)
277 {
278         int ret, dret;
279         bam_header_t *header = bam_header_init();
280         kstring_t *str = fp->str;
281         while ((ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret)) >= 0 && str->s[0] == '@') { // skip header
282                 str->s[str->l] = dret; // note that str->s is NOT null terminated!!
283                 append_text(header, str);
284                 if (dret != '\n') {
285                         ret = ks_getuntil(fp->ks, '\n', str, &dret);
286                         str->s[str->l] = '\n'; // NOT null terminated!!
287                         append_text(header, str);
288                 }
289                 ++fp->n_lines;
290         }
291         sam_header_parse(header);
292         bam_init_header_hash(header);
293         fp->is_first = 1;
294         return header;
295 }
296
297 int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b)
298 {
299         int ret, doff, doff0, dret, z = 0;
300         bam1_core_t *c = &b->core;
301         kstring_t *str = fp->str;
302         kstream_t *ks = fp->ks;
303
304         if (fp->is_first) {
305                 fp->is_first = 0;
306                 ret = str->l;
307         } else {
308                 do { // special consideration for empty lines
309                         ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret);
310                         if (ret >= 0) z += str->l + 1;
311                 } while (ret == 0);
312         }
313         if (ret < 0) return -1;
314         ++fp->n_lines;
315         doff = 0;
316
317         { // name
318                 c->l_qname = strlen(str->s) + 1;
319                 memcpy(alloc_data(b, doff + c->l_qname) + doff, str->s, c->l_qname);
320                 doff += c->l_qname;
321         }
322         { // flag
323                 long flag;
324                 char *s;
325                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
326                 flag = strtol((char*)str->s, &s, 0);
327                 if (*s) { // not the end of the string
328                         flag = 0;
329                         for (s = str->s; *s; ++s)
330                                 flag |= bam_char2flag_table[(int)*s];
331                 }
332                 c->flag = flag;
333         }
334         { // tid, pos, qual
335                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->tid = bam_get_tid(header, str->s);
336                 if (c->tid < 0 && strcmp(str->s, "*")) {
337                         if (header->n_targets == 0) {
338                                 fprintf(stderr, "[sam_read1] missing header? Abort!\n");
339                                 exit(1);
340                         } else fprintf(stderr, "[sam_read1] reference '%s' is recognized as '*'.\n", str->s);
341                 }
342                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->pos = isdigit(str->s[0])? atoi(str->s) - 1 : -1;
343                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->qual = isdigit(str->s[0])? atoi(str->s) : 0;
344                 if (ret < 0) return -2;
345         }
346         { // cigar
347                 char *s, *t;
348                 int i, op;
349                 long x;
350                 c->n_cigar = 0;
351                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -3;
352                 z += str->l + 1;
353                 if (str->s[0] != '*') {
354                         for (s = str->s; *s; ++s) {
355                                 if (isalpha(*s)) ++c->n_cigar;
356                                 else if (!isdigit(*s)) parse_error(fp->n_lines, "invalid CIGAR character");
357                         }
358                         b->data = alloc_data(b, doff + c->n_cigar * 4);
359                         for (i = 0, s = str->s; i != c->n_cigar; ++i) {
360                                 x = strtol(s, &t, 10);
361                                 op = toupper(*t);
362                                 if (op == 'M' || op == '=' || op == 'X') op = BAM_CMATCH;
363                                 else if (op == 'I') op = BAM_CINS;
364                                 else if (op == 'D') op = BAM_CDEL;
365                                 else if (op == 'N') op = BAM_CREF_SKIP;
366                                 else if (op == 'S') op = BAM_CSOFT_CLIP;
367                                 else if (op == 'H') op = BAM_CHARD_CLIP;
368                                 else if (op == 'P') op = BAM_CPAD;
369                                 else parse_error(fp->n_lines, "invalid CIGAR operation");
370                                 s = t + 1;
371                                 bam1_cigar(b)[i] = x << BAM_CIGAR_SHIFT | op;
372                         }
373                         if (*s) parse_error(fp->n_lines, "unmatched CIGAR operation");
374                         c->bin = bam_reg2bin(c->pos, bam_calend(c, bam1_cigar(b)));
375                         doff += c->n_cigar * 4;
376                 } else {
377                         if (!(c->flag&BAM_FUNMAP)) {
378                                 fprintf(stderr, "Parse warning at line %lld: mapped sequence without CIGAR\n", (long long)fp->n_lines);
379                                 c->flag |= BAM_FUNMAP;
380                         }
381                         c->bin = bam_reg2bin(c->pos, c->pos + 1);
382                 }
383         }
384         { // mtid, mpos, isize
385                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
386                 c->mtid = strcmp(str->s, "=")? bam_get_tid(header, str->s) : c->tid;
387                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
388                 c->mpos = isdigit(str->s[0])? atoi(str->s) - 1 : -1;
389                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
390                 c->isize = (str->s[0] == '-' || isdigit(str->s[0]))? atoi(str->s) : 0;
391                 if (ret < 0) return -4;
392         }
393         { // seq and qual
394                 int i;
395                 uint8_t *p = 0;
396                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -5; // seq
397                 z += str->l + 1;
398                 if (strcmp(str->s, "*")) {
399                         c->l_qseq = strlen(str->s);
400                         if (c->n_cigar && c->l_qseq != (int32_t)bam_cigar2qlen(c, bam1_cigar(b)))
401                                 parse_error(fp->n_lines, "CIGAR and sequence length are inconsistent");
402                         p = (uint8_t*)alloc_data(b, doff + c->l_qseq + (c->l_qseq+1)/2) + doff;
403                         memset(p, 0, (c->l_qseq+1)/2);
404                         for (i = 0; i < c->l_qseq; ++i)
405                                 p[i/2] |= bam_nt16_table[(int)str->s[i]] << 4*(1-i%2);
406                 } else c->l_qseq = 0;
407                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -6; // qual
408                 z += str->l + 1;
409                 if (strcmp(str->s, "*") && c->l_qseq != strlen(str->s))
410                         parse_error(fp->n_lines, "sequence and quality are inconsistent");
411                 p += (c->l_qseq+1)/2;
412                 if (strcmp(str->s, "*") == 0) for (i = 0; i < c->l_qseq; ++i) p[i] = 0xff;
413                 else for (i = 0; i < c->l_qseq; ++i) p[i] = str->s[i] - 33;
414                 doff += c->l_qseq + (c->l_qseq+1)/2;
415         }
416         doff0 = doff;
417         if (dret != '\n' && dret != '\r') { // aux
418                 while (ks_getuntil(ks, KS_SEP_TAB, str, &dret) >= 0) {
419                         uint8_t *s, type, key[2];
420                         z += str->l + 1;
421                         if (str->l < 6 || str->s[2] != ':' || str->s[4] != ':')
422                                 parse_error(fp->n_lines, "missing colon in auxiliary data");
423                         key[0] = str->s[0]; key[1] = str->s[1];
424                         type = str->s[3];
425                         s = alloc_data(b, doff + 3) + doff;
426                         s[0] = key[0]; s[1] = key[1]; s += 2; doff += 2;
427                         if (type == 'A' || type == 'a' || type == 'c' || type == 'C') { // c and C for backward compatibility
428                                 s = alloc_data(b, doff + 2) + doff;
429                                 *s++ = 'A'; *s = str->s[5];
430                                 doff += 2;
431                         } else if (type == 'I' || type == 'i') {
432                                 long long x;
433                                 s = alloc_data(b, doff + 5) + doff;
434                                 x = (long long)atoll(str->s + 5);
435                                 if (x < 0) {
436                                         if (x >= -127) {
437                                                 *s++ = 'c'; *(int8_t*)s = (int8_t)x;
438                                                 s += 1; doff += 2;
439                                         } else if (x >= -32767) {
440                                                 *s++ = 's'; *(int16_t*)s = (int16_t)x;
441                                                 s += 2; doff += 3;
442                                         } else {
443                                                 *s++ = 'i'; *(int32_t*)s = (int32_t)x;
444                                                 s += 4; doff += 5;
445                                                 if (x < -2147483648ll)
446                                                         fprintf(stderr, "Parse warning at line %lld: integer %lld is out of range.",
447                                                                         (long long)fp->n_lines, x);
448                                         }
449                                 } else {
450                                         if (x <= 255) {
451                                                 *s++ = 'C'; *s++ = (uint8_t)x;
452                                                 doff += 2;
453                                         } else if (x <= 65535) {
454                                                 *s++ = 'S'; *(uint16_t*)s = (uint16_t)x;
455                                                 s += 2; doff += 3;
456                                         } else {
457                                                 *s++ = 'I'; *(uint32_t*)s = (uint32_t)x;
458                                                 s += 4; doff += 5;
459                                                 if (x > 4294967295ll)
460                                                         fprintf(stderr, "Parse warning at line %lld: integer %lld is out of range.",
461                                                                         (long long)fp->n_lines, x);
462                                         }
463                                 }
464                         } else if (type == 'f') {
465                                 s = alloc_data(b, doff + 5) + doff;
466                                 *s++ = 'f';
467                                 *(float*)s = (float)atof(str->s + 5);
468                                 s += 4; doff += 5;
469                         } else if (type == 'd') {
470                                 s = alloc_data(b, doff + 9) + doff;
471                                 *s++ = 'd';
472                                 *(float*)s = (float)atof(str->s + 9);
473                                 s += 8; doff += 9;
474                         } else if (type == 'Z' || type == 'H') {
475                                 int size = 1 + (str->l - 5) + 1;
476                                 if (type == 'H') { // check whether the hex string is valid
477                                         int i;
478                                         if ((str->l - 5) % 2 == 1) parse_error(fp->n_lines, "length of the hex string not even");
479                                         for (i = 0; i < str->l - 5; ++i) {
480                                                 int c = toupper(str->s[5 + i]);
481                                                 if (!((c >= '0' && c <= '9') || (c >= 'A' && c <= 'F')))
482                                                         parse_error(fp->n_lines, "invalid hex character");
483                                         }
484                                 }
485                                 s = alloc_data(b, doff + size) + doff;
486                                 *s++ = type;
487                                 memcpy(s, str->s + 5, str->l - 5);
488                                 s[str->l - 5] = 0;
489                                 doff += size;
490                         } else parse_error(fp->n_lines, "unrecognized type");
491                         if (dret == '\n' || dret == '\r') break;
492                 }
493         }
494         b->l_aux = doff - doff0;
495         b->data_len = doff;
496         return z;
497 }
498
499 tamFile sam_open(const char *fn)
500 {
501         tamFile fp;
502         gzFile gzfp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "rb") : gzopen(fn, "rb");
503         if (gzfp == 0) return 0;
504         fp = (tamFile)calloc(1, sizeof(struct __tamFile_t));
505         fp->str = (kstring_t*)calloc(1, sizeof(kstring_t));
506         fp->fp = gzfp;
507         fp->ks = ks_init(fp->fp);
508         return fp;
509 }
510
511 void sam_close(tamFile fp)
512 {
513         if (fp) {
514                 ks_destroy(fp->ks);
515                 gzclose(fp->fp);
516                 free(fp->str->s); free(fp->str);
517                 free(fp);
518         }
519 }