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 "bam.h"
   9 #include "kseq.h"
  10 #include "khash.h"
  11
  12 KSTREAM_INIT(gzFile, gzread, 8192)
  13 KHASH_MAP_INIT_STR(ref, uint64_t)
  14
  15 void bam_init_header_hash(bam_header_t *header);
  16 void bam_destroy_header_hash(bam_header_t *header);
  17 int32_t bam_get_tid(const bam_header_t *header, const char *seq_name);
  18
  19 unsigned char bam_nt16_table[256] = {
  20         15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15,
  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          1, 2, 4, 8, 15,15,15,15, 15,15,15,15, 15, 0 /*=*/,15,15,
  24         15, 1,14, 2, 13,15,15, 4, 11,15,15,12, 15, 3,15,15,
  25         15,15, 5, 6,  8,15, 7, 9, 15,10,15,15, 15,15,15,15,
  26         15, 1,14, 2, 13,15,15, 4, 11,15,15,12, 15, 3,15,15,
  27         15,15, 5, 6,  8,15, 7, 9, 15,10,15,15, 15,15,15,15,
  28         15,15,15,15, 15,15,15,15, 15,15,15,15, 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 };
  37
  38 char *bam_nt16_rev_table = "=ACMGRSVTWYHKDBN";
  39
  40 struct __tamFile_t {
  41         gzFile fp;
  42         kstream_t *ks;
  43         kstring_t *str;
  44         uint64_t n_lines;
  45 };
  46
  47 char **__bam_get_lines(const char *fn, int *_n) // for bam_plcmd.c only
  48 {
  49         char **list = 0, *s;
  50         int n = 0, dret, m = 0;
  51         gzFile fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
  52         kstream_t *ks;
  53         kstring_t *str;
  54         str = (kstring_t*)calloc(1, sizeof(kstring_t));
  55         ks = ks_init(fp);
  56         while (ks_getuntil(ks, '\n', str, &dret) > 0) {
  57                 if (n == m) {
  58                         m = m? m << 1 : 16;
  59                         list = (char**)realloc(list, m * sizeof(char*));
  60                 }
  61                 if (str->s[str->l-1] == '\r')
  62                         str->s[--str->l] = '\0';
  63                 s = list[n++] = (char*)calloc(str->l + 1, 1);
  64                 strcpy(s, str->s);
  65         }
  66         ks_destroy(ks);
  67         gzclose(fp);
  68         free(str->s); free(str);
  69         *_n = n;
  70         return list;
  71 }
  72
  73 static bam_header_t *hash2header(const kh_ref_t *hash)
  74 {
  75         bam_header_t *header;
  76         khiter_t k;
  77         header = bam_header_init();
  78         header->n_targets = kh_size(hash);
  79         header->target_name = (char**)calloc(kh_size(hash), sizeof(char*));
  80         header->target_len = (uint32_t*)calloc(kh_size(hash), 4);
  81         for (k = kh_begin(hash); k != kh_end(hash); ++k) {
  82                 if (kh_exist(hash, k)) {
  83                         int i = (int)kh_value(hash, k);
  84                         header->target_name[i] = (char*)kh_key(hash, k);
  85                         header->target_len[i] = kh_value(hash, k)>>32;
  86                 }
  87         }
  88         bam_init_header_hash(header);
  89         return header;
  90 }
  91 bam_header_t *sam_header_read2(const char *fn)
  92 {
  93         bam_header_t *header;
  94         int c, dret, ret;
  95         gzFile fp;
  96         kstream_t *ks;
  97         kstring_t *str;
  98         kh_ref_t *hash;
  99         khiter_t k;
 100         hash = kh_init(ref);
 101         fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
 102         assert(fp);
 103         ks = ks_init(fp);
 104         str = (kstring_t*)calloc(1, sizeof(kstring_t));
 105         while (ks_getuntil(ks, 0, str, &dret) > 0) {
 106                 char *s = strdup(str->s);
 107                 int len, i;
 108                 i = kh_size(hash);
 109                 ks_getuntil(ks, 0, str, &dret);
 110                 len = atoi(str->s);
 111                 k = kh_put(ref, hash, s, &ret);
 112                 kh_value(hash, k) = (uint64_t)len<<32 | i;
 113                 if (dret != '\n')
 114                         while ((c = ks_getc(ks)) != '\n' && c != -1);
 115         }
 116         ks_destroy(ks);
 117         gzclose(fp);
 118         free(str->s); free(str);
 119         fprintf(stderr, "[sam_header_read2] %d sequences loaded.\n", kh_size(hash));
 120         header = hash2header(hash);
 121         kh_destroy(ref, hash);
 122         return header;
 123 }
 124 static inline uint8_t *alloc_data(bam1_t *b, int size)
 125 {
 126         if (b->m_data < size) {
 127                 b->m_data = size;
 128                 kroundup32(b->m_data);
 129                 b->data = (uint8_t*)realloc(b->data, b->m_data);
 130         }
 131         return b->data;
 132 }
 133 static inline void parse_error(int64_t n_lines, const char * __restrict msg)
 134 {
 135         fprintf(stderr, "Parse error at line %lld: %s\n", (long long)n_lines, msg);
 136         abort();
 137 }
 138 static inline void append_text(bam_header_t *header, kstring_t *str)
 139 {
 140         int x = header->l_text, y = header->l_text + str->l + 2; // 2 = 1 byte dret + 1 byte null
 141         kroundup32(x); kroundup32(y);
 142         if (x < y) header->text = (char*)realloc(header->text, y);
 143         strncpy(header->text + header->l_text, str->s, str->l+1); // we cannot use strcpy() here.
 144         header->l_text += str->l + 1;
 145         header->text[header->l_text] = 0;
 146 }
 147 int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b)
 148 {
 149         int ret, doff, doff0, dret;
 150         bam1_core_t *c = &b->core;
 151         kstring_t *str = fp->str;
 152         kstream_t *ks = fp->ks;
 153
 154         while ((ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret)) >= 0 && str->s[0] == '@') { // skip header
 155                 str->s[str->l] = dret; // note that str->s is NOT null terminated!!
 156                 append_text(header, str);
 157                 if (dret != '\n') {
 158                         ret = ks_getuntil(fp->ks, '\n', str, &dret);
 159                         str->s[str->l] = '\n'; // NOT null terminated!!
 160                         append_text(header, str);
 161                 }
 162                 ++fp->n_lines;
 163         }
 164         while (ret == 0) ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret); // special consideration for "\r\n"
 165         if (ret < 0) return -1;
 166         ++fp->n_lines;
 167         doff = 0;
 168
 169         { // name
 170                 c->l_qname = strlen(str->s) + 1;
 171                 memcpy(alloc_data(b, doff + c->l_qname) + doff, str->s, c->l_qname);
 172                 doff += c->l_qname;
 173         }
 174         { // flag, tid, pos, qual
 175                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); c->flag = atoi(str->s);
 176                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); c->tid = bam_get_tid(header, str->s);
 177                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); c->pos = isdigit(str->s[0])? atoi(str->s) - 1 : -1;
 178                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); c->qual = isdigit(str->s[0])? atoi(str->s) : 0;
 179                 if (ret < 0) return -2;
 180         }
 181         { // cigar
 182                 char *s, *t;
 183                 int i, op;
 184                 long x;
 185                 c->n_cigar = 0;
 186                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -3;
 187                 if (str->s[0] != '*') {
 188                         for (s = str->s; *s; ++s) {
 189                                 if (isalpha(*s)) ++c->n_cigar;
 190                                 else if (!isdigit(*s)) parse_error(fp->n_lines, "invalid CIGAR character");
 191                         }
 192                         b->data = alloc_data(b, doff + c->n_cigar * 4);
 193                         for (i = 0, s = str->s; i != c->n_cigar; ++i) {
 194                                 x = strtol(s, &t, 10);
 195                                 op = toupper(*t);
 196                                 if (op == 'M') op = BAM_CMATCH;
 197                                 else if (op == 'I') op = BAM_CINS;
 198                                 else if (op == 'D') op = BAM_CDEL;
 199                                 else if (op == 'N') op = BAM_CREF_SKIP;
 200                                 else if (op == 'S') op = BAM_CSOFT_CLIP;
 201                                 else if (op == 'H') op = BAM_CHARD_CLIP;
 202                                 else if (op == 'P') op = BAM_CPAD;
 203                                 else parse_error(fp->n_lines, "invalid CIGAR operation");
 204                                 s = t + 1;
 205                                 bam1_cigar(b)[i] = x << BAM_CIGAR_SHIFT | op;
 206                         }
 207                         if (*s) parse_error(fp->n_lines, "unmatched CIGAR operation");
 208                         c->bin = bam_reg2bin(c->pos, bam_calend(c, bam1_cigar(b)));
 209                         doff += c->n_cigar * 4;
 210                 } else c->bin = bam_reg2bin(c->pos, c->pos + 1);
 211         }
 212         { // mtid, mpos, isize
 213                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); c->mtid = strcmp(str->s, "=")? bam_get_tid(header, str->s) : c->tid;
 214                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); c->mpos = isdigit(str->s[0])? atoi(str->s) - 1 : -1;
 215                 ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); c->isize = (str->s[0] == '-' || isdigit(str->s[0]))? atoi(str->s) : 0;
 216                 if (ret < 0) return -4;
 217         }
 218         { // seq and qual
 219                 int i;
 220                 uint8_t *p;
 221                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -5; // seq
 222                 c->l_qseq = strlen(str->s);
 223                 if (c->n_cigar && c->l_qseq != (int32_t)bam_cigar2qlen(c, bam1_cigar(b)))
 224                         parse_error(fp->n_lines, "CIGAR and sequence length are inconsistent");
 225                 p = (uint8_t*)alloc_data(b, doff + c->l_qseq + (c->l_qseq+1)/2) + doff;
 226                 bzero(p, (c->l_qseq+1)/2);
 227                 for (i = 0; i < c->l_qseq; ++i)
 228                         p[i/2] |= bam_nt16_table[(int)str->s[i]] << 4*(1-i%2);
 229                 if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -6; // qual
 230                 if (strcmp(str->s, "*") && c->l_qseq != strlen(str->s))
 231                         parse_error(fp->n_lines, "sequence and quality are inconsistent");
 232                 p += (c->l_qseq+1)/2;
 233                 if (strcmp(str->s, "*") == 0) for (i = 0; i < c->l_qseq; ++i) p[i] = 0xff;
 234                 else for (i = 0; i < c->l_qseq; ++i) p[i] = str->s[i] - 33;
 235                 doff += c->l_qseq + (c->l_qseq+1)/2;
 236         }
 237         doff0 = doff;
 238         if (dret != '\n' && dret != '\r') { // aux
 239                 while (ks_getuntil(ks, KS_SEP_TAB, str, &dret) >= 0) {
 240                         uint8_t *s, type, key[2];
 241                         if (str->l < 6 || str->s[2] != ':' || str->s[4] != ':')
 242                                 parse_error(fp->n_lines, "missing colon in auxiliary data");
 243                         key[0] = str->s[0]; key[1] = str->s[1];
 244                         type = str->s[3];
 245                         s = alloc_data(b, doff + 3) + doff;
 246                         s[0] = key[0]; s[1] = key[1]; s += 2; doff += 2;
 247                         if (type == 'A' || type == 'a' || type == 'c' || type == 'C') { // c and C for backward compatibility
 248                                 s = alloc_data(b, doff + 2) + doff;
 249                                 *s++ = 'A'; *s = str->s[5];
 250                                 doff += 2;
 251                         } else if (type == 'I' || type == 'i') {
 252                                 long long x;
 253                                 s = alloc_data(b, doff + 5) + doff;
 254                                 x = (long long)atoll(str->s + 5);
 255                                 if (x < 0) {
 256                                         if (x >= -127) {
 257                                                 *s++ = 'c'; *(int8_t*)s = (int8_t)x;
 258                                                 s += 1; doff += 2;
 259                                         } else if (x >= -32767) {
 260                                                 *s++ = 's'; *(int16_t*)s = (int16_t)x;
 261                                                 s += 2; doff += 3;
 262                                         } else {
 263                                                 *s++ = 'i'; *(int32_t*)s = (int32_t)x;
 264                                                 s += 4; doff += 5;
 265                                                 if (x < -2147483648ll)
 266                                                         fprintf(stderr, "Parse warning at line %lld: integer %lld is out of range.",
 267                                                                         (long long)fp->n_lines, x);
 268                                         }
 269                                 } else {
 270                                         if (x <= 255) {
 271                                                 *s++ = 'C'; *s++ = (uint8_t)x;
 272                                                 doff += 2;
 273                                         } else if (x <= 65535) {
 274                                                 *s++ = 'S'; *(uint16_t*)s = (uint16_t)x;
 275                                                 s += 2; doff += 3;
 276                                         } else {
 277                                                 *s++ = 'I'; *(uint32_t*)s = (uint32_t)x;
 278                                                 s += 4; doff += 5;
 279                                                 if (x > 4294967295ll)
 280                                                         fprintf(stderr, "Parse warning at line %lld: integer %lld is out of range.",
 281                                                                         (long long)fp->n_lines, x);
 282                                         }
 283                                 }
 284                         } else if (type == 'f') {
 285                                 s = alloc_data(b, doff + 5) + doff;
 286                                 *s++ = 'f';
 287                                 *(float*)s = (float)atof(str->s + 5);
 288                                 s += 4; doff += 5;
 289                         } else if (type == 'd') {
 290                                 s = alloc_data(b, doff + 9) + doff;
 291                                 *s++ = 'd';
 292                                 *(float*)s = (float)atof(str->s + 9);
 293                                 s += 8; doff += 9;
 294                         } else if (type == 'Z' || type == 'H') {
 295                                 int size = 1 + (str->l - 5) + 1;
 296                                 if (type == 'H') { // check whether the hex string is valid
 297                                         int i;
 298                                         if ((str->l - 5) % 2 == 1) parse_error(fp->n_lines, "length of the hex string not even");
 299                                         for (i = 0; i < str->l - 5; ++i) {
 300                                                 int c = toupper(str->s[5 + i]);
 301                                                 if (!((c >= '0' && c <= '9') || (c >= 'A' && c <= 'F')))
 302                                                         parse_error(fp->n_lines, "invalid hex character");
 303                                         }
 304                                 }
 305                                 s = alloc_data(b, doff + size) + doff;
 306                                 *s++ = type;
 307                                 memcpy(s, str->s + 5, str->l - 5);
 308                                 s[str->l - 5] = 0;
 309                                 doff += size;
 310                         } else parse_error(fp->n_lines, "unrecognized type");
 311                         if (dret == '\n' || dret == '\r') break;
 312                 }
 313         }
 314         b->l_aux = doff - doff0;
 315         b->data_len = doff;
 316         return 0;
 317 }
 318
 319 tamFile sam_open(const char *fn)
 320 {
 321         tamFile fp;
 322         fp = (tamFile)calloc(1, sizeof(struct __tamFile_t));
 323         fp->str = (kstring_t*)calloc(1, sizeof(kstring_t));
 324         fp->fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r");
 325         fp->ks = ks_init(fp->fp);
 326         fp->n_lines = 0;
 327         return fp;
 328 }
 329
 330 void sam_close(tamFile fp)
 331 {
 332         if (fp) {
 333                 ks_destroy(fp->ks);
 334                 gzclose(fp->fp);
 335                 free(fp->str->s); free(fp->str);
 336                 free(fp);
 337         }
 338 }
 339
 340 static void taf2baf_core(const char *fntaf, const char *fnbaf, bam_header_t *header)
 341 {
 342         bamFile fpbaf;
 343         bam1_t *b;
 344         tamFile fp;
 345         int ret;
 346
 347         b = (bam1_t*)calloc(1, sizeof(bam1_t));
 348         fpbaf = (strcmp(fnbaf, "-") == 0)? bam_dopen(fileno(stdout), "w") : bam_open(fnbaf, "w");
 349         fp = sam_open(fntaf);
 350         ret = sam_read1(fp, header, b);
 351         bam_header_write(fpbaf, header);
 352         if (ret >= 0) {
 353                 bam_write1(fpbaf, b);
 354                 while (sam_read1(fp, header, b) >= 0) bam_write1(fpbaf, b);
 355         }
 356         bam_close(fpbaf);
 357         free(b->data); free(b);
 358         sam_close(fp);
 359 }
 360
 361 int bam_taf2baf(int argc, char *argv[])
 362 {
 363         int c;
 364         bam_header_t *header;
 365
 366         while ((c = getopt(argc, argv, "")) >= 0) {
 367         }
 368         if (optind + 3 > argc) {
 369                 fprintf(stderr, "Usage: bamtk import <in.ref_list> <in.sam> <out.bam>\n");
 370                 return 1;
 371         }
 372         header = sam_header_read2(argv[optind]);
 373         taf2baf_core(argv[optind+1], argv[optind+2], header);
 374         bam_header_destroy(header);
 375         return 0;
 376 }