Merge commit 'upstream/0.1.18'
[samtools.git] / bcftools / vcf.c
1 #include <zlib.h>
2 #include <stdlib.h>
3 #include <stdio.h>
4 #include <string.h>
5 #include "bcf.h"
6 #include "kstring.h"
7 #include "kseq.h"
8 KSTREAM_INIT(gzFile, gzread, 4096)
9
10 typedef struct {
11         gzFile fp;
12         FILE *fpout;
13         kstream_t *ks;
14         void *refhash;
15         kstring_t line;
16         int max_ref;
17 } vcf_t;
18
19 bcf_hdr_t *vcf_hdr_read(bcf_t *bp)
20 {
21         kstring_t meta, smpl;
22         int dret;
23         vcf_t *v;
24         bcf_hdr_t *h;
25         if (!bp->is_vcf) return bcf_hdr_read(bp);
26         h = calloc(1, sizeof(bcf_hdr_t));
27         v = (vcf_t*)bp->v;
28         v->line.l = 0;
29         memset(&meta, 0, sizeof(kstring_t));
30         memset(&smpl, 0, sizeof(kstring_t));
31         while (ks_getuntil(v->ks, '\n', &v->line, &dret) >= 0) {
32                 if (v->line.l < 2) continue;
33                 if (v->line.s[0] != '#') return 0; // no sample line
34                 if (v->line.s[0] == '#' && v->line.s[1] == '#') {
35                         kputsn(v->line.s, v->line.l, &meta); kputc('\n', &meta);
36                 } else if (v->line.s[0] == '#') {
37                         int k;
38                         ks_tokaux_t aux;
39                         char *p;
40                         for (p = kstrtok(v->line.s, "\t\n", &aux), k = 0; p; p = kstrtok(0, 0, &aux), ++k) {
41                                 if (k >= 9) {
42                                         kputsn(p, aux.p - p, &smpl);
43                                         kputc('\0', &smpl);
44                                 }
45                         }
46                         break;
47                 }
48         }
49         kputc('\0', &meta);
50         h->name = 0;
51         h->sname = smpl.s; h->l_smpl = smpl.l;
52         h->txt = meta.s; h->l_txt = meta.l;
53         bcf_hdr_sync(h);
54         return h;
55 }
56
57 bcf_t *vcf_open(const char *fn, const char *mode)
58 {
59         bcf_t *bp;
60         vcf_t *v;
61         if (strchr(mode, 'b')) return bcf_open(fn, mode);
62         bp = calloc(1, sizeof(bcf_t));
63         v = calloc(1, sizeof(vcf_t));
64         bp->is_vcf = 1;
65         bp->v = v;
66         v->refhash = bcf_str2id_init();
67         if (strchr(mode, 'r')) {
68                 v->fp = strcmp(fn, "-")? gzopen(fn, "r") : gzdopen(fileno(stdin), "r");
69                 v->ks = ks_init(v->fp);
70         } else if (strchr(mode, 'w'))
71                 v->fpout = strcmp(fn, "-")? fopen(fn, "w") : stdout;
72         return bp;
73 }
74
75 int vcf_dictread(bcf_t *bp, bcf_hdr_t *h, const char *fn)
76 {
77         vcf_t *v;
78         gzFile fp;
79         kstream_t *ks;
80         kstring_t s, rn;
81         int dret;
82         if (bp == 0) return -1;
83         if (!bp->is_vcf) return 0;
84         s.l = s.m = 0; s.s = 0;
85         rn.m = rn.l = h->l_nm; rn.s = h->name;
86         v = (vcf_t*)bp->v;
87         fp = gzopen(fn, "r");
88         ks = ks_init(fp);
89         while (ks_getuntil(ks, 0, &s, &dret) >= 0) {
90                 bcf_str2id_add(v->refhash, strdup(s.s));
91                 kputs(s.s, &rn); kputc('\0', &rn);
92                 if (dret != '\n') ks_getuntil(ks, '\n', &s, &dret);
93         }
94         ks_destroy(ks);
95         gzclose(fp);
96         h->l_nm = rn.l; h->name = rn.s;
97         bcf_hdr_sync(h);
98         free(s.s);
99         return 0;
100 }
101
102 int vcf_close(bcf_t *bp)
103 {
104         vcf_t *v;
105         if (bp == 0) return -1;
106         if (!bp->is_vcf) return bcf_close(bp);
107         v = (vcf_t*)bp->v;
108         if (v->fp) {
109                 ks_destroy(v->ks);
110                 gzclose(v->fp);
111         }
112         if (v->fpout) fclose(v->fpout);
113         free(v->line.s);
114         bcf_str2id_thorough_destroy(v->refhash);
115         free(v);
116         free(bp);
117         return 0;
118 }
119
120 int vcf_hdr_write(bcf_t *bp, const bcf_hdr_t *h)
121 {
122         vcf_t *v = (vcf_t*)bp->v;
123         int i, has_ver = 0;
124         if (!bp->is_vcf) return bcf_hdr_write(bp, h);
125         if (h->l_txt > 0) {
126                 if (strstr(h->txt, "##fileformat=")) has_ver = 1;
127                 if (has_ver == 0) fprintf(v->fpout, "##fileformat=VCFv4.1\n");
128                 fwrite(h->txt, 1, h->l_txt - 1, v->fpout);
129         }
130         if (h->l_txt == 0) fprintf(v->fpout, "##fileformat=VCFv4.1\n");
131         fprintf(v->fpout, "#CHROM\tPOS\tID\tREF\tALT\tQUAL\tFILTER\tINFO\tFORMAT");
132         for (i = 0; i < h->n_smpl; ++i)
133                 fprintf(v->fpout, "\t%s", h->sns[i]);
134         fputc('\n', v->fpout);
135         return 0;
136 }
137
138 int vcf_write(bcf_t *bp, bcf_hdr_t *h, bcf1_t *b)
139 {
140         vcf_t *v = (vcf_t*)bp->v;
141         extern void bcf_fmt_core(const bcf_hdr_t *h, bcf1_t *b, kstring_t *s);
142         if (!bp->is_vcf) return bcf_write(bp, h, b);
143         bcf_fmt_core(h, b, &v->line);
144         fwrite(v->line.s, 1, v->line.l, v->fpout);
145         fputc('\n', v->fpout);
146         return v->line.l + 1;
147 }
148
149 int vcf_read(bcf_t *bp, bcf_hdr_t *h, bcf1_t *b)
150 {
151         int dret, k, i, sync = 0;
152         vcf_t *v = (vcf_t*)bp->v;
153         char *p, *q;
154         kstring_t str, rn;
155         ks_tokaux_t aux, a2;
156         if (!bp->is_vcf) return bcf_read(bp, h, b);
157         v->line.l = 0;
158         str.l = 0; str.m = b->m_str; str.s = b->str;
159         rn.l = rn.m = h->l_nm; rn.s = h->name;
160         if (ks_getuntil(v->ks, '\n', &v->line, &dret) < 0) return -1;
161         b->n_smpl = h->n_smpl;
162         for (p = kstrtok(v->line.s, "\t", &aux), k = 0; p; p = kstrtok(0, 0, &aux), ++k) {
163                 *(char*)aux.p = 0;
164                 if (k == 0) { // ref
165                         int tid = bcf_str2id(v->refhash, p);
166                         if (tid < 0) {
167                                 tid = bcf_str2id_add(v->refhash, strdup(p));
168                                 kputs(p, &rn); kputc('\0', &rn);
169                                 sync = 1;
170                         }
171                         b->tid = tid;
172                 } else if (k == 1) { // pos
173                         b->pos = atoi(p) - 1;
174                 } else if (k == 5) { // qual
175                         b->qual = (p[0] >= '0' && p[0] <= '9')? atof(p) : 0;
176                 } else if (k <= 8) { // variable length strings
177                         kputs(p, &str); kputc('\0', &str);
178                         b->l_str = str.l; b->m_str = str.m; b->str = str.s;
179                         if (k == 8) bcf_sync(b);
180                 } else { // k > 9
181                         if (strncmp(p, "./.", 3) == 0) {
182                                 for (i = 0; i < b->n_gi; ++i) {
183                                         if (b->gi[i].fmt == bcf_str2int("GT", 2)) {
184                                                 ((uint8_t*)b->gi[i].data)[k-9] = 1<<7;
185                                         } else if (b->gi[i].fmt == bcf_str2int("GQ", 2)) {
186                                                 ((uint8_t*)b->gi[i].data)[k-9] = 0;
187                                         } else if (b->gi[i].fmt == bcf_str2int("SP", 2)) {
188                                                 ((int32_t*)b->gi[i].data)[k-9] = 0;
189                                         } else if (b->gi[i].fmt == bcf_str2int("DP", 2)) {
190                                                 ((uint16_t*)b->gi[i].data)[k-9] = 0;
191                                         } else if (b->gi[i].fmt == bcf_str2int("PL", 2)) {
192                                                 int y = b->n_alleles * (b->n_alleles + 1) / 2;
193                                                 memset((uint8_t*)b->gi[i].data + (k - 9) * y, 0, y);
194                                         } else if (b->gi[i].fmt == bcf_str2int("GL", 2)) {
195                                                 int y = b->n_alleles * (b->n_alleles + 1) / 2;
196                                                 memset((float*)b->gi[i].data + (k - 9) * y, 0, y * 4);
197                                         }
198                                 }
199                                 goto endblock;
200                         }
201                         for (q = kstrtok(p, ":", &a2), i = 0; q && i < b->n_gi; q = kstrtok(0, 0, &a2), ++i) {
202                                 if (b->gi[i].fmt == bcf_str2int("GT", 2)) {
203                                         ((uint8_t*)b->gi[i].data)[k-9] = (q[0] - '0')<<3 | (q[2] - '0') | (q[1] == '/'? 0 : 1) << 6;
204                                 } else if (b->gi[i].fmt == bcf_str2int("GQ", 2)) {
205                                         double _x = strtod(q, &q);
206                                         int x = (int)(_x + .499);
207                                         if (x > 255) x = 255;
208                                         ((uint8_t*)b->gi[i].data)[k-9] = x;
209                                 } else if (b->gi[i].fmt == bcf_str2int("SP", 2)) {
210                                         int x = strtol(q, &q, 10);
211                                         if (x > 0xffff) x = 0xffff;
212                                         ((uint32_t*)b->gi[i].data)[k-9] = x;
213                                 } else if (b->gi[i].fmt == bcf_str2int("DP", 2)) {
214                                         int x = strtol(q, &q, 10);
215                                         if (x > 0xffff) x = 0xffff;
216                                         ((uint16_t*)b->gi[i].data)[k-9] = x;
217                                 } else if (b->gi[i].fmt == bcf_str2int("PL", 2)) {
218                                         int x, y, j;
219                                         uint8_t *data = (uint8_t*)b->gi[i].data;
220                                         y = b->n_alleles * (b->n_alleles + 1) / 2;
221                                         for (j = 0; j < y; ++j) {
222                                                 x = strtol(q, &q, 10);
223                                                 if (x > 255) x = 255;
224                                                 data[(k-9) * y + j] = x;
225                                                 ++q;
226                                         }
227                                 } else if (b->gi[i].fmt == bcf_str2int("GL", 2)) {
228                                         int j, y;
229                                         float x, *data = (float*)b->gi[i].data;
230                                         y = b->n_alleles * (b->n_alleles + 1) / 2;
231                                         for (j = 0; j < y; ++j) {
232                                                 x = strtod(q, &q);
233                                                 data[(k-9) * y + j] = x > 0? -x/10. : x;
234                                                 ++q;
235                                         }
236                                 }
237                         }
238                 endblock: i = i;
239                 }
240         }
241         h->l_nm = rn.l; h->name = rn.s;
242         if (sync) bcf_hdr_sync(h);
243         return v->line.l + 1;
244 }