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