]> git.donarmstrong.com Git - samtools.git/blob - padding.c
Include leading CIGAR P operators in 'samtools depad'
[samtools.git] / padding.c
1 #include <string.h>
2 #include <assert.h>
3 #include "kstring.h"
4 #include "bam.h"
5
6 static void replace_cigar(bam1_t *b, int n, uint32_t *cigar)
7 {
8         if (n != b->core.n_cigar) {
9                 int o = b->core.l_qname + b->core.n_cigar * 4;
10                 if (b->data_len + (n - b->core.n_cigar) * 4 > b->m_data) {
11                         b->m_data = b->data_len + (n - b->core.n_cigar) * 4;
12                         kroundup32(b->m_data);
13                         b->data = (uint8_t*)realloc(b->data, b->m_data);
14                 }
15                 memmove(b->data + b->core.l_qname + n * 4, b->data + o, b->data_len - o);
16                 memcpy(b->data + b->core.l_qname, cigar, n * 4);
17                 b->data_len += (n - b->core.n_cigar) * 4;
18                 b->core.n_cigar = n;
19         } else memcpy(b->data + b->core.l_qname, cigar, n * 4);
20 }
21
22 #define write_cigar(_c, _n, _m, _v) do { \
23                 if (_n == _m) { \
24                         _m = _m? _m<<1 : 4; \
25                         _c = (uint32_t*)realloc(_c, _m * 4); \
26                 } \
27                 _c[_n++] = (_v); \
28         } while (0)
29
30 static void unpad_seq(bam1_t *b, kstring_t *s)
31 {
32         int k, j, i;
33         uint32_t *cigar = bam1_cigar(b);
34         uint8_t *seq = bam1_seq(b);
35         ks_resize(s, b->core.l_qseq);
36         for (k = 0, s->l = 0, j = 0; k < b->core.n_cigar; ++k) {
37                 int op, ol;
38                 op = bam_cigar_op(cigar[k]);
39                 ol = bam_cigar_oplen(cigar[k]);
40                 if (op == BAM_CMATCH || op == BAM_CEQUAL || op == BAM_CDIFF) {
41                         for (i = 0; i < ol; ++i, ++j) s->s[s->l++] = bam1_seqi(seq, j);
42                 } else if (op == BAM_CSOFT_CLIP) {
43                         j += ol;
44                 } else if (op == BAM_CDEL || op == BAM_CPAD) {
45                         for (i = 0; i < ol; ++i) s->s[s->l++] = 0;
46                 } else {
47                         fprintf(stderr, "[depad] ERROR: Didn't expect CIGAR op %c in read %s\n", BAM_CIGAR_STR[op], bam1_qname(b));
48                         assert(-1);
49                 }
50         }
51 }
52
53 int bam_pad2unpad(bamFile in, bamFile out)
54 {
55         bam_header_t *h;
56         bam1_t *b;
57         kstring_t r, q;
58         int r_tid = -1;
59         uint32_t *cigar2 = 0;
60         int n2 = 0, m2 = 0, *posmap = 0;
61
62         h = bam_header_read(in);
63         /* TODO - The reference sequence lengths in the BAM + SAM headers should be updated */
64         bam_header_write(out, h);
65         b = bam_init1();
66         r.l = r.m = q.l = q.m = 0; r.s = q.s = 0;
67         while (bam_read1(in, b) >= 0) {
68                 uint32_t *cigar = bam1_cigar(b);
69                 n2 = 0;
70                 if (b->core.pos == 0 && b->core.tid >= 0 && strcmp(bam1_qname(b), h->target_name[b->core.tid]) == 0) {
71                         int i, k;
72                         /*
73                         fprintf(stderr, "[depad] Found embedded reference %s\n", bam1_qname(b));
74                         */
75                         r_tid = b->core.tid;
76                         unpad_seq(b, &r);
77                         write_cigar(cigar2, n2, m2, bam_cigar_gen(b->core.l_qseq, BAM_CMATCH));
78                         replace_cigar(b, n2, cigar2);
79                         posmap = realloc(posmap, r.m * sizeof(int));
80                         for (i = k = 0; i < r.l; ++i) {
81                                 posmap[i] = k;
82                                 if (r.s[i]) ++k;
83                         }
84                 } else if (b->core.n_cigar > 0) {
85                         int i, k, op;
86                         if (b->core.tid < 0) {
87                                 fprintf(stderr, "[depad] ERROR: Read %s has CIGAR but no RNAME\n", bam1_qname(b));
88                                 return -1;
89                         } else if (b->core.tid != r_tid) {
90                                 fprintf(stderr, "[depad] ERROR: Missing %s embedded reference sequence\n", h->target_name[b->core.tid]);
91                                 return -1;
92                         }
93                         unpad_seq(b, &q);
94                         if (bam_cigar_op(cigar[0]) == BAM_CSOFT_CLIP) write_cigar(cigar2, n2, m2, cigar[0]);
95                         /* Include any pads if starts with an insert */
96                         for (k = 0; k+1 < b->core.pos && !r.s[b->core.pos - k - 1]; ++k);
97                         if (k) write_cigar(cigar2, n2, m2, bam_cigar_gen(k, BAM_CPAD));
98                         /* Determine CIGAR operator for each base in the aligned read */
99                         for (i = 0, k = b->core.pos; i < q.l; ++i, ++k)
100                                 q.s[i] = q.s[i]? (r.s[k]? BAM_CMATCH : BAM_CINS) : (r.s[k]? BAM_CDEL : BAM_CPAD);
101                         /* Count consecutive CIGAR operators to turn into a CIGAR string */
102                         for (i = k = 1, op = q.s[0]; i < q.l; ++i) {
103                                 if (op != q.s[i]) {
104                                         write_cigar(cigar2, n2, m2, bam_cigar_gen(k, op));
105                                         op = q.s[i]; k = 1;
106                                 } else ++k;
107                         }
108                         write_cigar(cigar2, n2, m2, bam_cigar_gen(k, op));
109                         if (bam_cigar_op(cigar[b->core.n_cigar-1]) == BAM_CSOFT_CLIP) write_cigar(cigar2, n2, m2, cigar[b->core.n_cigar-1]);
110                         /* Remove redundant P operators between M operators, e.g. 5M2P10M -> 15M */
111                         for (i = 2; i < n2; ++i)
112                                 if (bam_cigar_op(cigar2[i]) == BAM_CMATCH && bam_cigar_op(cigar2[i-1]) == BAM_CPAD && bam_cigar_op(cigar2[i-2]) == BAM_CMATCH)
113                                         cigar2[i] += cigar2[i-2], cigar2[i-2] = cigar2[i-1] = 0;
114                         for (i = k = 0; i < n2; ++i)
115                                 if (cigar2[i]) cigar2[k++] = cigar2[i];
116                         n2 = k;
117                         replace_cigar(b, n2, cigar2);
118                         b->core.pos = posmap[b->core.pos];
119                 }
120                 bam_write1(out, b);
121         }
122         free(r.s); free(q.s); free(posmap);
123         bam_destroy1(b);
124         bam_header_destroy(h);
125         return 0;
126 }
127
128 int main_pad2unpad(int argc, char *argv[])
129 {
130         bamFile in, out;
131         int result=0;
132         if (argc == 1) {
133                 fprintf(stderr, "Usage: samtools depad <in.bam>\n");
134                 return 1;
135         }
136         in = strcmp(argv[1], "-")? bam_open(argv[1], "r") : bam_dopen(fileno(stdin), "r");
137         out = bam_dopen(fileno(stdout), "w");
138         result = bam_pad2unpad(in, out);
139         bam_close(in); bam_close(out);
140         return result;
141 }