Merge commit 'upstream/0.1.14'
[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 "sam_header.h"
14 #include "kseq.h"
15 #include "khash.h"
16
17 KSTREAM_INIT(gzFile, gzread, 8192)
18 KHASH_MAP_INIT_STR(ref, uint64_t)
19
20 void bam_init_header_hash(bam_header_t *header);
21 void bam_destroy_header_hash(bam_header_t *header);
22 int32_t bam_get_tid(const bam_header_t *header, const char *seq_name);
23
24 unsigned char bam_nt16_table[256] = {
25         15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
26         15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
27         15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
28          1, 2, 4, 8, 15,15,15,15, 15,15,15,15, 15, 0 /*=*/,15,15,
29         15, 1,14, 2, 13,15,15, 4, 11,15,15,12, 15, 3,15,15,
30         15,15, 5, 6,  8,15, 7, 9, 15,10,15,15, 15,15,15,15,
31         15, 1,14, 2, 13,15,15, 4, 11,15,15,12, 15, 3,15,15,
32         15,15, 5, 6,  8,15, 7, 9, 15,10,15,15, 15,15,15,15,
33         15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
34         15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
35         15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
36         15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
37         15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
38         15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
39         15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
40         15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15
41 };
42
43 unsigned short bam_char2flag_table[256] = {
44         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
45         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
46         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
47         0,BAM_FREAD1,BAM_FREAD2,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
48         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
49         BAM_FPROPER_PAIR,0,BAM_FMREVERSE,0, 0,BAM_FMUNMAP,0,0, 0,0,0,0, 0,0,0,0,
50         0,0,0,0, BAM_FDUP,0,BAM_FQCFAIL,0, 0,0,0,0, 0,0,0,0,
51         BAM_FPAIRED,0,BAM_FREVERSE,BAM_FSECONDARY, 0,BAM_FUNMAP,0,0, 0,0,0,0, 0,0,0,0,
52         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
53         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
54         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
55         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
56         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
57         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
58         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0,
59         0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0
60 };
61
62 char *bam_nt16_rev_table = "=ACMGRSVTWYHKDBN";
63
64 struct __tamFile_t {
65         gzFile fp;
66         kstream_t *ks;
67         kstring_t *str;
68         uint64_t n_lines;
69         int is_first;
70 };
71
72 char **__bam_get_lines(const char *fn, int *_n) // for bam_plcmd.c only
73 {
74         char **list = 0, *s;
75         int n = 0, dret, m = 0;
76         gzFile fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
77         kstream_t *ks;
78         kstring_t *str;
79         str = (kstring_t*)calloc(1, sizeof(kstring_t));
80         ks = ks_init(fp);
81         while (ks_getuntil(ks, '\n', str, &dret) > 0) {
82                 if (n == m) {
83                         m = m? m << 1 : 16;
84                         list = (char**)realloc(list, m * sizeof(char*));
85                 }
86                 if (str->s[str->l-1] == '\r')
87                         str->s[--str->l] = '\0';
88                 s = list[n++] = (char*)calloc(str->l + 1, 1);
89                 strcpy(s, str->s);
90         }
91         ks_destroy(ks);
92         gzclose(fp);
93         free(str->s); free(str);
94         *_n = n;
95         return list;
96 }
97
98 static bam_header_t *hash2header(const kh_ref_t *hash)
99 {
100         bam_header_t *header;
101         khiter_t k;
102         header = bam_header_init();
103         header->n_targets = kh_size(hash);
104         header->target_name = (char**)calloc(kh_size(hash), sizeof(char*));
105         header->target_len = (uint32_t*)calloc(kh_size(hash), 4);
106         for (k = kh_begin(hash); k != kh_end(hash); ++k) {
107                 if (kh_exist(hash, k)) {
108                         int i = (int)kh_value(hash, k);
109                         header->target_name[i] = (char*)kh_key(hash, k);
110                         header->target_len[i] = kh_value(hash, k)>>32;
111                 }
112         }
113         bam_init_header_hash(header);
114         return header;
115 }
116 bam_header_t *sam_header_read2(const char *fn)
117 {
118         bam_header_t *header;
119         int c, dret, ret, error = 0;
120         gzFile fp;
121         kstream_t *ks;
122         kstring_t *str;
123         kh_ref_t *hash;
124         khiter_t k;
125         if (fn == 0) return 0;
126         fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
127         if (fp == 0) return 0;
128         hash = kh_init(ref);
129         ks = ks_init(fp);
130         str = (kstring_t*)calloc(1, sizeof(kstring_t));
131         while (ks_getuntil(ks, 0, str, &dret) > 0) {
132                 char *s = strdup(str->s);
133                 int len, i;
134                 i = kh_size(hash);
135                 ks_getuntil(ks, 0, str, &dret);
136                 len = atoi(str->s);
137                 k = kh_put(ref, hash, s, &ret);
138                 if (ret == 0) {
139                         fprintf(stderr, "[sam_header_read2] duplicated sequence name: %s\n", s);
140                         error = 1;
141                 }
142                 kh_value(hash, k) = (uint64_t)len<<32 | i;
143                 if (dret != '\n')
144                         while ((c = ks_getc(ks)) != '\n' && c != -1);
145         }
146         ks_destroy(ks);
147         gzclose(fp);
148         free(str->s); free(str);
149         fprintf(stderr, "[sam_header_read2] %d sequences loaded.\n", kh_size(hash));
150         if (error) return 0;
151         header = hash2header(hash);
152         kh_destroy(ref, hash);
153         return header;
154 }
155 static inline uint8_t *alloc_data(bam1_t *b, int size)
156 {
157         if (b->m_data < size) {
158                 b->m_data = size;
159                 kroundup32(b->m_data);
160                 b->data = (uint8_t*)realloc(b->data, b->m_data);
161         }
162         return b->data;
163 }
164 static inline void parse_error(int64_t n_lines, const char * __restrict msg)
165 {
166         fprintf(stderr, "Parse error at line %lld: %s\n", (long long)n_lines, msg);
167         abort();
168 }
169 static inline void append_text(bam_header_t *header, kstring_t *str)
170 {
171         size_t x = header->l_text, y = header->l_text + str->l + 2; // 2 = 1 byte dret + 1 byte null
172         kroundup32(x); kroundup32(y);
173         if (x < y) 
174     {
175         header->n_text = y;
176         header->text = (char*)realloc(header->text, y);
177         if ( !header->text ) 
178         {
179             fprintf(stderr,"realloc failed to alloc %ld bytes\n", y);
180             abort();
181         }
182     }
183     // Sanity check
184     if ( header->l_text+str->l+1 >= header->n_text )
185     {
186         fprintf(stderr,"append_text FIXME: %ld>=%ld, x=%ld,y=%ld\n",  header->l_text+str->l+1,header->n_text,x,y);
187         abort();
188     }
189         strncpy(header->text + header->l_text, str->s, str->l+1); // we cannot use strcpy() here.
190         header->l_text += str->l + 1;
191         header->text[header->l_text] = 0;
192 }
193
194 int sam_header_parse(bam_header_t *h)
195 {
196         char **tmp;
197         int i;
198         free(h->target_len); free(h->target_name);
199         h->n_targets = 0; h->target_len = 0; h->target_name = 0;
200         if (h->l_text < 3) return 0;
201         if (h->dict == 0) h->dict = sam_header_parse2(h->text);
202         tmp = sam_header2list(h->dict, "SQ", "SN", &h->n_targets);
203         if (h->n_targets == 0) return 0;
204         h->target_name = calloc(h->n_targets, sizeof(void*));
205         for (i = 0; i < h->n_targets; ++i)
206                 h->target_name[i] = strdup(tmp[i]);
207         free(tmp);
208         tmp = sam_header2list(h->dict, "SQ", "LN", &h->n_targets);
209         h->target_len = calloc(h->n_targets, 4);
210         for (i = 0; i < h->n_targets; ++i)
211                 h->target_len[i] = atoi(tmp[i]);
212         free(tmp);
213         return h->n_targets;
214 }
215
216 bam_header_t *sam_header_read(tamFile fp)
217 {
218         int ret, dret;
219         bam_header_t *header = bam_header_init();
220         kstring_t *str = fp->str;
221         while ((ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret)) >= 0 && str->s[0] == '@') { // skip header
222                 str->s[str->l] = dret; // note that str->s is NOT null terminated!!
223                 append_text(header, str);
224                 if (dret != '\n') {
225                         ret = ks_getuntil(fp->ks, '\n', str, &dret);
226                         str->s[str->l] = '\n'; // NOT null terminated!!
227                         append_text(header, str);
228                 }
229                 ++fp->n_lines;
230         }
231         sam_header_parse(header);
232         bam_init_header_hash(header);
233         fp->is_first = 1;
234         return header;
235 }
236
237 int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b)
238 {
239         int ret, doff, doff0, dret, z = 0;
240         bam1_core_t *c = &b->core;
241         kstring_t *str = fp->str;
242         kstream_t *ks = fp->ks;
243
244         if (fp->is_first) {
245                 fp->is_first = 0;
246                 ret = str->l;
247         } else {
248                 do { // special consideration for empty lines
249                         ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret);
250                         if (ret >= 0) z += str->l + 1;
251                 } while (ret == 0);
252         }
253         if (ret < 0) return -1;
254         ++fp->n_lines;
255         doff = 0;
256
257         { // name
258                 c->l_qname = strlen(str->s) + 1;
259                 memcpy(alloc_data(b, doff + c->l_qname) + doff, str->s, c->l_qname);
260                 doff += c->l_qname;
261         }
262         { // flag
263                 long flag;
264                 char *s;
265                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
266                 flag = strtol((char*)str->s, &s, 0);
267                 if (*s) { // not the end of the string
268                         flag = 0;
269                         for (s = str->s; *s; ++s)
270                                 flag |= bam_char2flag_table[(int)*s];
271                 }
272                 c->flag = flag;
273         }
274         { // tid, pos, qual
275                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->tid = bam_get_tid(header, str->s);
276                 if (c->tid < 0 && strcmp(str->s, "*")) {
277                         if (header->n_targets == 0) {
278                                 fprintf(stderr, "[sam_read1] missing header? Abort!\n");
279                                 exit(1);
280                         } else fprintf(stderr, "[sam_read1] reference '%s' is recognized as '*'.\n", str->s);
281                 }
282                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->pos = isdigit(str->s[0])? atoi(str->s) - 1 : -1;
283                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->qual = isdigit(str->s[0])? atoi(str->s) : 0;
284                 if (ret < 0) return -2;
285         }
286         { // cigar
287                 char *s, *t;
288                 int i, op;
289                 long x;
290                 c->n_cigar = 0;
291                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -3;
292                 z += str->l + 1;
293                 if (str->s[0] != '*') {
294                         for (s = str->s; *s; ++s) {
295                                 if (isalpha(*s)) ++c->n_cigar;
296                                 else if (!isdigit(*s)) parse_error(fp->n_lines, "invalid CIGAR character");
297                         }
298                         b->data = alloc_data(b, doff + c->n_cigar * 4);
299                         for (i = 0, s = str->s; i != c->n_cigar; ++i) {
300                                 x = strtol(s, &t, 10);
301                                 op = toupper(*t);
302                                 if (op == 'M' || op == '=' || op == 'X') op = BAM_CMATCH;
303                                 else if (op == 'I') op = BAM_CINS;
304                                 else if (op == 'D') op = BAM_CDEL;
305                                 else if (op == 'N') op = BAM_CREF_SKIP;
306                                 else if (op == 'S') op = BAM_CSOFT_CLIP;
307                                 else if (op == 'H') op = BAM_CHARD_CLIP;
308                                 else if (op == 'P') op = BAM_CPAD;
309                                 else parse_error(fp->n_lines, "invalid CIGAR operation");
310                                 s = t + 1;
311                                 bam1_cigar(b)[i] = x << BAM_CIGAR_SHIFT | op;
312                         }
313                         if (*s) parse_error(fp->n_lines, "unmatched CIGAR operation");
314                         c->bin = bam_reg2bin(c->pos, bam_calend(c, bam1_cigar(b)));
315                         doff += c->n_cigar * 4;
316                 } else {
317                         if (!(c->flag&BAM_FUNMAP)) {
318                                 fprintf(stderr, "Parse warning at line %lld: mapped sequence without CIGAR\n", (long long)fp->n_lines);
319                                 c->flag |= BAM_FUNMAP;
320                         }
321                         c->bin = bam_reg2bin(c->pos, c->pos + 1);
322                 }
323         }
324         { // mtid, mpos, isize
325                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
326                 c->mtid = strcmp(str->s, "=")? bam_get_tid(header, str->s) : c->tid;
327                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
328                 c->mpos = isdigit(str->s[0])? atoi(str->s) - 1 : -1;
329                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
330                 c->isize = (str->s[0] == '-' || isdigit(str->s[0]))? atoi(str->s) : 0;
331                 if (ret < 0) return -4;
332         }
333         { // seq and qual
334                 int i;
335                 uint8_t *p = 0;
336                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -5; // seq
337                 z += str->l + 1;
338                 if (strcmp(str->s, "*")) {
339                         c->l_qseq = strlen(str->s);
340                         if (c->n_cigar && c->l_qseq != (int32_t)bam_cigar2qlen(c, bam1_cigar(b)))
341                                 parse_error(fp->n_lines, "CIGAR and sequence length are inconsistent");
342                         p = (uint8_t*)alloc_data(b, doff + c->l_qseq + (c->l_qseq+1)/2) + doff;
343                         memset(p, 0, (c->l_qseq+1)/2);
344                         for (i = 0; i < c->l_qseq; ++i)
345                                 p[i/2] |= bam_nt16_table[(int)str->s[i]] << 4*(1-i%2);
346                 } else c->l_qseq = 0;
347                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -6; // qual
348                 z += str->l + 1;
349                 if (strcmp(str->s, "*") && c->l_qseq != strlen(str->s))
350                         parse_error(fp->n_lines, "sequence and quality are inconsistent");
351                 p += (c->l_qseq+1)/2;
352                 if (strcmp(str->s, "*") == 0) for (i = 0; i < c->l_qseq; ++i) p[i] = 0xff;
353                 else for (i = 0; i < c->l_qseq; ++i) p[i] = str->s[i] - 33;
354                 doff += c->l_qseq + (c->l_qseq+1)/2;
355         }
356         doff0 = doff;
357         if (dret != '\n' && dret != '\r') { // aux
358                 while (ks_getuntil(ks, KS_SEP_TAB, str, &dret) >= 0) {
359                         uint8_t *s, type, key[2];
360                         z += str->l + 1;
361                         if (str->l < 6 || str->s[2] != ':' || str->s[4] != ':')
362                                 parse_error(fp->n_lines, "missing colon in auxiliary data");
363                         key[0] = str->s[0]; key[1] = str->s[1];
364                         type = str->s[3];
365                         s = alloc_data(b, doff + 3) + doff;
366                         s[0] = key[0]; s[1] = key[1]; s += 2; doff += 2;
367                         if (type == 'A' || type == 'a' || type == 'c' || type == 'C') { // c and C for backward compatibility
368                                 s = alloc_data(b, doff + 2) + doff;
369                                 *s++ = 'A'; *s = str->s[5];
370                                 doff += 2;
371                         } else if (type == 'I' || type == 'i') {
372                                 long long x;
373                                 s = alloc_data(b, doff + 5) + doff;
374                                 x = (long long)atoll(str->s + 5);
375                                 if (x < 0) {
376                                         if (x >= -127) {
377                                                 *s++ = 'c'; *(int8_t*)s = (int8_t)x;
378                                                 s += 1; doff += 2;
379                                         } else if (x >= -32767) {
380                                                 *s++ = 's'; *(int16_t*)s = (int16_t)x;
381                                                 s += 2; doff += 3;
382                                         } else {
383                                                 *s++ = 'i'; *(int32_t*)s = (int32_t)x;
384                                                 s += 4; doff += 5;
385                                                 if (x < -2147483648ll)
386                                                         fprintf(stderr, "Parse warning at line %lld: integer %lld is out of range.",
387                                                                         (long long)fp->n_lines, x);
388                                         }
389                                 } else {
390                                         if (x <= 255) {
391                                                 *s++ = 'C'; *s++ = (uint8_t)x;
392                                                 doff += 2;
393                                         } else if (x <= 65535) {
394                                                 *s++ = 'S'; *(uint16_t*)s = (uint16_t)x;
395                                                 s += 2; doff += 3;
396                                         } else {
397                                                 *s++ = 'I'; *(uint32_t*)s = (uint32_t)x;
398                                                 s += 4; doff += 5;
399                                                 if (x > 4294967295ll)
400                                                         fprintf(stderr, "Parse warning at line %lld: integer %lld is out of range.",
401                                                                         (long long)fp->n_lines, x);
402                                         }
403                                 }
404                         } else if (type == 'f') {
405                                 s = alloc_data(b, doff + 5) + doff;
406                                 *s++ = 'f';
407                                 *(float*)s = (float)atof(str->s + 5);
408                                 s += 4; doff += 5;
409                         } else if (type == 'd') {
410                                 s = alloc_data(b, doff + 9) + doff;
411                                 *s++ = 'd';
412                                 *(float*)s = (float)atof(str->s + 9);
413                                 s += 8; doff += 9;
414                         } else if (type == 'Z' || type == 'H') {
415                                 int size = 1 + (str->l - 5) + 1;
416                                 if (type == 'H') { // check whether the hex string is valid
417                                         int i;
418                                         if ((str->l - 5) % 2 == 1) parse_error(fp->n_lines, "length of the hex string not even");
419                                         for (i = 0; i < str->l - 5; ++i) {
420                                                 int c = toupper(str->s[5 + i]);
421                                                 if (!((c >= '0' && c <= '9') || (c >= 'A' && c <= 'F')))
422                                                         parse_error(fp->n_lines, "invalid hex character");
423                                         }
424                                 }
425                                 s = alloc_data(b, doff + size) + doff;
426                                 *s++ = type;
427                                 memcpy(s, str->s + 5, str->l - 5);
428                                 s[str->l - 5] = 0;
429                                 doff += size;
430                         } else parse_error(fp->n_lines, "unrecognized type");
431                         if (dret == '\n' || dret == '\r') break;
432                 }
433         }
434         b->l_aux = doff - doff0;
435         b->data_len = doff;
436         return z;
437 }
438
439 tamFile sam_open(const char *fn)
440 {
441         tamFile fp;
442         gzFile gzfp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "rb") : gzopen(fn, "rb");
443         if (gzfp == 0) return 0;
444         fp = (tamFile)calloc(1, sizeof(struct __tamFile_t));
445         fp->str = (kstring_t*)calloc(1, sizeof(kstring_t));
446         fp->fp = gzfp;
447         fp->ks = ks_init(fp->fp);
448         return fp;
449 }
450
451 void sam_close(tamFile fp)
452 {
453         if (fp) {
454                 ks_destroy(fp->ks);
455                 gzclose(fp->fp);
456                 free(fp->str->s); free(fp->str);
457                 free(fp);
458         }
459 }