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