static void phased_push(args_t *args, bcf1_t *arec, bcf1_t *brec)
{
if ( arec && arec->errcode )
error("Parse error at %s:%d, cannot proceed: %s\n", bcf_seqname(args->files->readers[0].header,arec),arec->pos+1, args->files->readers[0].fname);
if ( brec && brec->errcode )
error("Parse error at %s:%d, cannot proceed: %s\n", bcf_seqname(args->files->readers[1].header,brec),brec->pos+1, args->files->readers[1].fname);
int i, nsmpl = bcf_hdr_nsamples(args->out_hdr);
int chr_id = bcf_hdr_name2id(args->out_hdr, bcf_seqname(args->files->readers[0].header,arec));
if ( args->prev_chr<0 || args->prev_chr!=chr_id )
{
if ( args->prev_chr>=0 ) phased_flush(args);
for (i=0; i<nsmpl; i++)
args->phase_set[i] = arec->pos+1;
args->phase_set_changed = 1;
if ( args->seen_seq[chr_id] ) error("The chromosome block %s is not contiguous\n", bcf_seqname(args->files->readers[0].header,arec));
args->seen_seq[chr_id] = 1;
args->prev_chr = chr_id;
args->prev_pos_check = -1;
}
if ( !brec )
{
bcf_translate(args->out_hdr, args->files->readers[0].header, arec);
if ( args->nswap )
phase_update(args, args->out_hdr, arec);
if ( !args->compact_PS || args->phase_set_changed )
{
bcf_update_format_int32(args->out_hdr,arec,"PS",args->phase_set,nsmpl);
args->phase_set_changed = 0;
}
bcf_write(args->out_fh, args->out_hdr, arec);
if ( arec->pos < args->prev_pos_check )
error("FIXME, disorder: %s:%d in %s vs %d written [3]\n", bcf_seqname(args->files->readers[0].header,arec), arec->pos+1,args->files->readers[0].fname, args->prev_pos_check+1);
args->prev_pos_check = arec->pos;
return;
}
int m = args->mbuf;
args->nbuf += 2;
hts_expand(bcf1_t*,args->nbuf,args->mbuf,args->buf);
for (i=m; i<args->mbuf; i++)
args->buf[i] = bcf_init1();
SWAP(bcf1_t*, args->files->readers[0].buffer[0], args->buf[args->nbuf-2]);
SWAP(bcf1_t*, args->files->readers[1].buffer[0], args->buf[args->nbuf-1]);
}
static void concat(args_t *args)
{
int i;
if ( args->phased_concat ) // phased concat
{
// keep only two open files at a time
while ( args->ifname < args->nfnames )
{
int new_file = 0;
while ( args->files->nreaders < 2 && args->ifname < args->nfnames )
{
if ( !bcf_sr_add_reader(args->files,args->fnames[args->ifname]) ) error("Failed to open %s: %s\n", args->fnames[args->ifname],bcf_sr_strerror(args->files->errnum));
new_file = 1;
args->ifname++;
if ( args->start_pos[args->ifname-1]==-1 ) break; // new chromosome, start with only one file open
if ( args->ifname < args->nfnames && args->start_pos[args->ifname]==-1 ) break; // next file starts on a different chromosome
}
// is there a line from the previous run? Seek the newly opened reader to that position
int seek_pos = -1;
int seek_chr = -1;
if ( bcf_sr_has_line(args->files,0) )
{
bcf1_t *line = bcf_sr_get_line(args->files,0);
bcf_sr_seek(args->files, bcf_seqname(args->files->readers[0].header,line), line->pos);
seek_pos = line->pos;
seek_chr = bcf_hdr_name2id(args->out_hdr, bcf_seqname(args->files->readers[0].header,line));
}
else if ( new_file )
bcf_sr_seek(args->files,NULL,0); // set to start
int nret;
while ( (nret = bcf_sr_next_line(args->files)) )
{
if ( !bcf_sr_has_line(args->files,0) ) // no input from the first reader
{
// We are assuming that there is a perfect overlap, sites which are not present in both files are dropped
if ( ! bcf_sr_region_done(args->files,0) ) continue;
phased_flush(args);
bcf_sr_remove_reader(args->files, 0);
}
// Get a line to learn about current position
for (i=0; i<args->files->nreaders; i++)
if ( bcf_sr_has_line(args->files,i) ) break;
bcf1_t *line = bcf_sr_get_line(args->files,i);
// This can happen after bcf_sr_seek: indel may start before the coordinate which we seek to.
if ( seek_chr>=0 && seek_pos>line->pos && seek_chr==bcf_hdr_name2id(args->out_hdr, bcf_seqname(args->files->readers[i].header,line)) ) continue;
seek_pos = seek_chr = -1;
// Check if the position overlaps with the next, yet unopened, reader
int must_seek = 0;
while ( args->ifname < args->nfnames && args->start_pos[args->ifname]!=-1 && line->pos >= args->start_pos[args->ifname] )
{
must_seek = 1;
if ( !bcf_sr_add_reader(args->files,args->fnames[args->ifname]) ) error("Failed to open %s: %s\n", args->fnames[args->ifname],bcf_sr_strerror(args->files->errnum));
args->ifname++;
}
if ( must_seek )
{
bcf_sr_seek(args->files, bcf_seqname(args->files->readers[i].header,line), line->pos);
seek_pos = line->pos;
seek_chr = bcf_hdr_name2id(args->out_hdr, bcf_seqname(args->files->readers[i].header,line));
continue;
}
// We are assuming that there is a perfect overlap, sites which are not present in both files are dropped
if ( args->files->nreaders>1 && !bcf_sr_has_line(args->files,1) && !bcf_sr_region_done(args->files,1) ) continue;
phased_push(args, bcf_sr_get_line(args->files,0), args->files->nreaders>1 ? bcf_sr_get_line(args->files,1) : NULL);
}
if ( args->files->nreaders )
{
phased_flush(args);
while ( args->files->nreaders )
bcf_sr_remove_reader(args->files, 0);
}
}
}
else if ( args->files ) // combining overlapping files, using synced reader
{
while ( bcf_sr_next_line(args->files) )
{
for (i=0; i<args->files->nreaders; i++)
{
bcf1_t *line = bcf_sr_get_line(args->files,i);
if ( !line ) continue;
bcf_translate(args->out_hdr, args->files->readers[i].header, line);
bcf_write1(args->out_fh, args->out_hdr, line);
if ( args->remove_dups ) break;
}
}
}
else // concatenating
{
kstring_t tmp = {0,0,0};
int prev_chr_id = -1, prev_pos;
bcf1_t *line = bcf_init();
for (i=0; i<args->nfnames; i++)
{
htsFile *fp = hts_open(args->fnames[i], "r"); if ( !fp ) error("Failed to open: %s\n", args->fnames[i]);
bcf_hdr_t *hdr = bcf_hdr_read(fp); if ( !hdr ) error("Failed to parse header: %s\n", args->fnames[i]);
if ( !fp->is_bin && args->output_type&FT_VCF )
{
line->max_unpack = BCF_UN_STR;
// if VCF is on both input and output, avoid VCF to BCF conversion
while ( hts_getline(fp, KS_SEP_LINE, &fp->line) >=0 )
{
char *str = fp->line.s;
while ( *str && *str!='\t' ) str++;
tmp.l = 0;
kputsn(fp->line.s,str-fp->line.s,&tmp);
int chr_id = bcf_hdr_name2id(args->out_hdr, tmp.s);
if ( chr_id<0 ) error("The sequence \"%s\" not defined in the header: %s\n(Quick workaround: index the file.)\n", tmp.s, args->fnames[i]);
if ( prev_chr_id!=chr_id )
{
prev_pos = -1;
if ( args->seen_seq[chr_id] )
error("\nThe chromosome block %s is not contiguous, consider running with -a.\n", tmp.s);
}
char *end;
int pos = strtol(str+1,&end,10) - 1;
if ( end==str+1 ) error("Could not parse line: %s\n", fp->line.s);
if ( prev_pos > pos )
error("The chromosome block %s is not sorted, consider running with -a.\n", tmp.s);
args->seen_seq[chr_id] = 1;
prev_chr_id = chr_id;
if ( vcf_write_line(args->out_fh, &fp->line)!=0 ) error("Failed to write %d bytes\n", fp->line.l);
}
}
else
{
// BCF conversion is required
line->max_unpack = 0;
while ( bcf_read(fp, hdr, line)==0 )
{
bcf_translate(args->out_hdr, hdr, line);
if ( prev_chr_id!=line->rid )
{
prev_pos = -1;
if ( args->seen_seq[line->rid] )
error("\nThe chromosome block %s is not contiguous, consider running with -a.\n", bcf_seqname(args->out_hdr, line));
}
if ( prev_pos > line->pos )
error("The chromosome block %s is not sorted, consider running with -a.\n", bcf_seqname(args->out_hdr, line));
args->seen_seq[line->rid] = 1;
prev_chr_id = line->rid;
if ( bcf_write(args->out_fh, args->out_hdr, line)!=0 ) error("Failed to write\n");
}
}
bcf_hdr_destroy(hdr);
hts_close(fp);
}
bcf_destroy(line);
free(tmp.s);
}
}
static void phased_flush(args_t *args)
{
if ( !args->nbuf ) return;
bcf_hdr_t *ahdr = args->files->readers[0].header;
bcf_hdr_t *bhdr = args->files->readers[1].header;
int i, j, nsmpl = bcf_hdr_nsamples(args->out_hdr);
static int gt_absent_warned = 0;
for (i=0; i<args->nbuf; i+=2)
{
bcf1_t *arec = args->buf[i];
bcf1_t *brec = args->buf[i+1];
int nGTs = bcf_get_genotypes(ahdr, arec, &args->GTa, &args->mGTa);
if ( nGTs < 0 )
{
if ( !gt_absent_warned )
{
fprintf(stderr,"GT is not present at %s:%d. (This warning is printed only once.)\n", bcf_seqname(ahdr,arec), arec->pos+1);
gt_absent_warned = 1;
}
continue;
}
if ( nGTs != 2*nsmpl ) continue; // not diploid
nGTs = bcf_get_genotypes(bhdr, brec, &args->GTb, &args->mGTb);
if ( nGTs < 0 )
{
if ( !gt_absent_warned )
{
fprintf(stderr,"GT is not present at %s:%d. (This warning is printed only once.)\n", bcf_seqname(bhdr,brec), brec->pos+1);
gt_absent_warned = 1;
}
continue;
}
if ( nGTs != 2*nsmpl ) continue; // not diploid
for (j=0; j<nsmpl; j++)
{
int *gta = &args->GTa[j*2];
int *gtb = &args->GTb[j*2];
if ( gta[1]==bcf_int32_vector_end || gtb[1]==bcf_int32_vector_end ) continue;
if ( bcf_gt_is_missing(gta[0]) || bcf_gt_is_missing(gta[1]) || bcf_gt_is_missing(gtb[0]) || bcf_gt_is_missing(gtb[1]) ) continue;
if ( !bcf_gt_is_phased(gta[1]) || !bcf_gt_is_phased(gtb[1]) ) continue;
if ( bcf_gt_allele(gta[0])==bcf_gt_allele(gta[1]) || bcf_gt_allele(gtb[0])==bcf_gt_allele(gtb[1]) ) continue;
if ( bcf_gt_allele(gta[0])==bcf_gt_allele(gtb[0]) && bcf_gt_allele(gta[1])==bcf_gt_allele(gtb[1]) )
{
if ( args->swap_phase[j] ) args->nmism[j]++; else args->nmatch[j]++;
}
if ( bcf_gt_allele(gta[0])==bcf_gt_allele(gtb[1]) && bcf_gt_allele(gta[1])==bcf_gt_allele(gtb[0]) )
{
if ( args->swap_phase[j] ) args->nmatch[j]++; else args->nmism[j]++;
}
}
}
for (i=0; i<args->nbuf/2; i+=2)
{
bcf1_t *arec = args->buf[i];
bcf_translate(args->out_hdr, args->files->readers[0].header, arec);
if ( args->nswap )
phase_update(args, args->out_hdr, arec);
if ( !args->compact_PS || args->phase_set_changed )
{
bcf_update_format_int32(args->out_hdr,arec,"PS",args->phase_set,nsmpl);
args->phase_set_changed = 0;
}
bcf_write(args->out_fh, args->out_hdr, arec);
if ( arec->pos < args->prev_pos_check ) error("FIXME, disorder: %s:%d vs %d [1]\n", bcf_seqname(args->files->readers[0].header,arec),arec->pos+1,args->prev_pos_check+1);
args->prev_pos_check = arec->pos;
}
args->nswap = 0;
for (j=0; j<nsmpl; j++)
{
if ( args->nmatch[j] >= args->nmism[j] )
args->swap_phase[j] = 0;
else
{
args->swap_phase[j] = 1;
args->nswap++;
}
if ( args->nmatch[j] && args->nmism[j] )
{
// Entropy-inspired quality. The factor 0.7 shifts and scales to (0,1)
double f = (double)args->nmatch[j]/(args->nmatch[j]+args->nmism[j]);
args->phase_qual[j] = 99*(0.7 + f*log(f) + (1-f)*log(1-f))/0.7;
}
else
args->phase_qual[j] = 99;
args->nmatch[j] = 0;
args->nmism[j] = 0;
}
int PQ_printed = 0;
for (; i<args->nbuf; i+=2)
{
bcf1_t *brec = args->buf[i+1];
bcf_translate(args->out_hdr, args->files->readers[1].header, brec);
if ( !PQ_printed )
{
bcf_update_format_int32(args->out_hdr,brec,"PQ",args->phase_qual,nsmpl);
PQ_printed = 1;
for (j=0; j<nsmpl; j++)
if ( args->phase_qual[j] < args->min_PQ )
{
args->phase_set[j] = brec->pos+1;
args->phase_set_changed = 1;
}
else if ( args->compact_PS ) args->phase_set[j] = bcf_int32_missing;
}
if ( args->nswap )
phase_update(args, args->out_hdr, brec);
if ( !args->compact_PS || args->phase_set_changed )
{
bcf_update_format_int32(args->out_hdr,brec,"PS",args->phase_set,nsmpl);
args->phase_set_changed = 0;
}
bcf_write(args->out_fh, args->out_hdr, brec);
if ( brec->pos < args->prev_pos_check ) error("FIXME, disorder: %s:%d vs %d [2]\n", bcf_seqname(args->files->readers[1].header,brec),brec->pos+1,args->prev_pos_check+1);
args->prev_pos_check = brec->pos;
}
args->nbuf = 0;
}
int main(int argc, char **argv)
{
char *fname = argc>1 ? argv[1] : "/dev/null";
htsFile *fp = hts_open(fname, "w");
bcf_hdr_t *hdr1, *hdr2;
hdr1 = bcf_hdr_init("w");
hdr2 = bcf_hdr_init("w");
// Add two shared and two private annotations
bcf_hdr_append(hdr1, "##contig=<ID=1>");
bcf_hdr_append(hdr1, "##contig=<ID=2>");
bcf_hdr_append(hdr2, "##contig=<ID=2>");
bcf_hdr_append(hdr2, "##contig=<ID=1>");
bcf_hdr_append(hdr1, "##FILTER=<ID=FLT1,Description=\"Filter 1\">");
bcf_hdr_append(hdr1, "##FILTER=<ID=FLT2,Description=\"Filter 2\">");
bcf_hdr_append(hdr1, "##FILTER=<ID=FLT3,Description=\"Filter 3\">");
bcf_hdr_append(hdr2, "##FILTER=<ID=FLT4,Description=\"Filter 4\">");
bcf_hdr_append(hdr2, "##FILTER=<ID=FLT3,Description=\"Filter 3\">");
bcf_hdr_append(hdr2, "##FILTER=<ID=FLT2,Description=\"Filter 2\">");
bcf_hdr_append(hdr1, "##INFO=<ID=INF1,Number=.,Type=Integer,Description=\"Info 1\">");
bcf_hdr_append(hdr1, "##INFO=<ID=INF2,Number=.,Type=Integer,Description=\"Info 2\">");
bcf_hdr_append(hdr1, "##INFO=<ID=INF3,Number=.,Type=Integer,Description=\"Info 3\">");
bcf_hdr_append(hdr2, "##INFO=<ID=INF4,Number=.,Type=Integer,Description=\"Info 4\">");
bcf_hdr_append(hdr2, "##INFO=<ID=INF3,Number=.,Type=Integer,Description=\"Info 3\">");
bcf_hdr_append(hdr2, "##INFO=<ID=INF2,Number=.,Type=Integer,Description=\"Info 2\">");
bcf_hdr_append(hdr1, "##FORMAT=<ID=FMT1,Number=.,Type=Integer,Description=\"FMT 1\">");
bcf_hdr_append(hdr1, "##FORMAT=<ID=FMT2,Number=.,Type=Integer,Description=\"FMT 2\">");
bcf_hdr_append(hdr1, "##FORMAT=<ID=FMT3,Number=.,Type=Integer,Description=\"FMT 3\">");
bcf_hdr_append(hdr2, "##FORMAT=<ID=FMT4,Number=.,Type=Integer,Description=\"FMT 4\">");
bcf_hdr_append(hdr2, "##FORMAT=<ID=FMT3,Number=.,Type=Integer,Description=\"FMT 3\">");
bcf_hdr_append(hdr2, "##FORMAT=<ID=FMT2,Number=.,Type=Integer,Description=\"FMT 2\">");
bcf_hdr_add_sample(hdr1,"SMPL1");
bcf_hdr_add_sample(hdr1,"SMPL2");
bcf_hdr_add_sample(hdr2,"SMPL1");
bcf_hdr_add_sample(hdr2,"SMPL2");
bcf_hdr_sync(hdr1);
bcf_hdr_sync(hdr2);
hdr2 = bcf_hdr_merge(hdr2,hdr1);
bcf_hdr_sync(hdr2);
if ( bcf_hdr_write(fp, hdr2)!=0 ) error("Failed to write to %s\n", fname);
bcf1_t *rec = bcf_init1();
rec->rid = bcf_hdr_name2id(hdr1, "1");
rec->pos = 0;
bcf_update_alleles_str(hdr1, rec, "G,A");
int32_t tmpi[3];
tmpi[0] = bcf_hdr_id2int(hdr1, BCF_DT_ID, "FLT1");
tmpi[1] = bcf_hdr_id2int(hdr1, BCF_DT_ID, "FLT2");
tmpi[2] = bcf_hdr_id2int(hdr1, BCF_DT_ID, "FLT3");
bcf_update_filter(hdr1, rec, tmpi, 3);
tmpi[0] = 1; bcf_update_info_int32(hdr1, rec, "INF1", tmpi, 1);
tmpi[0] = 2; bcf_update_info_int32(hdr1, rec, "INF2", tmpi, 1);
tmpi[0] = 3; bcf_update_info_int32(hdr1, rec, "INF3", tmpi, 1);
tmpi[0] = tmpi[1] = 1; bcf_update_format_int32(hdr1, rec, "FMT1", tmpi, 2);
tmpi[0] = tmpi[1] = 2; bcf_update_format_int32(hdr1, rec, "FMT2", tmpi, 2);
tmpi[0] = tmpi[1] = 3; bcf_update_format_int32(hdr1, rec, "FMT3", tmpi, 2);
bcf_remove_filter(hdr1, rec, bcf_hdr_id2int(hdr1, BCF_DT_ID, "FLT2"), 0);
bcf_update_info_int32(hdr1, rec, "INF2", NULL, 0);
bcf_update_format_int32(hdr1, rec, "FMT2", NULL, 0);
bcf_translate(hdr2, hdr1, rec);
if ( bcf_write(fp, hdr2, rec)!=0 ) error("Faild to write to %s\n", fname);
// Clean
bcf_destroy1(rec);
bcf_hdr_destroy(hdr1);
bcf_hdr_destroy(hdr2);
int ret;
if ( (ret=hts_close(fp)) )
{
fprintf(stderr,"hts_close(%s): non-zero status %d\n",fname,ret);
exit(ret);
}
return 0;
}
