bcf_hdr_t *bcf_hdr_read(BGZF *fp)
{
uint8_t magic[5];
bcf_hdr_t *h;
h = bcf_hdr_init();
bgzf_read(fp, magic, 5);
if (strncmp((char*)magic, "BCF\2\1", 5) != 0) {
if (hts_verbose >= 2)
fprintf(stderr, "[E::%s] invalid BCF2 magic string\n", __func__);
bcf_hdr_destroy(h);
return 0;
}
bgzf_read(fp, &h->l_text, 4);
h->text = (char*)malloc(h->l_text);
bgzf_read(fp, h->text, h->l_text);
bcf_hdr_parse(h);
return h;
}
static bcf_hdr_t* make_vcf_hdr(cJSON *json, const char *in_path,
bool is_breakpoint, size_t kmer_size,
char const*const* ref_paths, size_t nref_paths,
read_t *chroms, size_t nchroms)
{
ctx_assert(json != NULL);
StrBuf hdrbuf;
strbuf_alloc(&hdrbuf, 1024);
char datestr[9];
time_t date = time(NULL);
strftime(datestr, 9, "%Y%m%d", localtime(&date));
strbuf_append_str(&hdrbuf, "##fileformat=VCFv4.2\n##fileDate=");
strbuf_append_str(&hdrbuf, datestr);
strbuf_append_str(&hdrbuf, "\n");
// Print commands used to generate header
cJSON *commands = json_hdr_get(json, "commands", cJSON_Array, in_path);
cJSON *command = commands->child;
// Print this command
char keystr[8];
char *prevstr = NULL;
size_t i;
if(command) {
cJSON *key = json_hdr_get(command, "key", cJSON_String, in_path);
prevstr = key->valuestring;
}
// Print command entry for this command
strbuf_append_str(&hdrbuf, "##mccortex_");
strbuf_append_str(&hdrbuf, hex_rand_str(keystr, sizeof(keystr)));
strbuf_append_str(&hdrbuf, "=<prev=\"");
strbuf_append_str(&hdrbuf, prevstr ? prevstr : "NULL");
strbuf_append_str(&hdrbuf, "\",cmd=\"");
strbuf_append_str(&hdrbuf, cmd_get_cmdline());
strbuf_append_str(&hdrbuf, "\",cwd=\"");
strbuf_append_str(&hdrbuf, cmd_get_cwd());
strbuf_append_str(&hdrbuf, "\",version="CTX_VERSION">\n");
// Print previous commands
vcf_hdrtxt_append_commands(command, &hdrbuf, in_path);
// Print field definitions
if(is_breakpoint)
strbuf_append_str(&hdrbuf, "##INFO=<ID=BRKPNT,Number=1,Type=String,Description=\"Breakpoint call\">\n");
else
strbuf_append_str(&hdrbuf, "##INFO=<ID=BUBBLE,Number=1,Type=String,Description=\"Bubble call\">\n");
strbuf_sprintf(&hdrbuf, "##INFO=<ID=K%zu,Number=0,Type=Flag,Description=\"Found at k=%zu\">\n", kmer_size, kmer_size);
strbuf_append_str(&hdrbuf, "##FORMAT=<ID=GT,Number=1,Type=String,Description=\"Genotype\">\n");
strbuf_append_str(&hdrbuf, "##FILTER=<ID=PASS,Description=\"All filters passed\">\n");
// Print reference paths
strbuf_append_str(&hdrbuf, "##reference=");
strbuf_append_str(&hdrbuf, ref_paths[0]);
for(i = 1; i < nref_paths; i++) {
strbuf_append_char(&hdrbuf, ',');
strbuf_append_str(&hdrbuf, ref_paths[i]);
}
strbuf_append_str(&hdrbuf, "\n");
// Print contigs lengths
for(i = 0; i < nchroms; i++) {
strbuf_sprintf(&hdrbuf, "##contig=<ID=%s,length=%zu>\n",
chroms[i].name.b, chroms[i].seq.end);
}
// Print VCF column header
strbuf_append_str(&hdrbuf, "#CHROM\tPOS\tID\tREF\tALT\tQUAL\tFILTER\tINFO\tFORMAT");
if(is_breakpoint)
{
// Print a column for each sample
cJSON *graph_json = json_hdr_get(json, "graph", cJSON_Object, in_path);
cJSON *colours_json = json_hdr_get(graph_json, "colours", cJSON_Array, in_path);
cJSON *colour_json = colours_json->child;
if(colour_json == NULL) die("Missing colours");
for(; colour_json; colour_json = colour_json->next)
{
if(!json_hdr_colour_is_ref(colour_json)) {
cJSON *sample_json = json_hdr_get(colour_json, "sample", cJSON_String, in_path);
strbuf_append_str(&hdrbuf, "\t");
strbuf_append_str(&hdrbuf, sample_json->valuestring);
}
}
}
strbuf_append_char(&hdrbuf, '\n');
bcf_hdr_t *hdr = bcf_hdr_init("w");
if(bcf_hdr_parse(hdr, hdrbuf.b) != 0) die("Cannot construct VCF header");
strbuf_dealloc(&hdrbuf);
return hdr;
}
bcf_hdr_t *vcf_hdr_read(htsFile *fp)
{
if (!fp->is_bin) {
kstring_t txt, *s = &fp->line;
bcf_hdr_t *h;
h = bcf_hdr_init();
txt.l = txt.m = 0; txt.s = 0;
while (hts_getline(fp, KS_SEP_LINE, s) >= 0) {
if (s->l == 0) continue;
if (s->s[0] != '#') {
if (hts_verbose >= 2)
fprintf(stderr, "[E::%s] no sample line\n", __func__);
free(txt.s);
bcf_hdr_destroy(h);
return 0;
}
if (s->s[1] != '#' && fp->fn_aux) { // insert contigs here
int dret;
gzFile f;
kstream_t *ks;
kstring_t tmp;
tmp.l = tmp.m = 0; tmp.s = 0;
f = gzopen(fp->fn_aux, "r");
ks = ks_init(f);
while (ks_getuntil(ks, 0, &tmp, &dret) >= 0) {
int c;
kputs("##contig=<ID=", &txt); kputs(tmp.s, &txt);
ks_getuntil(ks, 0, &tmp, &dret);
kputs(",length=", &txt); kputw(atol(tmp.s), &txt);
kputsn(">\n", 2, &txt);
if (dret != '\n')
while ((c = ks_getc(ks)) != '\n' && c != -1); // skip the rest of the line
}
free(tmp.s);
ks_destroy(ks);
gzclose(f);
}
kputsn(s->s, s->l, &txt);
if (s->s[1] != '#') break;
kputc('\n', &txt);
}
h->l_text = txt.l + 1; // including NULL
h->text = txt.s;
bcf_hdr_parse(h);
// check tabix index, are all contigs listed in the header? add the missing ones
tbx_t *idx = tbx_index_load(fp->fn);
if ( idx )
{
int i, n, need_sync = 0;
const char **names = tbx_seqnames(idx, &n);
for (i=0; i<n; i++)
{
bcf_hrec_t *hrec = bcf_hdr_get_hrec(h, BCF_DT_CTG, (char*) names[i]);
if ( hrec ) continue;
hrec = (bcf_hrec_t*) calloc(1,sizeof(bcf_hrec_t));
hrec->key = strdup("contig");
bcf_hrec_add_key(hrec, "ID", strlen("ID"));
bcf_hrec_set_val(hrec, hrec->nkeys-1, (char*) names[i], strlen(names[i]), 0);
bcf_hrec_add_key(hrec, "length", strlen("length"));
bcf_hrec_set_val(hrec, hrec->nkeys-1, "-1", strlen("-1"), 0); // what is a good default value?
bcf_hdr_add_hrec(h, hrec);
need_sync = 1;
}
free(names);
tbx_destroy(idx);
if ( need_sync )
{
bcf_hdr_sync(h);
bcf_hdr_fmt_text(h);
}
}
return h;
} else return bcf_hdr_read((BGZF*)fp->fp);
}
static void reheader_bcf(args_t *args, int is_compressed)
{
htsFile *fp = hts_open(args->fname, "r"); if ( !fp ) error("Failed to open: %s\n", args->fname);
bcf_hdr_t *hdr = bcf_hdr_read(fp); if ( !hdr ) error("Failed to read the header: %s\n", args->fname);
kstring_t htxt = {0,0,0};
int hlen;
htxt.s = bcf_hdr_fmt_text(hdr, 1, &hlen);
htxt.l = hlen;
int i, nsamples = 0;
char **samples = NULL;
if ( args->samples_fname )
samples = hts_readlines(args->samples_fname, &nsamples);
if ( args->header_fname )
{
free(htxt.s); htxt.s = NULL; htxt.l = htxt.m = 0;
read_header_file(args->header_fname, &htxt);
}
if ( samples )
{
set_samples(samples, nsamples, &htxt);
for (i=0; i<nsamples; i++) free(samples[i]);
free(samples);
}
bcf_hdr_t *hdr_out = bcf_hdr_init("r");
bcf_hdr_parse(hdr_out, htxt.s);
if ( args->header_fname ) hdr_out = strip_header(hdr, hdr_out);
// write the header and the body
htsFile *fp_out = hts_open("-",is_compressed ? "wb" : "wbu");
bcf_hdr_write(fp_out, hdr_out);
bcf1_t *rec = bcf_init();
while ( bcf_read(fp, hdr, rec)==0 )
{
// sanity checking, this slows things down. Make it optional?
bcf_unpack(rec, BCF_UN_ALL);
if ( rec->rid >= hdr_out->n[BCF_DT_CTG] || strcmp(bcf_hdr_int2id(hdr,BCF_DT_CTG,rec->rid),bcf_hdr_int2id(hdr_out,BCF_DT_CTG,rec->rid)) )
error("The CHROM is not defined: \"%s\"\n", bcf_hdr_int2id(hdr,BCF_DT_CTG,rec->rid));
for (i=0; i<rec->d.n_flt; i++)
{
int id = rec->d.flt[i];
if ( id >= hdr_out->n[BCF_DT_ID] ) break;
if ( !bcf_hdr_idinfo_exists(hdr_out,BCF_HL_FLT,id) ) break;
if ( strcmp(hdr->id[BCF_DT_ID][id].key,hdr_out->id[BCF_DT_ID][id].key) )
error("FIXME: Broken FILTER ids: %s vs %s\n", hdr->id[BCF_DT_ID][id].key,hdr_out->id[BCF_DT_ID][id].key);
}
if ( i!=rec->d.n_flt )
error("The FILTER is not defined: \"%s\"\n", bcf_hdr_int2id(hdr,BCF_DT_ID,rec->d.flt[i]));
for (i=0; i<rec->n_info; i++)
{
int id = rec->d.info[i].key;
if ( id >= hdr_out->n[BCF_DT_ID] ) break;
if ( !hdr_out->id[BCF_DT_ID][id].key ) break;
if ( !bcf_hdr_idinfo_exists(hdr_out,BCF_HL_INFO,id) ) break;
if ( strcmp(hdr->id[BCF_DT_ID][id].key,hdr_out->id[BCF_DT_ID][id].key) )
error("FIXME: Broken INFO ids: %s vs %s\n", hdr->id[BCF_DT_ID][id].key,hdr_out->id[BCF_DT_ID][id].key);
}
if ( i!=rec->n_info )
error("The INFO tag is not defined: \"%s\"\n", bcf_hdr_int2id(hdr,BCF_DT_ID,rec->d.info[i].key));
for (i=0; i<rec->n_fmt; i++)
{
int id = rec->d.fmt[i].id;
if ( id >= hdr_out->n[BCF_DT_ID] ) break;
if ( !hdr_out->id[BCF_DT_ID][id].key ) break;
if ( !bcf_hdr_idinfo_exists(hdr_out,BCF_HL_FMT,id) ) break;
if ( strcmp(hdr->id[BCF_DT_ID][id].key,hdr_out->id[BCF_DT_ID][id].key) )
error("FIXME: Broken FORMAT ids: %s vs %s\n", hdr->id[BCF_DT_ID][id].key,hdr_out->id[BCF_DT_ID][id].key);
}
if ( i!=rec->n_fmt )
error("The FORMAT tag is not defined: \"%s\"\n", bcf_hdr_int2id(hdr,BCF_DT_ID,rec->d.fmt[i].id));
bcf_write(fp_out,hdr_out,rec);
}
bcf_destroy(rec);
free(htxt.s);
hts_close(fp_out);
hts_close(fp);
bcf_hdr_destroy(hdr_out);
bcf_hdr_destroy(hdr);
}
