X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=bam_import.c;h=427140db9cbe9e3c80d173b0d69e31b1774c3037;hb=d0e30eec1158752010659982342a611fc91ae8e3;hp=9daa29f42a1676077b29f3f63ab7a7027896dbad;hpb=f1222861f3824c28d35cd143f5565ca09d80f056;p=samtools.git diff --git a/bam_import.c b/bam_import.c index 9daa29f..427140d 100644 --- a/bam_import.c +++ b/bam_import.c @@ -5,7 +5,12 @@ #include #include #include +#ifdef _WIN32 +#include +#endif +#include "kstring.h" #include "bam.h" +#include "sam_header.h" #include "kseq.h" #include "khash.h" @@ -35,6 +40,25 @@ unsigned char bam_nt16_table[256] = { 15,15,15,15, 15,15,15,15, 15,15,15,15, 15,15,15,15 }; +unsigned short bam_char2flag_table[256] = { + 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0, + 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0, + 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0, + 0,BAM_FREAD1,BAM_FREAD2,0, 0,0,0,0, 0,0,0,0, 0,0,0,0, + 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0, + BAM_FPROPER_PAIR,0,BAM_FMREVERSE,0, 0,BAM_FMUNMAP,0,0, 0,0,0,0, 0,0,0,0, + 0,0,0,0, BAM_FDUP,0,BAM_FQCFAIL,0, 0,0,0,0, 0,0,0,0, + BAM_FPAIRED,0,BAM_FREVERSE,BAM_FSECONDARY, 0,BAM_FUNMAP,0,0, 0,0,0,0, 0,0,0,0, + 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0, + 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0, + 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0, + 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0, + 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0, + 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0, + 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0, + 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0 +}; + char *bam_nt16_rev_table = "=ACMGRSVTWYHKDBN"; struct __tamFile_t { @@ -42,6 +66,7 @@ struct __tamFile_t { kstream_t *ks; kstring_t *str; uint64_t n_lines; + int is_first; }; char **__bam_get_lines(const char *fn, int *_n) // for bam_plcmd.c only @@ -97,9 +122,10 @@ bam_header_t *sam_header_read2(const char *fn) kstring_t *str; kh_ref_t *hash; khiter_t k; - hash = kh_init(ref); + if (fn == 0) return 0; fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r"); - assert(fp); + if (fp == 0) return 0; + hash = kh_init(ref); ks = ks_init(fp); str = (kstring_t*)calloc(1, sizeof(kstring_t)); while (ks_getuntil(ks, 0, str, &dret) > 0) { @@ -144,28 +170,76 @@ static inline void append_text(bam_header_t *header, kstring_t *str) header->l_text += str->l + 1; header->text[header->l_text] = 0; } -int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b) + +int sam_header_parse_rg(bam_header_t *h) { - int ret, doff, doff0, dret, z = 0; - bam1_core_t *c = &b->core; - kstring_t *str = fp->str; - kstream_t *ks = fp->ks; + if (h->dict == 0) h->dict = sam_header_parse2(h->text); + if (h->rg2lib) h->rg2lib = sam_header2tbl(h->dict, "RG", "ID", "LB"); + return sam_tbl_size(h->rg2lib); +} +int sam_header_parse(bam_header_t *h) +{ + void *tbl; + char **tmp; + int i; + free(h->target_len); free(h->target_name); + h->n_targets = 0; h->target_len = 0; h->target_name = 0; + if (h->l_text < 3) return 0; + if (h->dict == 0) h->dict = sam_header_parse2(h->text); + tbl = sam_header2tbl(h->dict, "SQ", "SN", "LN"); + h->n_targets = sam_tbl_size(tbl); + if (h->n_targets == 0) { + sam_tbl_destroy(tbl); + return 0; + } + h->target_len = (uint32_t*)calloc(h->n_targets, 4); + h->target_name = (char**)calloc(h->n_targets, sizeof(void*)); + tmp = (char**)calloc(h->n_targets, sizeof(void*)); + sam_tbl_pair(tbl, h->target_name, tmp); + for (i = 0; i < h->n_targets; ++i) + h->target_len[i] = atoi(tmp[i]); + free(tmp); + sam_tbl_destroy(tbl); + return h->n_targets; +} + +bam_header_t *sam_header_read(tamFile fp) +{ + int ret, dret; + bam_header_t *header = bam_header_init(); + kstring_t *str = fp->str; while ((ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret)) >= 0 && str->s[0] == '@') { // skip header - z += str->l + 1; str->s[str->l] = dret; // note that str->s is NOT null terminated!! append_text(header, str); if (dret != '\n') { ret = ks_getuntil(fp->ks, '\n', str, &dret); - z += str->l + 1; str->s[str->l] = '\n'; // NOT null terminated!! append_text(header, str); } ++fp->n_lines; } - while (ret == 0) { // special consideration for "\r\n" and empty lines - ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret); - if (ret >= 0) z += str->l + 1; + sam_header_parse(header); + bam_init_header_hash(header); + fp->is_first = 1; + return header; +} + +int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b) +{ + int ret, doff, doff0, dret, z = 0; + bam1_core_t *c = &b->core; + kstring_t *str = fp->str; + kstream_t *ks = fp->ks; + + if (fp->is_first) { + fp->is_first = 0; + ret = str->l; + } else { + do { // special consideration for empty lines + ret = ks_getuntil(fp->ks, KS_SEP_TAB, str, &dret); + if (ret >= 0) z += str->l + 1; + } while (ret == 0); } if (ret < 0) return -1; ++fp->n_lines; @@ -176,9 +250,26 @@ int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b) memcpy(alloc_data(b, doff + c->l_qname) + doff, str->s, c->l_qname); doff += c->l_qname; } - { // flag, tid, pos, qual - ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->flag = atoi(str->s); + { // flag + long flag; + char *s; + ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; + flag = strtol((char*)str->s, &s, 0); + if (*s) { // not the end of the string + flag = 0; + for (s = str->s; *s; ++s) + flag |= bam_char2flag_table[(int)*s]; + } + c->flag = flag; + } + { // tid, pos, qual ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->tid = bam_get_tid(header, str->s); + if (c->tid < 0 && strcmp(str->s, "*")) { + if (header->n_targets == 0) { + fprintf(stderr, "[sam_read1] missing header? Abort!\n"); + exit(1); + } else fprintf(stderr, "[sam_read1] reference '%s' is recognized as '*'.\n", str->s); + } ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->pos = isdigit(str->s[0])? atoi(str->s) - 1 : -1; ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; c->qual = isdigit(str->s[0])? atoi(str->s) : 0; if (ret < 0) return -2; @@ -199,7 +290,7 @@ int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b) for (i = 0, s = str->s; i != c->n_cigar; ++i) { x = strtol(s, &t, 10); op = toupper(*t); - if (op == 'M') op = BAM_CMATCH; + if (op == 'M' || op == '=' || op == 'X') op = BAM_CMATCH; else if (op == 'I') op = BAM_CINS; else if (op == 'D') op = BAM_CDEL; else if (op == 'N') op = BAM_CREF_SKIP; @@ -213,7 +304,13 @@ int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b) if (*s) parse_error(fp->n_lines, "unmatched CIGAR operation"); c->bin = bam_reg2bin(c->pos, bam_calend(c, bam1_cigar(b))); doff += c->n_cigar * 4; - } else c->bin = bam_reg2bin(c->pos, c->pos + 1); + } else { + if (!(c->flag&BAM_FUNMAP)) { + fprintf(stderr, "Parse warning at line %lld: mapped sequence without CIGAR\n", (long long)fp->n_lines); + c->flag |= BAM_FUNMAP; + } + c->bin = bam_reg2bin(c->pos, c->pos + 1); + } } { // mtid, mpos, isize ret = ks_getuntil(ks, KS_SEP_TAB, str, &dret); z += str->l + 1; @@ -226,16 +323,18 @@ int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b) } { // seq and qual int i; - uint8_t *p; + uint8_t *p = 0; if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -5; // seq z += str->l + 1; - c->l_qseq = strlen(str->s); - if (c->n_cigar && c->l_qseq != (int32_t)bam_cigar2qlen(c, bam1_cigar(b))) - parse_error(fp->n_lines, "CIGAR and sequence length are inconsistent"); - p = (uint8_t*)alloc_data(b, doff + c->l_qseq + (c->l_qseq+1)/2) + doff; - bzero(p, (c->l_qseq+1)/2); - for (i = 0; i < c->l_qseq; ++i) - p[i/2] |= bam_nt16_table[(int)str->s[i]] << 4*(1-i%2); + if (strcmp(str->s, "*")) { + c->l_qseq = strlen(str->s); + if (c->n_cigar && c->l_qseq != (int32_t)bam_cigar2qlen(c, bam1_cigar(b))) + parse_error(fp->n_lines, "CIGAR and sequence length are inconsistent"); + p = (uint8_t*)alloc_data(b, doff + c->l_qseq + (c->l_qseq+1)/2) + doff; + memset(p, 0, (c->l_qseq+1)/2); + for (i = 0; i < c->l_qseq; ++i) + p[i/2] |= bam_nt16_table[(int)str->s[i]] << 4*(1-i%2); + } else c->l_qseq = 0; if (ks_getuntil(ks, KS_SEP_TAB, str, &dret) < 0) return -6; // qual z += str->l + 1; if (strcmp(str->s, "*") && c->l_qseq != strlen(str->s)) @@ -331,11 +430,12 @@ int sam_read1(tamFile fp, bam_header_t *header, bam1_t *b) tamFile sam_open(const char *fn) { tamFile fp; + gzFile gzfp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "rb") : gzopen(fn, "rb"); + if (gzfp == 0) return 0; fp = (tamFile)calloc(1, sizeof(struct __tamFile_t)); fp->str = (kstring_t*)calloc(1, sizeof(kstring_t)); - fp->fp = (strcmp(fn, "-") == 0)? gzdopen(fileno(stdin), "r") : gzopen(fn, "r"); + fp->fp = gzfp; fp->ks = ks_init(fp->fp); - fp->n_lines = 0; return fp; }