Imported Upstream version 0.1.6~dfsg
[samtools.git] / bam_import.c
index fccaa022208131b27093a2b44f32e74d13a469c0..1dc906eb49f36428460be547b1582ab52432f651 100644 (file)
@@ -5,6 +5,9 @@
 #include <stdlib.h>
 #include <unistd.h>
 #include <assert.h>
+#ifdef _WIN32
+#include <fcntl.h>
+#endif
 #include "kstring.h"
 #include "bam.h"
 #include "kseq.h"
@@ -36,6 +39,25 @@ 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 {
@@ -99,9 +121,10 @@ 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) {
@@ -296,8 +319,19 @@ 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) {
@@ -325,7 +359,7 @@ 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;
@@ -339,7 +373,13 @@ 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;
@@ -352,16 +392,18 @@ 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))
@@ -457,9 +499,11 @@ 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);
        return fp;
 }