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;
 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                 kh_value(hash, k) = (uint64_t)len<<32 | i;
 139                 if (dret != '\n')
 140                         while ((c = ks_getc(ks)) != '\n' && c != -1);
 141         }
 142         ks_destroy(ks);
 143         gzclose(fp);
 144         free(str->s); free(str);
 145         fprintf(stderr, "[sam_header_read2] %d sequences loaded.\n", kh_size(hash));
 146         header = hash2header(hash);
 147         kh_destroy(ref, hash);
 148         return header;
 149 }
 150 static inline uint8_t *alloc_data(bam1_t *b, int size)
 151 {
 152         if (b->m_data < size) {
 153                 b->m_data = size;
 154                 kroundup32(b->m_data);
 155                 b->data = (uint8_t*)realloc(b->data, b->m_data);
 156         }
 157         return b->data;
 158 }
 159 static inline void parse_error(int64_t n_lines, const char * __restrict msg)
 160 {
 161         fprintf(stderr, "Parse error at line %lld: %s\n", (long long)n_lines, msg);
 162         abort();
 163 }
 164 static inline void append_text(bam_header_t *header, kstring_t *str)
 165 {
 166         int x = header->l_text, y = header->l_text + str->l + 2; // 2 = 1 byte dret + 1 byte null
 167         kroundup32(x); kroundup32(y);
 168         if (x < y) header->text = (char*)realloc(header->text, y);
 169         strncpy(header->text + header->l_text, str->s, str->l+1); // we cannot use strcpy() here.
 170         header->l_text += str->l + 1;
 171         header->text[header->l_text] = 0;
 172 }
 173
 174 int sam_header_parse(bam_header_t *h)
 175 {
 176         char **tmp;
 177         int i;
 178         free(h->target_len); free(h->target_name);
 179         h->n_targets = 0; h->target_len = 0; h->target_name = 0;
 180         if (h->l_text < 3) return 0;
 181         if (h->dict == 0) h->dict = sam_header_parse2(h->text);
 182         tmp = sam_header2list(h->dict, "SQ", "SN", &h->n_targets);
 183         if (h->n_targets == 0) return 0;
 184         h->target_name = calloc(h->n_targets, sizeof(void*));
 185         for (i = 0; i < h->n_targets; ++i)
 186                 h->target_name[i] = strdup(tmp[i]);
 187         free(tmp);
 188         tmp = sam_header2list(h->dict, "SQ", "LN", &h->n_targets);
 189         h->target_len = calloc(h->n_targets, 4);
 190         for (i = 0; i < h->n_targets; ++i)
 191                 h->target_len[i] = atoi(tmp[i]);
 192         free(tmp);
 193         return h->n_targets;
 194 }
 195
 196 bam_header_t *sam_header_read(tamFile fp)
 197 {
 198         int ret, dret;
 199         bam_header_t *header = bam_header_init();
 200         kstring_t *str = fp->str;
 201         while ((ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret)) >= 0 && str->s[0] == '@') { // skip header
 202                 str->s[str->l] = dret; // note that str->s is NOT null terminated!!
 203                 append_text(header, str);
 204                 if (dret != '\n') {
 205                         ret = ks_getuntil(fp->ks, '\n', str, &dret);
 206                         str->s[str->l] = '\n'; // NOT null terminated!!
 207                         append_text(header, str);
 208                 }
 209                 ++fp->n_lines;
 210         }
 211         sam_header_parse(header);
 212         bam_init_header_hash(header);
 213         fp->is_first = 1;
 214         return header;
 215 }
 216
 217 int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b)
 218 {
 219         int ret, doff, doff0, dret, z = 0;
 220         bam1_core_t *c = &b->core;
 221         kstring_t *str = fp->str;
 222         kstream_t *ks = fp->ks;
 223
 224         if (fp->is_first) {
 225                 fp->is_first = 0;
 226                 ret = str->l;
 227         } else {
 228                 do { // special consideration for empty lines
 229                         ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret);
 230                         if (ret >= 0) z += str->l + 1;
 231                 } while (ret == 0);
 232         }
 233         if (ret < 0) return -1;
 234         ++fp->n_lines;
 235         doff = 0;
 236
 237         { // name
 238                 c->l_qname = strlen(str->s) + 1;
 239                 memcpy(alloc_data(b, doff + c->l_qname) + doff, str->s, c->l_qname);
 240                 doff += c->l_qname;
 241         }
 242         { // flag
 243                 long flag;
 244                 char *s;
 245                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
 246                 flag = strtol((char*)str->s, &s, 0);
 247                 if (*s) { // not the end of the string
 248                         flag = 0;
 249                         for (s = str->s; *s; ++s)
 250                                 flag |= bam_char2flag_table[(int)*s];
 251                 }
 252                 c->flag = flag;
 253         }
 254         { // tid, pos, qual
 255                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->tid = bam_get_tid(header, str->s);
 256                 if (c->tid < 0 && strcmp(str->s, "*")) {
 257                         if (header->n_targets == 0) {
 258                                 fprintf(stderr, "[sam_read1] missing header? Abort!\n");
 259                                 exit(1);
 260                         } else fprintf(stderr, "[sam_read1] reference '%s' is recognized as '*'.\n", str->s);
 261                 }
 262                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->pos = isdigit(str->s[0])? atoi(str->s) - 1 : -1;
 263                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->qual = isdigit(str->s[0])? atoi(str->s) : 0;
 264                 if (ret < 0) return -2;
 265         }
 266         { // cigar
 267                 char *s, *t;
 268                 int i, op;
 269                 long x;
 270                 c->n_cigar = 0;
 271                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -3;
 272                 z += str->l + 1;
 273                 if (str->s[0] != '*') {
 274                         for (s = str->s; *s; ++s) {
 275                                 if (isalpha(*s)) ++c->n_cigar;
 276                                 else if (!isdigit(*s)) parse_error(fp->n_lines, "invalid CIGAR character");
 277                         }
 278                         b->data = alloc_data(b, doff + c->n_cigar * 4);
 279                         for (i = 0, s = str->s; i != c->n_cigar; ++i) {
 280                                 x = strtol(s, &t, 10);
 281                                 op = toupper(*t);
 282                                 if (op == 'M' || op == '=' || op == 'X') op = BAM_CMATCH;
 283                                 else if (op == 'I') op = BAM_CINS;
 284                                 else if (op == 'D') op = BAM_CDEL;
 285                                 else if (op == 'N') op = BAM_CREF_SKIP;
 286                                 else if (op == 'S') op = BAM_CSOFT_CLIP;
 287                                 else if (op == 'H') op = BAM_CHARD_CLIP;
 288                                 else if (op == 'P') op = BAM_CPAD;
 289                                 else parse_error(fp->n_lines, "invalid CIGAR operation");
 290                                 s = t + 1;
 291                                 bam1_cigar(b)[i] = x << BAM_CIGAR_SHIFT | op;
 292                         }
 293                         if (*s) parse_error(fp->n_lines, "unmatched CIGAR operation");
 294                         c->bin = bam_reg2bin(c->pos, bam_calend(c, bam1_cigar(b)));
 295                         doff += c->n_cigar * 4;
 296                 } else {
 297                         if (!(c->flag&BAM_FUNMAP)) {
 298                                 fprintf(stderr, "Parse warning at line %lld: mapped sequence without CIGAR\n", (long long)fp->n_lines);
 299                                 c->flag |= BAM_FUNMAP;
 300                         }
 301                         c->bin = bam_reg2bin(c->pos, c->pos + 1);
 302                 }
 303         }
 304         { // mtid, mpos, isize
 305                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
 306                 c->mtid = strcmp(str->s, "=")? bam_get_tid(header, str->s) : c->tid;
 307                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
 308                 c->mpos = isdigit(str->s[0])? atoi(str->s) - 1 : -1;
 309                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1;
 310                 c->isize = (str->s[0] == '-' || isdigit(str->s[0]))? atoi(str->s) : 0;
 311                 if (ret < 0) return -4;
 312         }
 313         { // seq and qual
 314                 int i;
 315                 uint8_t *p = 0;
 316                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -5; // seq
 317                 z += str->l + 1;
 318                 if (strcmp(str->s, "*")) {
 319                         c->l_qseq = strlen(str->s);
 320                         if (c->n_cigar && c->l_qseq != (int32_t)bam_cigar2qlen(c, bam1_cigar(b)))
 321                                 parse_error(fp->n_lines, "CIGAR and sequence length are inconsistent");
 322                         p = (uint8_t*)alloc_data(b, doff + c->l_qseq + (c->l_qseq+1)/2) + doff;
 323                         memset(p, 0, (c->l_qseq+1)/2);
 324                         for (i = 0; i < c->l_qseq; ++i)
 325                                 p[i/2] |= bam_nt16_table[(int)str->s[i]] << 4*(1-i%2);
 326                 } else c->l_qseq = 0;
 327                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -6; // qual
 328                 z += str->l + 1;
 329                 if (strcmp(str->s, "*") && c->l_qseq != strlen(str->s))
 330                         parse_error(fp->n_lines, "sequence and quality are inconsistent");
 331                 p += (c->l_qseq+1)/2;
 332                 if (strcmp(str->s, "*") == 0) for (i = 0; i < c->l_qseq; ++i) p[i] = 0xff;
 333                 else for (i = 0; i < c->l_qseq; ++i) p[i] = str->s[i] - 33;
 334                 doff += c->l_qseq + (c->l_qseq+1)/2;
 335         }
 336         doff0 = doff;
 337         if (dret != '\n' && dret != '\r') { // aux
 338                 while (ks_getuntil(ks, KS_SEP_TAB, str, &dret) >= 0) {
 339                         uint8_t *s, type, key[2];
 340                         z += str->l + 1;
 341                         if (str->l < 6 || str->s[2] != ':' || str->s[4] != ':')
 342                                 parse_error(fp->n_lines, "missing colon in auxiliary data");
 343                         key[0] = str->s[0]; key[1] = str->s[1];
 344                         type = str->s[3];
 345                         s = alloc_data(b, doff + 3) + doff;
 346                         s[0] = key[0]; s[1] = key[1]; s += 2; doff += 2;
 347                         if (type == 'A' || type == 'a' || type == 'c' || type == 'C') { // c and C for backward compatibility
 348                                 s = alloc_data(b, doff + 2) + doff;
 349                                 *s++ = 'A'; *s = str->s[5];
 350                                 doff += 2;
 351                         } else if (type == 'I' || type == 'i') {
 352                                 long long x;
 353                                 s = alloc_data(b, doff + 5) + doff;
 354                                 x = (long long)atoll(str->s + 5);
 355                                 if (x < 0) {
 356                                         if (x >= -127) {
 357                                                 *s++ = 'c'; *(int8_t*)s = (int8_t)x;
 358                                                 s += 1; doff += 2;
 359                                         } else if (x >= -32767) {
 360                                                 *s++ = 's'; *(int16_t*)s = (int16_t)x;
 361                                                 s += 2; doff += 3;
 362                                         } else {
 363                                                 *s++ = 'i'; *(int32_t*)s = (int32_t)x;
 364                                                 s += 4; doff += 5;
 365                                                 if (x < -2147483648ll)
 366                                                         fprintf(stderr, "Parse warning at line %lld: integer %lld is out of range.",
 367                                                                         (long long)fp->n_lines, x);
 368                                         }
 369                                 } else {
 370                                         if (x <= 255) {
 371                                                 *s++ = 'C'; *s++ = (uint8_t)x;
 372                                                 doff += 2;
 373                                         } else if (x <= 65535) {
 374                                                 *s++ = 'S'; *(uint16_t*)s = (uint16_t)x;
 375                                                 s += 2; doff += 3;
 376                                         } else {
 377                                                 *s++ = 'I'; *(uint32_t*)s = (uint32_t)x;
 378                                                 s += 4; doff += 5;
 379                                                 if (x > 4294967295ll)
 380                                                         fprintf(stderr, "Parse warning at line %lld: integer %lld is out of range.",
 381                                                                         (long long)fp->n_lines, x);
 382                                         }
 383                                 }
 384                         } else if (type == 'f') {
 385                                 s = alloc_data(b, doff + 5) + doff;
 386                                 *s++ = 'f';
 387                                 *(float*)s = (float)atof(str->s + 5);
 388                                 s += 4; doff += 5;
 389                         } else if (type == 'd') {
 390                                 s = alloc_data(b, doff + 9) + doff;
 391                                 *s++ = 'd';
 392                                 *(float*)s = (float)atof(str->s + 9);
 393                                 s += 8; doff += 9;
 394                         } else if (type == 'Z' || type == 'H') {
 395                                 int size = 1 + (str->l - 5) + 1;
 396                                 if (type == 'H') { // check whether the hex string is valid
 397                                         int i;
 398                                         if ((str->l - 5) % 2 == 1) parse_error(fp->n_lines, "length of the hex string not even");
 399                                         for (i = 0; i < str->l - 5; ++i) {
 400                                                 int c = toupper(str->s[5 + i]);
 401                                                 if (!((c >= '0' && c <= '9') || (c >= 'A' && c <= 'F')))
 402                                                         parse_error(fp->n_lines, "invalid hex character");
 403                                         }
 404                                 }
 405                                 s = alloc_data(b, doff + size) + doff;
 406                                 *s++ = type;
 407                                 memcpy(s, str->s + 5, str->l - 5);
 408                                 s[str->l - 5] = 0;
 409                                 doff += size;
 410                         } else parse_error(fp->n_lines, "unrecognized type");
 411                         if (dret == '\n' || dret == '\r') break;
 412                 }
 413         }
 414         b->l_aux = doff - doff0;
 415         b->data_len = doff;
 416         return z;
 417 }
 418
 419 tamFile sam_open(const char *fn)
 420 {
 421         tamFile fp;
 422         gzFile gzfp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "rb") : gzopen(fn, "rb");
 423         if (gzfp == 0) return 0;
 424         fp = (tamFile)calloc(1, sizeof(struct __tamFile_t));
 425         fp->str = (kstring_t*)calloc(1, sizeof(kstring_t));
 426         fp->fp = gzfp;
 427         fp->ks = ks_init(fp->fp);
 428         return fp;
 429 }
 430
 431 void sam_close(tamFile fp)
 432 {
 433         if (fp) {
 434                 ks_destroy(fp->ks);
 435                 gzclose(fp->fp);
 436                 free(fp->str->s); free(fp->str);
 437                 free(fp);
 438         }
 439 }