static void print_index_sequences(ostream& fout, Ebwt& ebwt)
{
EList<string>* refnames = &(ebwt.refnames());
TStr cat_ref;
ebwt.restore(cat_ref);
uint32_t curr_ref = 0xffffffff;
string curr_ref_seq = "";
uint32_t curr_ref_len = 0xffffffff;
uint32_t last_text_off = 0;
size_t orig_len = cat_ref.length();
uint32_t tlen = 0xffffffff;
bool first = true;
for(size_t i = 0; i < orig_len; i++) {
uint32_t tidx = 0xffffffff;
uint32_t textoff = 0xffffffff;
tlen = 0xffffffff;
bool straddled = false;
ebwt.joinedToTextOff(1 /* qlen */, (uint32_t)i, tidx, textoff, tlen, true, straddled);
if (tidx != 0xffffffff && textoff < tlen)
{
if (curr_ref != tidx)
{
if (curr_ref != 0xffffffff)
{
// Add trailing gaps, if any exist
if(curr_ref_seq.length() < curr_ref_len) {
curr_ref_seq += string(curr_ref_len - curr_ref_seq.length(), 'N');
}
print_fasta_record(fout, (*refnames)[curr_ref], curr_ref_seq);
}
curr_ref = tidx;
curr_ref_seq = "";
curr_ref_len = tlen;
last_text_off = 0;
first = true;
}
uint32_t textoff_adj = textoff;
if(first && textoff > 0) textoff_adj++;
if (textoff_adj - last_text_off > 1)
curr_ref_seq += string(textoff_adj - last_text_off - 1, 'N');
curr_ref_seq.push_back(cat_ref[i]);
last_text_off = textoff;
first = false;
}
}
if (curr_ref < refnames->size())
{
// Add trailing gaps, if any exist
if(curr_ref_seq.length() < curr_ref_len) {
curr_ref_seq += string(curr_ref_len - curr_ref_seq.length(), 'N');
}
print_fasta_record(fout, (*refnames)[curr_ref], curr_ref_seq);
}
}
static uint64_t count_idx_kmers ( Ebwt<index_t>& ebwt)
{
TStr cat_ref;
ebwt.restore(cat_ref);
cerr << "Index loaded" << endl;
#ifdef TEST_KMER_COUNTING
std::set<uint64_t> my_set;
#endif
HyperLogLogPlusMinus<uint64_t> kmer_counter(16);
uint64_t word = 0;
uint64_t curr_length = 0;
uint8_t k = 32;
TIndexOffU curr_ref = OFF_MASK;
TIndexOffU last_text_off = 0;
size_t orig_len = cat_ref.length();
TIndexOffU tlen = OFF_MASK;
bool first = true;
for(size_t i = 0; i < orig_len; i++) {
TIndexOffU tidx = OFF_MASK;
TIndexOffU textoff = OFF_MASK;
tlen = OFF_MASK;
bool straddled = false;
ebwt.joinedToTextOff(1 /* qlen */, (TIndexOffU)i, tidx, textoff, tlen, true, straddled);
if (tidx != OFF_MASK && textoff < tlen) {
if (curr_ref != tidx) {
// End of the sequence - reset word and counter
curr_ref = tidx;
word = 0; curr_length = 0;
last_text_off = 0;
first = true;
}
TIndexOffU textoff_adj = textoff;
if(first && textoff > 0) textoff_adj++;
if (textoff_adj - last_text_off > 1) {
// there's an N - reset word and counter
word = 0; curr_length = 0;
}
// add another char.
int bp = (int)cat_ref[i];
// shift the first two bits off the word
word = word << 2;
// put the base-pair code from pos at that position
word |= bp;
++curr_length;
//cerr << "[" << i << "; " << curr_length << "; " << word << ":" << kmer_counter.cardinality() << "]" << endl;
if (curr_length >= k) {
kmer_counter.add(word);
#ifdef TEST_KMER_COUNTING
my_set.insert(word);
cerr << " " << kmer_counter.cardinality() << " vs " << my_set.size() << endl;
#endif
}
last_text_off = textoff;
first = false;
}
}
if (curr_length >= k) {
kmer_counter.add(word);
#ifdef TEST_KMER_COUNTING
my_set.insert(word);
#endif
}
#ifdef TEST_KMER_COUNTING
cerr << "Exact count: " << my_set.size() << endl;
#endif
return kmer_counter.cardinality();
}
static void print_index_sequences(ostream& fout, Ebwt<index_t>& ebwt)
{
EList<string>* refnames = &(ebwt.refnames());
TStr cat_ref;
ebwt.restore(cat_ref);
HyperLogLogPlusMinus<uint64_t> kmer_counter;
TIndexOffU curr_ref = OFF_MASK;
string curr_ref_seq = "";
TIndexOffU curr_ref_len = OFF_MASK;
TIndexOffU last_text_off = 0;
size_t orig_len = cat_ref.length();
TIndexOffU tlen = OFF_MASK;
bool first = true;
for(size_t i = 0; i < orig_len; i++) {
TIndexOffU tidx = OFF_MASK;
TIndexOffU textoff = OFF_MASK;
tlen = OFF_MASK;
bool straddled = false;
ebwt.joinedToTextOff(1 /* qlen */, (TIndexOffU)i, tidx, textoff, tlen, true, straddled);
if (tidx != OFF_MASK && textoff < tlen)
{
if (curr_ref != tidx)
{
if (curr_ref != OFF_MASK)
{
// Add trailing gaps, if any exist
if(curr_ref_seq.length() < curr_ref_len) {
curr_ref_seq += string(curr_ref_len - curr_ref_seq.length(), 'N');
}
print_fasta_record(fout, (*refnames)[curr_ref], curr_ref_seq);
}
curr_ref = tidx;
curr_ref_seq = "";
curr_ref_len = tlen;
last_text_off = 0;
first = true;
}
TIndexOffU textoff_adj = textoff;
if(first && textoff > 0) textoff_adj++;
if (textoff_adj - last_text_off > 1)
curr_ref_seq += string(textoff_adj - last_text_off - 1, 'N');
curr_ref_seq.push_back("ACGT"[int(cat_ref[i])]);
last_text_off = textoff;
first = false;
}
}
if (curr_ref < refnames->size())
{
// Add trailing gaps, if any exist
if(curr_ref_seq.length() < curr_ref_len) {
curr_ref_seq += string(curr_ref_len - curr_ref_seq.length(), 'N');
}
print_fasta_record(fout, (*refnames)[curr_ref], curr_ref_seq);
}
}
void print_index_sequences(
ostream& fout,
Ebwt<TStr>& ebwt,
const BitPairReference& refs)
{
vector<string>* refnames = &(ebwt.refnames());
TStr cat_ref;
ebwt.restore(cat_ref);
TIndexOffU curr_ref = OFF_MASK;
string curr_ref_seq = "";
TIndexOffU curr_ref_len = OFF_MASK;
uint32_t last_text_off = 0;
size_t orig_len = seqan::length(cat_ref);
TIndexOffU tlen = OFF_MASK;
bool first = true;
for(size_t i = 0; i < orig_len; i++) {
TIndexOffU tidx = OFF_MASK;
TIndexOffU textoff = OFF_MASK;
tlen = OFF_MASK;
ebwt.joinedToTextOff(1 /* qlen */, (TIndexOffU)i, tidx, textoff, tlen);
if (tidx != OFF_MASK && textoff < tlen)
{
if (curr_ref != tidx)
{
if (curr_ref != OFF_MASK)
{
// Add trailing gaps, if any exist
if(curr_ref_seq.length() < curr_ref_len) {
curr_ref_seq += string(curr_ref_len - curr_ref_seq.length(), 'N');
}
print_fasta_record(fout, (*refnames)[curr_ref], curr_ref_seq);
}
curr_ref = tidx;
curr_ref_seq = "";
curr_ref_len = tlen;
last_text_off = 0;
first = true;
}
TIndexOffU textoff_adj = textoff;
if(first && textoff > 0) textoff_adj++;
if (textoff_adj - last_text_off > 1)
curr_ref_seq += string(textoff_adj - last_text_off - 1, 'N');
curr_ref_seq.push_back(getValue(cat_ref,i));
last_text_off = textoff;
first = false;
}
}
if (curr_ref < refnames->size())
{
// Add trailing gaps, if any exist
if(curr_ref_seq.length() < curr_ref_len) {
curr_ref_seq += string(curr_ref_len - curr_ref_seq.length(), 'N');
}
print_fasta_record(fout, (*refnames)[curr_ref], curr_ref_seq);
}
}