#include "bam.h"
#include "bam_endian.h"
#include "kstring.h"
+#include "sam_header.h"
-int bam_is_be = 0;
+int bam_is_be = 0, bam_verbose = 2, bam_no_B = 0;
char *bam_flag2char_table = "pPuUrR12sfd\0\0\0\0\0";
/**************************
* CIGAR related routines *
**************************/
-int bam_segreg(int32_t pos, const bam1_core_t *c, const uint32_t *cigar, bam_segreg_t *reg)
-{
- unsigned k;
- int32_t x = c->pos, y = 0;
- int state = 0;
- for (k = 0; k < c->n_cigar; ++k) {
- int op = cigar[k] & BAM_CIGAR_MASK; // operation
- int l = cigar[k] >> BAM_CIGAR_SHIFT; // length
- if (state == 0 && (op == BAM_CMATCH || op == BAM_CDEL || op == BAM_CINS) && x + l > pos) {
- reg->tbeg = x; reg->qbeg = y; reg->cbeg = k;
- state = 1;
- }
- if (op == BAM_CMATCH) { x += l; y += l; }
- else if (op == BAM_CDEL || op == BAM_CREF_SKIP) x += l;
- else if (op == BAM_CINS || op == BAM_CSOFT_CLIP) y += l;
- if (state == 1 && (op == BAM_CSOFT_CLIP || op == BAM_CHARD_CLIP || op == BAM_CREF_SKIP || k == c->n_cigar - 1)) {
- reg->tend = x; reg->qend = y; reg->cend = k;
- }
- }
- return state? 0 : -1;
-}
-
uint32_t bam_calend(const bam1_core_t *c, const uint32_t *cigar)
{
- uint32_t k, end;
- end = c->pos;
+ int k, end = c->pos;
for (k = 0; k < c->n_cigar; ++k) {
- int op = cigar[k] & BAM_CIGAR_MASK;
- if (op == BAM_CMATCH || op == BAM_CDEL || op == BAM_CREF_SKIP)
- end += cigar[k] >> BAM_CIGAR_SHIFT;
+ int op = bam_cigar_op(cigar[k]);
+ int len = bam_cigar_oplen(cigar[k]);
+ if (op == BAM_CBACK) { // move backward
+ int l, u, v;
+ if (k == c->n_cigar - 1) break; // skip trailing 'B'
+ for (l = k - 1, u = v = 0; l >= 0; --l) {
+ int op1 = bam_cigar_op(cigar[l]);
+ int len1 = bam_cigar_oplen(cigar[l]);
+ if (bam_cigar_type(op1)&1) { // consume query
+ if (u + len1 >= len) { // stop
+ if (bam_cigar_type(op1)&2) v += len - u;
+ break;
+ } else u += len1;
+ }
+ if (bam_cigar_type(op1)&2) v += len1;
+ }
+ end = l < 0? c->pos : end - v;
+ } else if (bam_cigar_type(op)&2) end += bam_cigar_oplen(cigar[k]);
}
return end;
}
{
uint32_t k;
int32_t l = 0;
- for (k = 0; k < c->n_cigar; ++k) {
- int op = cigar[k] & BAM_CIGAR_MASK;
- if (op == BAM_CMATCH || op == BAM_CINS || op == BAM_CSOFT_CLIP)
- l += cigar[k] >> BAM_CIGAR_SHIFT;
- }
+ for (k = 0; k < c->n_cigar; ++k)
+ if (bam_cigar_type(bam_cigar_op(cigar[k]))&1)
+ l += bam_cigar_oplen(cigar[k]);
return l;
}
free(header->target_len);
}
free(header->text);
-#ifndef BAM_NO_HASH
- if (header->rg2lib) bam_strmap_destroy(header->rg2lib);
+ if (header->dict) sam_header_free(header->dict);
+ if (header->rg2lib) sam_tbl_destroy(header->rg2lib);
bam_destroy_header_hash(header);
-#endif
free(header);
}
{
bam_header_t *header;
char buf[4];
+ int magic_len;
int32_t i = 1, name_len;
// check EOF
i = bgzf_check_EOF(fp);
// with ESPIPE. Suppress the error message in this case.
if (errno != ESPIPE) perror("[bam_header_read] bgzf_check_EOF");
}
- else if (i == 0) fprintf(stderr, "[bam_header_read] EOF marker is absent.\n");
+ else if (i == 0) fprintf(stderr, "[bam_header_read] EOF marker is absent. The input is probably truncated.\n");
// read "BAM1"
- if (bam_read(fp, buf, 4) != 4) return 0;
- if (strncmp(buf, "BAM\001", 4)) {
- fprintf(stderr, "[bam_header_read] wrong header\n");
+ magic_len = bam_read(fp, buf, 4);
+ if (magic_len != 4 || strncmp(buf, "BAM\001", 4) != 0) {
+ fprintf(stderr, "[bam_header_read] invalid BAM binary header (this is not a BAM file).\n");
return 0;
}
header = bam_header_init();
bam_write(fp, &x, 4);
} else bam_write(fp, &header->target_len[i], 4);
}
+ bgzf_flush(fp);
return 0;
}
else if (type == 'I' || type == 'F') { bam_swap_endian_4p(s); s += 4; }
else if (type == 'D') { bam_swap_endian_8p(s); s += 8; }
else if (type == 'Z' || type == 'H') { while (*s) ++s; ++s; }
+ else if (type == 'B') {
+ int32_t n, Bsize = bam_aux_type2size(*s);
+ memcpy(&n, s + 1, 4);
+ if (1 == Bsize) {
+ } else if (2 == Bsize) {
+ for (i = 0; i < n; i += 2)
+ bam_swap_endian_2p(s + 5 + i);
+ } else if (4 == Bsize) {
+ for (i = 0; i < n; i += 4)
+ bam_swap_endian_4p(s + 5 + i);
+ }
+ bam_swap_endian_4p(s+1);
+ }
}
}
if (bam_read(fp, b->data, b->data_len) != b->data_len) return -4;
b->l_aux = b->data_len - c->n_cigar * 4 - c->l_qname - c->l_qseq - (c->l_qseq+1)/2;
if (bam_is_be) swap_endian_data(c, b->data_len, b->data);
+ if (bam_no_B) bam_remove_B(b);
return 4 + block_len;
}
x[5] = c->mtid;
x[6] = c->mpos;
x[7] = c->isize;
+ bgzf_flush_try(fp, 4 + block_len);
if (bam_is_be) {
for (i = 0; i < 8; ++i) bam_swap_endian_4p(x + i);
y = block_len;
kstring_t str;
str.l = str.m = 0; str.s = 0;
- ksprintf(&str, "%s\t", bam1_qname(b));
- if (of == BAM_OFDEC) ksprintf(&str, "%d\t", c->flag);
+ kputsn(bam1_qname(b), c->l_qname-1, &str); kputc('\t', &str);
+ if (of == BAM_OFDEC) { kputw(c->flag, &str); kputc('\t', &str); }
else if (of == BAM_OFHEX) ksprintf(&str, "0x%x\t", c->flag);
else { // BAM_OFSTR
for (i = 0; i < 16; ++i)
kputc(bam_flag2char_table[i], &str);
kputc('\t', &str);
}
- if (c->tid < 0) kputs("*\t", &str);
- else ksprintf(&str, "%s\t", header->target_name[c->tid]);
- ksprintf(&str, "%d\t%d\t", c->pos + 1, c->qual);
+ if (c->tid < 0) kputsn("*\t", 2, &str);
+ else {
+ if (header) kputs(header->target_name[c->tid] , &str);
+ else kputw(c->tid, &str);
+ kputc('\t', &str);
+ }
+ kputw(c->pos + 1, &str); kputc('\t', &str); kputw(c->qual, &str); kputc('\t', &str);
if (c->n_cigar == 0) kputc('*', &str);
else {
- for (i = 0; i < c->n_cigar; ++i)
- ksprintf(&str, "%d%c", bam1_cigar(b)[i]>>BAM_CIGAR_SHIFT, "MIDNSHP"[bam1_cigar(b)[i]&BAM_CIGAR_MASK]);
+ uint32_t *cigar = bam1_cigar(b);
+ for (i = 0; i < c->n_cigar; ++i) {
+ kputw(bam1_cigar(b)[i]>>BAM_CIGAR_SHIFT, &str);
+ kputc(bam_cigar_opchr(cigar[i]), &str);
+ }
}
kputc('\t', &str);
- if (c->mtid < 0) kputs("*\t", &str);
- else if (c->mtid == c->tid) kputs("=\t", &str);
- else ksprintf(&str, "%s\t", header->target_name[c->mtid]);
- ksprintf(&str, "%d\t%d\t", c->mpos + 1, c->isize);
+ if (c->mtid < 0) kputsn("*\t", 2, &str);
+ else if (c->mtid == c->tid) kputsn("=\t", 2, &str);
+ else {
+ if (header) kputs(header->target_name[c->mtid], &str);
+ else kputw(c->mtid, &str);
+ kputc('\t', &str);
+ }
+ kputw(c->mpos + 1, &str); kputc('\t', &str); kputw(c->isize, &str); kputc('\t', &str);
if (c->l_qseq) {
for (i = 0; i < c->l_qseq; ++i) kputc(bam_nt16_rev_table[bam1_seqi(s, i)], &str);
kputc('\t', &str);
if (t[0] == 0xff) kputc('*', &str);
else for (i = 0; i < c->l_qseq; ++i) kputc(t[i] + 33, &str);
- } else ksprintf(&str, "*\t*");
+ } else kputsn("*\t*", 3, &str);
s = bam1_aux(b);
while (s < b->data + b->data_len) {
uint8_t type, key[2];
key[0] = s[0]; key[1] = s[1];
s += 2; type = *s; ++s;
- ksprintf(&str, "\t%c%c:", key[0], key[1]);
- if (type == 'A') { ksprintf(&str, "A:%c", *s); ++s; }
- else if (type == 'C') { ksprintf(&str, "i:%u", *s); ++s; }
- else if (type == 'c') { ksprintf(&str, "i:%d", *s); ++s; }
- else if (type == 'S') { ksprintf(&str, "i:%u", *(uint16_t*)s); s += 2; }
- else if (type == 's') { ksprintf(&str, "i:%d", *(int16_t*)s); s += 2; }
- else if (type == 'I') { ksprintf(&str, "i:%u", *(uint32_t*)s); s += 4; }
- else if (type == 'i') { ksprintf(&str, "i:%d", *(int32_t*)s); s += 4; }
+ kputc('\t', &str); kputsn((char*)key, 2, &str); kputc(':', &str);
+ if (type == 'A') { kputsn("A:", 2, &str); kputc(*s, &str); ++s; }
+ else if (type == 'C') { kputsn("i:", 2, &str); kputw(*s, &str); ++s; }
+ else if (type == 'c') { kputsn("i:", 2, &str); kputw(*(int8_t*)s, &str); ++s; }
+ else if (type == 'S') { kputsn("i:", 2, &str); kputw(*(uint16_t*)s, &str); s += 2; }
+ else if (type == 's') { kputsn("i:", 2, &str); kputw(*(int16_t*)s, &str); s += 2; }
+ else if (type == 'I') { kputsn("i:", 2, &str); kputuw(*(uint32_t*)s, &str); s += 4; }
+ else if (type == 'i') { kputsn("i:", 2, &str); kputw(*(int32_t*)s, &str); s += 4; }
else if (type == 'f') { ksprintf(&str, "f:%g", *(float*)s); s += 4; }
else if (type == 'd') { ksprintf(&str, "d:%lg", *(double*)s); s += 8; }
- else if (type == 'Z' || type == 'H') { ksprintf(&str, "%c:", type); while (*s) kputc(*s++, &str); ++s; }
+ else if (type == 'Z' || type == 'H') { kputc(type, &str); kputc(':', &str); while (*s) kputc(*s++, &str); ++s; }
+ else if (type == 'B') {
+ uint8_t sub_type = *(s++);
+ int32_t n;
+ memcpy(&n, s, 4);
+ s += 4; // no point to the start of the array
+ kputc(type, &str); kputc(':', &str); kputc(sub_type, &str); // write the typing
+ for (i = 0; i < n; ++i) {
+ kputc(',', &str);
+ if ('c' == sub_type || 'c' == sub_type) { kputw(*(int8_t*)s, &str); ++s; }
+ else if ('C' == sub_type) { kputw(*(uint8_t*)s, &str); ++s; }
+ else if ('s' == sub_type) { kputw(*(int16_t*)s, &str); s += 2; }
+ else if ('S' == sub_type) { kputw(*(uint16_t*)s, &str); s += 2; }
+ else if ('i' == sub_type) { kputw(*(int32_t*)s, &str); s += 4; }
+ else if ('I' == sub_type) { kputuw(*(uint32_t*)s, &str); s += 4; }
+ else if ('f' == sub_type) { ksprintf(&str, "%g", *(float*)s); s += 4; }
+ }
+ }
}
return str.s;
}
void bam_view1(const bam_header_t *header, const bam1_t *b)
{
char *s = bam_format1(header, b);
- printf("%s\n", s);
+ puts(s);
free(s);
}
+
+int bam_validate1(const bam_header_t *header, const bam1_t *b)
+{
+ char *s;
+
+ if (b->core.tid < -1 || b->core.mtid < -1) return 0;
+ if (header && (b->core.tid >= header->n_targets || b->core.mtid >= header->n_targets)) return 0;
+
+ if (b->data_len < b->core.l_qname) return 0;
+ s = memchr(bam1_qname(b), '\0', b->core.l_qname);
+ if (s != &bam1_qname(b)[b->core.l_qname-1]) return 0;
+
+ // FIXME: Other fields could also be checked, especially the auxiliary data
+
+ return 1;
+}
+
+// FIXME: we should also check the LB tag associated with each alignment
+const char *bam_get_library(bam_header_t *h, const bam1_t *b)
+{
+ const uint8_t *rg;
+ if (h->dict == 0) h->dict = sam_header_parse2(h->text);
+ if (h->rg2lib == 0) h->rg2lib = sam_header2tbl(h->dict, "RG", "ID", "LB");
+ rg = bam_aux_get(b, "RG");
+ return (rg == 0)? 0 : sam_tbl_get(h->rg2lib, (const char*)(rg + 1));
+}
+
+/************
+ * Remove B *
+ ************/
+
+int bam_remove_B(bam1_t *b)
+{
+ int i, j, end_j, k, l, no_qual;
+ uint32_t *cigar, *new_cigar;
+ uint8_t *seq, *qual, *p;
+ // test if removal is necessary
+ if (b->core.flag & BAM_FUNMAP) return 0; // unmapped; do nothing
+ cigar = bam1_cigar(b);
+ for (k = 0; k < b->core.n_cigar; ++k)
+ if (bam_cigar_op(cigar[k]) == BAM_CBACK) break;
+ if (k == b->core.n_cigar) return 0; // no 'B'
+ if (bam_cigar_op(cigar[0]) == BAM_CBACK) goto rmB_err; // cannot be removed
+ // allocate memory for the new CIGAR
+ if (b->data_len + (b->core.n_cigar + 1) * 4 > b->m_data) { // not enough memory
+ b->m_data = b->data_len + b->core.n_cigar * 4;
+ kroundup32(b->m_data);
+ b->data = (uint8_t*)realloc(b->data, b->m_data);
+ cigar = bam1_cigar(b); // after realloc, cigar may be changed
+ }
+ new_cigar = (uint32_t*)(b->data + (b->m_data - b->core.n_cigar * 4)); // from the end of b->data
+ // the core loop
+ seq = bam1_seq(b); qual = bam1_qual(b);
+ no_qual = (qual[0] == 0xff); // test whether base quality is available
+ i = j = 0; end_j = -1;
+ for (k = l = 0; k < b->core.n_cigar; ++k) {
+ int op = bam_cigar_op(cigar[k]);
+ int len = bam_cigar_oplen(cigar[k]);
+ if (op == BAM_CBACK) { // the backward operation
+ int t, u;
+ if (k == b->core.n_cigar - 1) break; // ignore 'B' at the end of CIGAR
+ if (len > j) goto rmB_err; // an excessively long backward
+ for (t = l - 1, u = 0; t >= 0; --t) { // look back
+ int op1 = bam_cigar_op(new_cigar[t]);
+ int len1 = bam_cigar_oplen(new_cigar[t]);
+ if (bam_cigar_type(op1)&1) { // consume the query
+ if (u + len1 >= len) { // stop
+ new_cigar[t] -= (len - u) << BAM_CIGAR_SHIFT;
+ break;
+ } else u += len1;
+ }
+ }
+ if (bam_cigar_oplen(new_cigar[t]) == 0) --t; // squeeze out the zero-length operation
+ l = t + 1;
+ end_j = j; j -= len;
+ } else { // other CIGAR operations
+ new_cigar[l++] = cigar[k];
+ if (bam_cigar_type(op)&1) { // consume the query
+ if (i != j) { // no need to copy if i == j
+ int u, c, c0;
+ for (u = 0; u < len; ++u) { // construct the consensus
+ c = bam1_seqi(seq, i+u);
+ if (j + u < end_j) { // in an overlap
+ c0 = bam1_seqi(seq, j+u);
+ if (c != c0) { // a mismatch; choose the better base
+ if (qual[j+u] < qual[i+u]) { // the base in the 2nd segment is better
+ bam1_seq_seti(seq, j+u, c);
+ qual[j+u] = qual[i+u] - qual[j+u];
+ } else qual[j+u] -= qual[i+u]; // the 1st is better; reduce base quality
+ } else qual[j+u] = qual[j+u] > qual[i+u]? qual[j+u] : qual[i+u];
+ } else { // not in an overlap; copy over
+ bam1_seq_seti(seq, j+u, c);
+ qual[j+u] = qual[i+u];
+ }
+ }
+ }
+ i += len, j += len;
+ }
+ }
+ }
+ if (no_qual) qual[0] = 0xff; // in very rare cases, this may be modified
+ // merge adjacent operations if possible
+ for (k = 1; k < l; ++k)
+ if (bam_cigar_op(new_cigar[k]) == bam_cigar_op(new_cigar[k-1]))
+ new_cigar[k] += new_cigar[k-1] >> BAM_CIGAR_SHIFT << BAM_CIGAR_SHIFT, new_cigar[k-1] &= 0xf;
+ // kill zero length operations
+ for (k = i = 0; k < l; ++k)
+ if (new_cigar[k] >> BAM_CIGAR_SHIFT)
+ new_cigar[i++] = new_cigar[k];
+ l = i;
+ // update b
+ memcpy(cigar, new_cigar, l * 4); // set CIGAR
+ p = b->data + b->core.l_qname + l * 4;
+ memmove(p, seq, (j+1)>>1); p += (j+1)>>1; // set SEQ
+ memmove(p, qual, j); p += j; // set QUAL
+ memmove(p, bam1_aux(b), b->l_aux); p += b->l_aux; // set optional fields
+ b->core.n_cigar = l, b->core.l_qseq = j; // update CIGAR length and query length
+ b->data_len = p - b->data; // update record length
+ return 0;
+
+rmB_err:
+ b->core.flag |= BAM_FUNMAP;
+ return -1;
+}
projects / samtools.git / blobdiff