X-Git-Url: http://woldlab.caltech.edu/gitweb/?p=samtools.git;a=blobdiff_plain;f=bam_sort.c;h=76ab793196ca76fc5a63d619e230cf0fb5f59f1e;hp=9884f3d6273f680709ee0e44532a8371413fddf4;hb=016b50ab60e879a0b8f81cb76ce11ea360a03d4a;hpb=317f5e8dd22cc9e1e5e05fbcaeb3b9aca7447351 diff --git a/bam_sort.c b/bam_sort.c index 9884f3d..76ab793 100644 --- a/bam_sort.c +++ b/bam_sort.c @@ -1,6 +1,7 @@ #include #include #include +#include #include #include #include @@ -59,6 +60,9 @@ static void swap_header_text(bam_header_t *h1, bam_header_t *h2) temps = h1->text, h1->text = h2->text, h2->text = temps; } +#define MERGE_RG 1 +#define MERGE_UNCOMP 2 + /*! @abstract Merge multiple sorted BAM. @param is_by_qname whether to sort by query name @@ -71,7 +75,8 @@ static void swap_header_text(bam_header_t *h1, bam_header_t *h2) @discussion Padding information may NOT correctly maintained. This function is NOT thread safe. */ -void bam_merge_core(int by_qname, const char *out, const char *headers, int n, char * const *fn, int add_RG) +int bam_merge_core(int by_qname, const char *out, const char *headers, int n, char * const *fn, + int flag, const char *reg) { bamFile fpout, *fp; heap1_t *heap; @@ -80,22 +85,25 @@ void bam_merge_core(int by_qname, const char *out, const char *headers, int n, c int i, j, *RG_len = 0; uint64_t idx = 0; char **RG = 0; + bam_iter_t *iter = 0; if (headers) { tamFile fpheaders = sam_open(headers); if (fpheaders == 0) { - fprintf(stderr, "[bam_merge_core] Cannot open file `%s'. Continue anyway.\n", headers); - } else { - hheaders = sam_header_read(fpheaders); - sam_close(fpheaders); + const char *message = strerror(errno); + fprintf(stderr, "[bam_merge_core] cannot open '%s': %s\n", headers, message); + return -1; } + hheaders = sam_header_read(fpheaders); + sam_close(fpheaders); } g_is_by_qname = by_qname; fp = (bamFile*)calloc(n, sizeof(bamFile)); heap = (heap1_t*)calloc(n, sizeof(heap1_t)); + iter = (bam_iter_t*)calloc(n, sizeof(bam_iter_t)); // prepare RG tag - if (add_RG) { + if (flag & MERGE_RG) { RG = (char**)calloc(n, sizeof(void*)); RG_len = (int*)calloc(n, sizeof(int)); for (i = 0; i != n; ++i) { @@ -111,7 +119,6 @@ void bam_merge_core(int by_qname, const char *out, const char *headers, int n, c } // read the first for (i = 0; i != n; ++i) { - heap1_t *h; bam_header_t *hin; fp[i] = bam_open(fn[i], "r"); if (fp[i] == 0) { @@ -120,7 +127,7 @@ void bam_merge_core(int by_qname, const char *out, const char *headers, int n, c for (j = 0; j < i; ++j) bam_close(fp[j]); free(fp); free(heap); // FIXME: possible memory leak - return; + return -1; } hin = bam_header_read(fp[i]); if (i == 0) { // the first SAM @@ -130,11 +137,15 @@ void bam_merge_core(int by_qname, const char *out, const char *headers, int n, c // check that they are consistent with the existing binary list // of reference information. if (hheaders->n_targets > 0) { - if (hout->n_targets != hheaders->n_targets) + if (hout->n_targets != hheaders->n_targets) { fprintf(stderr, "[bam_merge_core] number of @SQ headers in `%s' differs from number of target sequences", headers); + if (!reg) return -1; + } for (j = 0; j < hout->n_targets; ++j) - if (strcmp(hout->target_name[j], hheaders->target_name[j]) != 0) + if (strcmp(hout->target_name[j], hheaders->target_name[j]) != 0) { fprintf(stderr, "[bam_merge_core] @SQ header '%s' in '%s' differs from target sequence", hheaders->target_name[j], headers); + if (!reg) return -1; + } } swap_header_text(hout, hheaders); bam_header_destroy(hheaders); @@ -142,39 +153,62 @@ void bam_merge_core(int by_qname, const char *out, const char *headers, int n, c } } else { // validate multiple baf if (hout->n_targets != hin->n_targets) { - fprintf(stderr, "[bam_merge_core] file '%s' has different number of target sequences. Abort!\n", fn[i]); - exit(1); - } - for (j = 0; j < hout->n_targets; ++j) { - if (strcmp(hout->target_name[j], hin->target_name[j])) { - fprintf(stderr, "[bam_merge_core] different target sequence name: '%s' != '%s' in file '%s'. Abort!\n", - hout->target_name[j], hin->target_name[j], fn[i]); - exit(1); + fprintf(stderr, "[bam_merge_core] file '%s' has different number of target sequences. Continue anyway!\n", fn[i]); + } else { + for (j = 0; j < hout->n_targets; ++j) { + if (strcmp(hout->target_name[j], hin->target_name[j])) { + fprintf(stderr, "[bam_merge_core] different target sequence name: '%s' != '%s' in file '%s'. Continue anyway!\n", + hout->target_name[j], hin->target_name[j], fn[i]); + } } } bam_header_destroy(hin); } - h = heap + i; + } + + if (reg) { + int tid, beg, end; + if (bam_parse_region(hout, reg, &tid, &beg, &end) < 0) { + fprintf(stderr, "[%s] Malformated region string or undefined reference name\n", __func__); + return -1; + } + for (i = 0; i < n; ++i) { + bam_index_t *idx; + idx = bam_index_load(fn[i]); + iter[i] = bam_iter_query(idx, tid, beg, end); + bam_index_destroy(idx); + } + } + + for (i = 0; i < n; ++i) { + heap1_t *h = heap + i; h->i = i; h->b = (bam1_t*)calloc(1, sizeof(bam1_t)); - if (bam_read1(fp[i], h->b) >= 0) { + if (bam_iter_read(fp[i], iter[i], h->b) >= 0) { h->pos = ((uint64_t)h->b->core.tid<<32) | (uint32_t)h->b->core.pos<<1 | bam1_strand(h->b); h->idx = idx++; } else h->pos = HEAP_EMPTY; } - fpout = strcmp(out, "-")? bam_open(out, "w") : bam_dopen(fileno(stdout), "w"); - assert(fpout); + if (flag & MERGE_UNCOMP) { + fpout = strcmp(out, "-")? bam_open(out, "wu") : bam_dopen(fileno(stdout), "wu"); + } else { + fpout = strcmp(out, "-")? bam_open(out, "w") : bam_dopen(fileno(stdout), "w"); + } + if (fpout == 0) { + fprintf(stderr, "[%s] fail to create the output file.\n", __func__); + return -1; + } bam_header_write(fpout, hout); bam_header_destroy(hout); ks_heapmake(heap, n, heap); while (heap->pos != HEAP_EMPTY) { bam1_t *b = heap->b; - if (add_RG && bam_aux_get(b, "RG") == 0) + if ((flag & MERGE_RG) && bam_aux_get(b, "RG") == 0) bam_aux_append(b, "RG", 'Z', RG_len[heap->i] + 1, (uint8_t*)RG[heap->i]); bam_write1_core(fpout, &b->core, b->data_len, b->data); - if ((j = bam_read1(fp[heap->i], b)) >= 0) { + if ((j = bam_iter_read(fp[heap->i], iter[heap->i], b)) >= 0) { heap->pos = ((uint64_t)b->core.tid<<32) | (uint32_t)b->core.pos<<1 | bam1_strand(b); heap->idx = idx++; } else if (j == -1) { @@ -185,24 +219,31 @@ void bam_merge_core(int by_qname, const char *out, const char *headers, int n, c ks_heapadjust(heap, 0, n, heap); } - if (add_RG) { + if (flag & MERGE_RG) { for (i = 0; i != n; ++i) free(RG[i]); free(RG); free(RG_len); } - for (i = 0; i != n; ++i) bam_close(fp[i]); + for (i = 0; i != n; ++i) { + bam_iter_destroy(iter[i]); + bam_close(fp[i]); + } bam_close(fpout); - free(fp); free(heap); + free(fp); free(heap); free(iter); + return 0; } + int bam_merge(int argc, char *argv[]) { - int c, is_by_qname = 0, add_RG = 0; - char *fn_headers = NULL; + int c, is_by_qname = 0, flag = 0, ret = 0; + char *fn_headers = NULL, *reg = 0; - while ((c = getopt(argc, argv, "h:nr")) >= 0) { + while ((c = getopt(argc, argv, "h:nruR:")) >= 0) { switch (c) { - case 'r': add_RG = 1; break; + case 'r': flag |= MERGE_RG; break; case 'h': fn_headers = strdup(optarg); break; case 'n': is_by_qname = 1; break; + case 'u': flag |= MERGE_UNCOMP; break; + case 'R': reg = strdup(optarg); break; } } if (optind + 2 >= argc) { @@ -210,15 +251,18 @@ int bam_merge(int argc, char *argv[]) fprintf(stderr, "Usage: samtools merge [-nr] [-h inh.sam] [...]\n\n"); fprintf(stderr, "Options: -n sort by read names\n"); fprintf(stderr, " -r attach RG tag (inferred from file names)\n"); + fprintf(stderr, " -u uncompressed BAM output\n"); + fprintf(stderr, " -R STR merge file in the specified region STR [all]\n"); fprintf(stderr, " -h FILE copy the header in FILE to [in1.bam]\n\n"); fprintf(stderr, "Note: Samtools' merge does not reconstruct the @RG dictionary in the header. Users\n"); fprintf(stderr, " must provide the correct header with -h, or uses Picard which properly maintains\n"); fprintf(stderr, " the header dictionary in merging.\n\n"); return 1; } - bam_merge_core(is_by_qname, argv[optind], fn_headers, argc - optind - 1, argv + optind + 1, add_RG); + if (bam_merge_core(is_by_qname, argv[optind], fn_headers, argc - optind - 1, argv + optind + 1, flag, reg) < 0) ret = 1; + free(reg); free(fn_headers); - return 0; + return ret; } typedef bam1_t *bam1_p; @@ -294,7 +338,7 @@ void bam_sort_core_ext(int is_by_qname, const char *fn, const char *prefix, size mem += ret; ++k; if (mem >= max_mem) { - sort_blocks(n++, k, buf, prefix, header, is_stdout); + sort_blocks(n++, k, buf, prefix, header, 0); mem = 0; k = 0; } } @@ -304,7 +348,7 @@ void bam_sort_core_ext(int is_by_qname, const char *fn, const char *prefix, size else { // then merge char **fns, *fnout; fprintf(stderr, "[bam_sort_core] merging from %d files...\n", n+1); - sort_blocks(n++, k, buf, prefix, header, is_stdout); + sort_blocks(n++, k, buf, prefix, header, 0); fnout = (char*)calloc(strlen(prefix) + 20, 1); if (is_stdout) sprintf(fnout, "-"); else sprintf(fnout, "%s.bam", prefix); @@ -313,7 +357,7 @@ void bam_sort_core_ext(int is_by_qname, const char *fn, const char *prefix, size fns[i] = (char*)calloc(strlen(prefix) + 20, 1); sprintf(fns[i], "%s.%.4d.bam", prefix, i); } - bam_merge_core(is_by_qname, fnout, 0, n, fns, 0); + bam_merge_core(is_by_qname, fnout, 0, n, fns, 0, 0); free(fnout); for (i = 0; i < n; ++i) { unlink(fns[i]);