bam_pileup.c

   1 #include <stdio.h>
   2 #include <stdlib.h>
   3 #include <ctype.h>
   4 #include <assert.h>
   5 #include "sam.h"
   6
   7 typedef struct {
   8         int k, x, y, end;
   9 } cstate_t;
  10
  11 static cstate_t g_cstate_null = { -1, 0, 0, 0 };
  12
  13 typedef struct __linkbuf_t {
  14         bam1_t b;
  15         uint32_t beg, end;
  16         cstate_t s;
  17         struct __linkbuf_t *next;
  18 } lbnode_t;
  19
  20 /* --- BEGIN: Memory pool */
  21
  22 typedef struct {
  23         int cnt, n, max;
  24         lbnode_t **buf;
  25 } mempool_t;
  26
  27 static mempool_t *mp_init()
  28 {
  29         mempool_t *mp;
  30         mp = (mempool_t*)calloc(1, sizeof(mempool_t));
  31         return mp;
  32 }
  33 static void mp_destroy(mempool_t *mp)
  34 {
  35         int k;
  36         for (k = 0; k < mp->n; ++k) {
  37                 free(mp->buf[k]->b.data);
  38                 free(mp->buf[k]);
  39         }
  40         free(mp->buf);
  41         free(mp);
  42 }
  43 static inline lbnode_t *mp_alloc(mempool_t *mp)
  44 {
  45         ++mp->cnt;
  46         if (mp->n == 0) return (lbnode_t*)calloc(1, sizeof(lbnode_t));
  47         else return mp->buf[--mp->n];
  48 }
  49 static inline void mp_free(mempool_t *mp, lbnode_t *p)
  50 {
  51         --mp->cnt; p->next = 0; // clear lbnode_t::next here
  52         if (mp->n == mp->max) {
  53                 mp->max = mp->max? mp->max<<1 : 256;
  54                 mp->buf = (lbnode_t**)realloc(mp->buf, sizeof(lbnode_t*) * mp->max);
  55         }
  56         mp->buf[mp->n++] = p;
  57 }
  58
  59 /* --- END: Memory pool */
  60
  61 /* --- BEGIN: Auxiliary functions */
  62
  63 /* s->k: the index of the CIGAR operator that has just been processed.
  64    s->x: the reference coordinate of the start of s->k
  65    s->y: the query coordiante of the start of s->k
  66  */
  67 static inline int resolve_cigar2(bam_pileup1_t *p, uint32_t pos, cstate_t *s)
  68 {
  69 #define _cop(c) ((c)&BAM_CIGAR_MASK)
  70 #define _cln(c) ((c)>>BAM_CIGAR_SHIFT)
  71
  72         bam1_t *b = p->b;
  73         bam1_core_t *c = &b->core;
  74         uint32_t *cigar = bam1_cigar(b);
  75         int k, is_head = 0;
  76         // determine the current CIGAR operation
  77 //      fprintf(stderr, "%s\tpos=%d\tend=%d\t(%d,%d,%d)\n", bam1_qname(b), pos, s->end, s->k, s->x, s->y);
  78         if (s->k == -1) { // never processed
  79                 is_head = 1;
  80                 if (c->n_cigar == 1) { // just one operation, save a loop
  81                         if (_cop(cigar[0]) == BAM_CMATCH) s->k = 0, s->x = c->pos, s->y = 0;
  82                 } else { // find the first match
  83                         for (k = 0, s->x = c->pos, s->y = 0; k < c->n_cigar; ++k) {
  84                                 int op = _cop(cigar[k]);
  85                                 int l = _cln(cigar[k]);
  86                                 if (op == BAM_CMATCH) break;
  87                                 else if (op == BAM_CDEL || op == BAM_CREF_SKIP) s->x += l;
  88                                 else if (op == BAM_CINS || op == BAM_CSOFT_CLIP) s->y += l;
  89                         }
  90                         assert(k < c->n_cigar);
  91                         s->k = k;
  92                 }
  93         } else { // the read has been processed before
  94                 int op, l = _cln(cigar[s->k]);
  95                 if (pos - s->x >= l) { // jump to the next operation
  96                         assert(s->k < c->n_cigar); // otherwise a bug: this function should not be called in this case
  97                         op = _cop(cigar[s->k+1]);
  98                         if (op == BAM_CMATCH || op == BAM_CDEL || op == BAM_CREF_SKIP) { // jump to the next without a loop
  99                                 if (_cop(cigar[s->k]) == BAM_CMATCH) s->y += l;
 100                                 s->x += l;
 101                                 ++s->k;
 102                         } else { // find the next M/D/N
 103                                 if (_cop(cigar[s->k]) == BAM_CMATCH) s->y += l;
 104                                 s->x += l;
 105                                 for (k = s->k + 1; k < c->n_cigar; ++k) {
 106                                         op = _cop(cigar[k]), l = _cln(cigar[k]);
 107                                         if (op == BAM_CMATCH || op == BAM_CDEL || op == BAM_CREF_SKIP) break;
 108                                         else if (op == BAM_CINS || op == BAM_CSOFT_CLIP) s->y += l;
 109                                 }
 110                                 s->k = k;
 111                         }
 112                         assert(s->k < c->n_cigar); // otherwise a bug
 113                 } // else, do nothing
 114         }
 115         { // collect pileup information
 116                 int op, l;
 117                 op = _cop(cigar[s->k]); l = _cln(cigar[s->k]);
 118                 p->is_del = p->indel = p->is_refskip = 0;
 119                 if (s->x + l - 1 == pos && s->k + 1 < c->n_cigar) { // peek the next operation
 120                         int op2 = _cop(cigar[s->k+1]);
 121                         int l2 = _cln(cigar[s->k+1]);
 122                         if (op2 == BAM_CDEL) p->indel = -(int)l2;
 123                         else if (op2 == BAM_CINS) p->indel = l2;
 124                         else if (op2 == BAM_CPAD && s->k + 2 < c->n_cigar) { // no working for adjacent padding
 125                                 int l3 = 0;
 126                                 for (k = s->k + 2; k < c->n_cigar; ++k) {
 127                                         op2 = _cop(cigar[k]); l2 = _cln(cigar[k]);
 128                                         if (op2 == BAM_CINS) l3 += l2;
 129                                         else if (op2 == BAM_CDEL || op2 == BAM_CMATCH || op2 == BAM_CREF_SKIP) break;
 130                                 }
 131                                 if (l3 > 0) p->indel = l3;
 132                         }
 133                 }
 134                 if (op == BAM_CMATCH) {
 135                         p->qpos = s->y + (pos - s->x);
 136                 } else if (op == BAM_CDEL || op == BAM_CREF_SKIP) {
 137                         p->is_del = 1; p->qpos = s->y; // FIXME: distinguish D and N!!!!!
 138                         p->is_refskip = (op == BAM_CREF_SKIP);
 139                 } // cannot be other operations; otherwise a bug
 140                 p->is_head = (pos == c->pos); p->is_tail = (pos == s->end);
 141         }
 142         return 1;
 143 }
 144
 145 /* --- END: Auxiliary functions */
 146
 147 /*******************
 148  * pileup iterator *
 149  *******************/
 150
 151 struct __bam_plp_t {
 152         mempool_t *mp;
 153         lbnode_t *head, *tail, *dummy;
 154         int32_t tid, pos, max_tid, max_pos;
 155         int is_eof, flag_mask, max_plp, error;
 156         bam_pileup1_t *plp;
 157         // for the "auto" interface only
 158         bam1_t *b;
 159         bam_plp_auto_f func;
 160         void *data;
 161 };
 162
 163 bam_plp_t bam_plp_init(bam_plp_auto_f func, void *data)
 164 {
 165         bam_plp_t iter;
 166         iter = calloc(1, sizeof(struct __bam_plp_t));
 167         iter->mp = mp_init();
 168         iter->head = iter->tail = mp_alloc(iter->mp);
 169         iter->dummy = mp_alloc(iter->mp);
 170         iter->max_tid = iter->max_pos = -1;
 171         iter->flag_mask = BAM_DEF_MASK;
 172         if (func) {
 173                 iter->func = func;
 174                 iter->data = data;
 175                 iter->b = bam_init1();
 176         }
 177         return iter;
 178 }
 179
 180 void bam_plp_destroy(bam_plp_t iter)
 181 {
 182         mp_free(iter->mp, iter->dummy);
 183         mp_free(iter->mp, iter->head);
 184         if (iter->mp->cnt != 0)
 185                 fprintf(stderr, "[bam_plp_destroy] memory leak: %d. Continue anyway.\n", iter->mp->cnt);
 186         mp_destroy(iter->mp);
 187         if (iter->b) bam_destroy1(iter->b);
 188         free(iter->plp);
 189         free(iter);
 190 }
 191
 192 const bam_pileup1_t *bam_plp_next(bam_plp_t iter, int *_tid, int *_pos, int *_n_plp)
 193 {
 194         if (iter->error) { *_n_plp = -1; return 0; }
 195         *_n_plp = 0;
 196         if (iter->is_eof && iter->head->next == 0) return 0;
 197         while (iter->is_eof || iter->max_tid > iter->tid || (iter->max_tid == iter->tid && iter->max_pos > iter->pos)) {
 198                 int n_plp = 0;
 199                 lbnode_t *p, *q;
 200                 // write iter->plp at iter->pos
 201                 iter->dummy->next = iter->head;
 202                 for (p = iter->head, q = iter->dummy; p->next; q = p, p = p->next) {
 203                         if (p->b.core.tid < iter->tid || (p->b.core.tid == iter->tid && p->end <= iter->pos)) { // then remove
 204                                 q->next = p->next; mp_free(iter->mp, p); p = q;
 205                         } else if (p->b.core.tid == iter->tid && p->beg <= iter->pos) { // here: p->end > pos; then add to pileup
 206                                 if (n_plp == iter->max_plp) { // then double the capacity
 207                                         iter->max_plp = iter->max_plp? iter->max_plp<<1 : 256;
 208                                         iter->plp = (bam_pileup1_t*)realloc(iter->plp, sizeof(bam_pileup1_t) * iter->max_plp);
 209                                 }
 210                                 iter->plp[n_plp].b = &p->b;
 211                                 if (resolve_cigar2(iter->plp + n_plp, iter->pos, &p->s)) ++n_plp; // actually always true...
 212                         }
 213                 }
 214                 iter->head = iter->dummy->next; // dummy->next may be changed
 215                 *_n_plp = n_plp; *_tid = iter->tid; *_pos = iter->pos;
 216                 // update iter->tid and iter->pos
 217                 if (iter->head->next) {
 218                         if (iter->tid > iter->head->b.core.tid) {
 219                                 fprintf(stderr, "[%s] unsorted input. Pileup aborts.\n", __func__);
 220                                 iter->error = 1;
 221                                 *_n_plp = -1;
 222                                 return 0;
 223                         }
 224                 }
 225                 if (iter->tid < iter->head->b.core.tid) { // come to a new reference sequence
 226                         iter->tid = iter->head->b.core.tid; iter->pos = iter->head->beg; // jump to the next reference
 227                 } else if (iter->pos < iter->head->beg) { // here: tid == head->b.core.tid
 228                         iter->pos = iter->head->beg; // jump to the next position
 229                 } else ++iter->pos; // scan contiguously
 230                 // return
 231                 if (n_plp) return iter->plp;
 232                 if (iter->is_eof && iter->head->next == 0) break;
 233         }
 234         return 0;
 235 }
 236
 237 int bam_plp_push(bam_plp_t iter, const bam1_t *b)
 238 {
 239         if (iter->error) return -1;
 240         if (b) {
 241                 if (b->core.tid < 0) return 0;
 242                 if (b->core.flag & iter->flag_mask) return 0;
 243                 bam_copy1(&iter->tail->b, b);
 244                 iter->tail->beg = b->core.pos; iter->tail->end = bam_calend(&b->core, bam1_cigar(b));
 245                 iter->tail->s = g_cstate_null; iter->tail->s.end = iter->tail->end - 1; // initialize cstate_t
 246                 if (b->core.tid < iter->max_tid) {
 247                         fprintf(stderr, "[bam_pileup_core] the input is not sorted (chromosomes out of order)\n");
 248                         iter->error = 1;
 249                         return -1;
 250                 }
 251                 if ((b->core.tid == iter->max_tid) && (iter->tail->beg < iter->max_pos)) {
 252                         fprintf(stderr, "[bam_pileup_core] the input is not sorted (reads out of order)\n");
 253                         iter->error = 1;
 254                         return -1;
 255                 }
 256                 iter->max_tid = b->core.tid; iter->max_pos = iter->tail->beg;
 257                 if (iter->tail->end > iter->pos || iter->tail->b.core.tid > iter->tid) {
 258                         iter->tail->next = mp_alloc(iter->mp);
 259                         iter->tail = iter->tail->next;
 260                 }
 261         } else iter->is_eof = 1;
 262         return 0;
 263 }
 264
 265 const bam_pileup1_t *bam_plp_auto(bam_plp_t iter, int *_tid, int *_pos, int *_n_plp)
 266 {
 267         const bam_pileup1_t *plp;
 268         if (iter->func == 0 || iter->error) { *_n_plp = -1; return 0; }
 269         if ((plp = bam_plp_next(iter, _tid, _pos, _n_plp)) != 0) return plp;
 270         else {
 271                 *_n_plp = 0;
 272                 if (iter->is_eof) return 0;
 273                 while (iter->func(iter->data, iter->b) >= 0) {
 274                         if (bam_plp_push(iter, iter->b) < 0) {
 275                                 *_n_plp = -1;
 276                                 return 0;
 277                         }
 278                         if ((plp = bam_plp_next(iter, _tid, _pos, _n_plp)) != 0) return plp;
 279                 }
 280                 bam_plp_push(iter, 0);
 281                 if ((plp = bam_plp_next(iter, _tid, _pos, _n_plp)) != 0) return plp;
 282                 return 0;
 283         }
 284 }
 285
 286 void bam_plp_reset(bam_plp_t iter)
 287 {
 288         lbnode_t *p, *q;
 289         iter->max_tid = iter->max_pos = -1;
 290         iter->tid = iter->pos = 0;
 291         iter->is_eof = 0;
 292         for (p = iter->head; p->next;) {
 293                 q = p->next;
 294                 mp_free(iter->mp, p);
 295                 p = q;
 296         }
 297         iter->head = iter->tail;
 298 }
 299
 300 void bam_plp_set_mask(bam_plp_t iter, int mask)
 301 {
 302         iter->flag_mask = mask < 0? BAM_DEF_MASK : (BAM_FUNMAP | mask);
 303 }
 304
 305 /*****************
 306  * callback APIs *
 307  *****************/
 308
 309 int bam_pileup_file(bamFile fp, int mask, bam_pileup_f func, void *func_data)
 310 {
 311         bam_plbuf_t *buf;
 312         int ret;
 313         bam1_t *b;
 314         b = bam_init1();
 315         buf = bam_plbuf_init(func, func_data);
 316         bam_plbuf_set_mask(buf, mask);
 317         while ((ret = bam_read1(fp, b)) >= 0)
 318                 bam_plbuf_push(b, buf);
 319         bam_plbuf_push(0, buf);
 320         bam_plbuf_destroy(buf);
 321         bam_destroy1(b);
 322         return 0;
 323 }
 324
 325 void bam_plbuf_set_mask(bam_plbuf_t *buf, int mask)
 326 {
 327         bam_plp_set_mask(buf->iter, mask);
 328 }
 329
 330 void bam_plbuf_reset(bam_plbuf_t *buf)
 331 {
 332         bam_plp_reset(buf->iter);
 333 }
 334
 335 bam_plbuf_t *bam_plbuf_init(bam_pileup_f func, void *data)
 336 {
 337         bam_plbuf_t *buf;
 338         buf = calloc(1, sizeof(bam_plbuf_t));
 339         buf->iter = bam_plp_init(0, 0);
 340         buf->func = func;
 341         buf->data = data;
 342         return buf;
 343 }
 344
 345 void bam_plbuf_destroy(bam_plbuf_t *buf)
 346 {
 347         bam_plp_destroy(buf->iter);
 348         free(buf);
 349 }
 350
 351 int bam_plbuf_push(const bam1_t *b, bam_plbuf_t *buf)
 352 {
 353         int ret, n_plp, tid, pos;
 354         const bam_pileup1_t *plp;
 355         ret = bam_plp_push(buf->iter, b);
 356         if (ret < 0) return ret;
 357         while ((plp = bam_plp_next(buf->iter, &tid, &pos, &n_plp)) != 0)
 358                 buf->func(tid, pos, n_plp, plp, buf->data);
 359         return 0;
 360 }
 361
 362 /***********
 363  * mpileup *
 364  ***********/
 365
 366 struct __bam_mplp_t {
 367         int n;
 368         uint64_t min, *pos;
 369         bam_plp_t *iter;
 370         int *n_plp;
 371         const bam_pileup1_t **plp;
 372 };
 373
 374 bam_mplp_t bam_mplp_init(int n, bam_plp_auto_f func, void **data)
 375 {
 376         int i;
 377         bam_mplp_t iter;
 378         iter = calloc(1, sizeof(struct __bam_mplp_t));
 379         iter->pos = calloc(n, 8);
 380         iter->n_plp = calloc(n, sizeof(int));
 381         iter->plp = calloc(n, sizeof(void*));
 382         iter->iter = calloc(n, sizeof(void*));
 383         iter->n = n;
 384         iter->min = (uint64_t)-1;
 385         for (i = 0; i < n; ++i) {
 386                 iter->iter[i] = bam_plp_init(func, data[i]);
 387                 iter->pos[i] = iter->min;
 388         }
 389         return iter;
 390 }
 391
 392 void bam_mplp_destroy(bam_mplp_t iter)
 393 {
 394         int i;
 395         for (i = 0; i < iter->n; ++i) bam_plp_destroy(iter->iter[i]);
 396         free(iter->iter); free(iter->pos); free(iter->n_plp); free(iter->plp);
 397         free(iter);
 398 }
 399
 400 int bam_mplp_auto(bam_mplp_t iter, int *_tid, int *_pos, int *n_plp, const bam_pileup1_t **plp)
 401 {
 402         int i, ret = 0;
 403         uint64_t new_min = (uint64_t)-1;
 404         for (i = 0; i < iter->n; ++i) {
 405                 if (iter->pos[i] == iter->min) {
 406                         int tid, pos;
 407                         iter->plp[i] = bam_plp_auto(iter->iter[i], &tid, &pos, &iter->n_plp[i]);
 408                         iter->pos[i] = (uint64_t)tid<<32 | pos;
 409                 }
 410                 if (iter->plp[i] && iter->pos[i] < new_min) new_min = iter->pos[i];
 411         }
 412         iter->min = new_min;
 413         if (new_min == (uint64_t)-1) return 0;
 414         *_tid = new_min>>32; *_pos = (uint32_t)new_min;
 415         for (i = 0; i < iter->n; ++i) {
 416                 if (iter->pos[i] == iter->min) { // FIXME: valgrind reports "uninitialised value(s) at this line"
 417                         n_plp[i] = iter->n_plp[i], plp[i] = iter->plp[i];
 418                         ++ret;
 419                 } else n_plp[i] = 0, plp[i] = 0;
 420         }
 421         return ret;
 422 }