PBWT *pbwtSubRange (PBWT *pOld, int start, int end)
{
int M = pOld->M ;
PBWT *pNew = pbwtCreate (M) ;
int i, j, k ;
uchar *x ;
PbwtCursor *uOld = pbwtCursorCreate (pOld, TRUE, TRUE) ;
pNew->yz = arrayCreate (pNew->N*8, uchar) ;
PbwtCursor *uNew = pbwtCursorCreate (pNew, TRUE, TRUE) ;
if (!pOld || !pOld->yz) die ("subrange without an existing pbwt") ;
if (start < 0 || end > pOld->N || end <= start)
die ("subrange invalid start %d, end %d", start, end) ;
x = myalloc (M, uchar) ;
if (pOld->sites) pNew->sites = arrayCreate (4096, Site) ;
for (i = 0 ; i < end ; ++i)
{ if (i >= start)
{ for (j = 0 ; j < M ; ++j) x[uOld->a[j]] = uOld->y[j] ;
for (j = 0 ; j < M ; ++j) uNew->y[j] = x[uNew->a[j]] ;
pbwtCursorWriteForwards (uNew) ;
if (pOld->sites) array(pNew->sites, pNew->N, Site) = arr(pOld->sites, i, Site) ;
++pNew->N ;
}
pbwtCursorForwardsRead (uOld) ;
}
pNew->aFend = myalloc (pNew->M, int) ; memcpy (pNew->aFend, uNew->a, pNew->M*sizeof(int)) ;
pbwtDestroy (pOld) ; pbwtCursorDestroy (uOld) ; pbwtCursorDestroy (uNew) ;
free(x) ;
return pNew ;
}
PBWT *pbwtSubSites (PBWT *pOld, double fmin, double frac)
{
int M = pOld->M ;
PBWT *pNew = pbwtCreate (M, 0) ;
int i, j, k, thresh = M*(1-fmin) ;
double bit = 0.0 ;
uchar *x ;
PbwtCursor *uOld = pbwtCursorCreate (pOld, TRUE, TRUE) ;
PbwtCursor *uNew = pbwtCursorCreate (pNew, TRUE, TRUE) ;
if (!pOld || !pOld->yz) die ("subsites without an existing pbwt") ;
if (fmin < 0 || fmin >= 1 || frac <= 0 || frac > 1)
die ("fmin %f, frac %f for subsites out of range\n", fmin, frac) ;
x = myalloc (M, uchar) ;
if (pOld->sites) pNew->sites = arrayCreate (4096, Site) ;
for (i = 0 ; i < pOld->N ; ++i)
{ if ((uOld->c < thresh) && ((bit += frac) > 1.0))
{ for (j = 0 ; j < M ; ++j) x[uOld->a[j]] = uOld->y[j] ;
for (j = 0 ; j < M ; ++j) uNew->y[j] = x[uNew->a[j]] ;
pbwtCursorWriteForwards (uNew) ;
if (pOld->sites) array(pNew->sites, pNew->N, Site) = arr(pOld->sites, i, Site) ;
++pNew->N ;
bit -= 1.0 ;
}
pbwtCursorForwardsRead (uOld) ;
}
pbwtCursorToAFend (uNew, pNew) ;
fprintf (logFile, "subsites with fmin %f, frac %f leaves %d sites\n", fmin, frac, pNew->N) ;
pNew->chrom = pOld->chrom ; pOld->chrom = 0 ;
pNew->samples = pOld->samples ; pOld->samples = 0 ;
pNew->missingOffset = pOld->missingOffset ; pOld->missingOffset = 0 ;
pNew->zMissing = pOld->zMissing ; pOld->zMissing = 0 ;
pbwtDestroy (pOld) ; pbwtCursorDestroy (uOld) ; pbwtCursorDestroy (uNew) ;
free(x) ;
return pNew ;
}
PBWT *pbwtSubRange (PBWT *pOld, int start, int end)
{
int M = pOld->M ;
PBWT *pNew = pbwtCreate (M, 0) ;
int i, j, k ;
uchar *x ;
PbwtCursor *uOld = pbwtCursorCreate (pOld, TRUE, TRUE) ;
PbwtCursor *uNew = pbwtCursorCreate (pNew, TRUE, TRUE) ;
if (!pOld || !pOld->yz) die ("subrange without an existing pbwt") ;
if (start < 0 || end > pOld->N || end <= start)
die ("subrange invalid start %d, end %d", start, end) ;
x = myalloc (M, uchar) ;
if (pOld->sites) pNew->sites = arrayCreate (4096, Site) ;
for (i = 0 ; i < end ; ++i)
{ if (i >= start)
{ for (j = 0 ; j < M ; ++j) x[uOld->a[j]] = uOld->y[j] ;
for (j = 0 ; j < M ; ++j) uNew->y[j] = x[uNew->a[j]] ;
pbwtCursorWriteForwards (uNew) ;
if (pOld->sites) array(pNew->sites, pNew->N, Site) = arr(pOld->sites, i, Site) ;
++pNew->N ;
}
pbwtCursorForwardsRead (uOld) ;
}
pbwtCursorToAFend (uNew, pNew) ;
pNew->chrom = pOld->chrom ; pOld->chrom = 0 ;
pNew->samples = pOld->samples ; pOld->samples = 0 ;
pNew->missingOffset = pOld->missingOffset ; pOld->missingOffset = 0 ;
pNew->zMissing = pOld->zMissing ; pOld->zMissing = 0 ;
pbwtDestroy (pOld) ; pbwtCursorDestroy (uOld) ; pbwtCursorDestroy (uNew) ;
free(x) ;
return pNew ;
}
PBWT *pbwtMerge(const char **fnames, int nfiles)
{
pbwt_reader_t *reader = pbwt_reader_init(fnames, nfiles);
int nhaps = 0, i;
for (i=0; i<nfiles; i++) nhaps += reader->pbwt[i]->M;
PBWT *out_pbwt = pbwtCreate(nhaps, 0);
PbwtCursor *cursor = pbwtNakedCursorCreate(nhaps, 0);
uchar *yseq = myalloc(nhaps, uchar);
out_pbwt->yz = arrayCreate (1<<20, uchar) ;
out_pbwt->sites = arrayReCreate(out_pbwt->sites, reader->pbwt[0]->N, Site);
out_pbwt->chrom = strdup(reader->pbwt[0]->chrom);
int pos, j;
while ( (pos=pbwt_reader_next(reader, nfiles)) )
{
// Merge only records shared by all files
for (i=0; i<nfiles; i++)
{
PBWT *p = reader->pbwt[i];
Site *site = arrp(p->sites, reader->cpos[i], Site);
// Both position and alleles must match. This requires that the records are sorted by alleles.
if ( site->x!=pos ) break;
char *als = dictName(variationDict, site->varD);
if ( strcmp(als,reader->mals) ) break;
}
if ( i!=nfiles )
{
// intersection: skip records which are not present in all files
for (i=0; i<nfiles; i++)
{
PBWT *p = reader->pbwt[i];
Site *site = arrp(p->sites, reader->cpos[i], Site);
if ( site->x!=pos ) continue;
char *als = dictName(variationDict, site->varD);
if ( strcmp(als,reader->mals) ) continue;
PbwtCursor *c = reader->cursor[i];
reader->unpacked[i] += unpack3(arrp(p->yz,reader->unpacked[i],uchar), p->M, c->y, 0);
pbwtCursorForwardsA(c);
}
continue;
}
// read and merge
int ihap = 0;
for (i=0; i<nfiles; i++)
{
PbwtCursor *c = reader->cursor[i];
PBWT *p = reader->pbwt[i];
Site *site = arrp(p->sites, reader->cpos[i], Site);
reader->unpacked[i] += unpack3(arrp(p->yz,reader->unpacked[i],uchar), p->M, c->y, 0);
for (j=0; j<p->M; j++) yseq[ihap + c->a[j]] = c->y[j];
pbwtCursorForwardsA(c);
ihap += p->M;
}
// pack merged haplotypes
for (j=0; j<nhaps; j++)
cursor->y[j] = yseq[cursor->a[j]];
pack3arrayAdd(cursor->y, out_pbwt->M, out_pbwt->yz);
pbwtCursorForwardsA(cursor);
// insert new site
arrayExtend(out_pbwt->sites, out_pbwt->N+1);
Site *site = arrayp(out_pbwt->sites, out_pbwt->N, Site);
site->x = pos;
dictAdd(variationDict, reader->mals, &site->varD);
out_pbwt->N++;
}
pbwtCursorToAFend (cursor, out_pbwt) ;
free(yseq);
pbwtCursorDestroy(cursor);
pbwt_reader_destroy(reader);
return out_pbwt;
}
