bool
CShortReadFastaInputSource::x_ReadFromTwoFiles(CBioseq_set& bioseq_set,
TSeqPos batch_size,
CShortReadFastaInputSource::EInputFormat format)
{
if (format == eFastc) {
NCBI_THROW(CInputException, eInvalidInput, "FASTC format cannot be "
"used with two files");
}
// tags to indicate paired sequences
CRef<CSeqdesc> seqdesc_first(new CSeqdesc);
seqdesc_first->SetUser().SetType().SetStr("Mapping");
seqdesc_first->SetUser().AddField("has_pair", eFirstSegment);
CRef<CSeqdesc> seqdesc_last(new CSeqdesc);
seqdesc_last->SetUser().SetType().SetStr("Mapping");
seqdesc_last->SetUser().AddField("has_pair", eLastSegment);
m_BasesAdded = 0;
while (m_BasesAdded < batch_size && !m_LineReader->AtEOF() &&
!m_SecondLineReader->AtEOF()) {
CRef<CSeq_entry> first;
CRef<CSeq_entry> second;
if (format == eFasta) {
first = x_ReadFastaOneSeq(m_LineReader);
second = x_ReadFastaOneSeq(m_SecondLineReader);
}
else {
first = x_ReadFastqOneSeq(m_LineReader);
second = x_ReadFastqOneSeq(m_SecondLineReader);
}
if (first.NotEmpty()) {
if (second.NotEmpty()) {
first->SetSeq().SetDescr().Set().push_back(seqdesc_first);
}
bioseq_set.SetSeq_set().push_back(first);
}
if (second.NotEmpty()) {
if (first.NotEmpty()) {
second->SetSeq().SetDescr().Set().push_back(seqdesc_last);
}
bioseq_set.SetSeq_set().push_back(second);
}
}
return true;
}
static void UnpackSeqSet(CBioseq_set& bss, SequenceSet *parent, SequenceSet::SequenceList& seqlist)
{
CBioseq_set::TSeq_set::iterator q, qe = bss.SetSeq_set().end();
for (q=bss.SetSeq_set().begin(); q!=qe; ++q) {
if (q->GetObject().IsSeq()) {
// only store amino acid or nucleotide sequences
if (q->GetObject().GetSeq().GetInst().GetMol() != CSeq_inst::eMol_aa &&
q->GetObject().GetSeq().GetInst().GetMol() != CSeq_inst::eMol_dna &&
q->GetObject().GetSeq().GetInst().GetMol() != CSeq_inst::eMol_rna &&
q->GetObject().GetSeq().GetInst().GetMol() != CSeq_inst::eMol_na)
continue;
const Sequence *sequence = new Sequence(parent, q->GetObject().SetSeq());
if (!sequence || !sequence->identifier)
FATALMSG("Can't create Sequence object, aborting...");
seqlist.push_back(sequence);
} else { // Bioseq-set
UnpackSeqSet(q->GetObject().SetSet(), parent, seqlist);
}
}
}
void
CShortReadFastaInputSource::x_ReadFastc(CBioseq_set& bioseq_set,
TSeqPos batch_size)
{
string id;
// tags to indicate paired sequences
CRef<CSeqdesc> seqdesc_first(new CSeqdesc);
seqdesc_first->SetUser().SetType().SetStr("Mapping");
seqdesc_first->SetUser().AddField("has_pair", eFirstSegment);
CRef<CSeqdesc> seqdesc_last(new CSeqdesc);
seqdesc_last->SetUser().SetType().SetStr("Mapping");
seqdesc_last->SetUser().AddField("has_pair", eLastSegment);
m_BasesAdded = 0;
while (m_BasesAdded < batch_size && !m_LineReader->AtEOF()) {
++(*m_LineReader);
m_Line = **m_LineReader;
// ignore empty lines
if (m_Line.empty()) {
continue;
}
// if defline
if (m_Line[0] == '>') {
id = x_ParseDefline(m_Line);
}
else {
// otherwise sequence
// make sure that a defline was read first
if (id.empty()) {
NCBI_THROW(CInputException, eInvalidInput,
(string)"Missing defline before line: " +
NStr::IntToString(m_LineReader->GetLineNumber()));
}
// find '><' that separate reads of a pair
size_t p = m_Line.find('>');
if (p == CTempString::npos || m_Line[p + 1] != '<') {
NCBI_THROW(CInputException, eInvalidInput,
(string)"FASTC parse error: Sequence separator '><'"
" was not found in line: " +
NStr::IntToString(m_LineReader->GetLineNumber()));
}
// set up reads, there are two sequences in the same line separated
char* first = (char*)m_Line.data();
char* second = (char*)m_Line.data() + p + 2;
size_t first_len = p;
size_t second_len = m_Line.length() - p - 2;
{{
CRef<CSeq_id> seqid(new CSeq_id);
seqid->Set(CSeq_id::e_Local, id + ".1");
CRef<CSeq_entry> seq_entry(new CSeq_entry);
CBioseq& bioseq = seq_entry->SetSeq();
bioseq.SetInst().SetMol(CSeq_inst::eMol_na);
bioseq.SetInst().SetRepr(CSeq_inst::eRepr_raw);
bioseq.SetId().clear();
bioseq.SetId().push_back(seqid);
bioseq.SetInst().SetLength(first_len);
first[first_len] = 0;
bioseq.SetInst().SetSeq_data().SetIupacna(CIUPACna(first));
bioseq.SetDescr().Set().push_back(seqdesc_first);
// add a sequence to the batch
bioseq_set.SetSeq_set().push_back(seq_entry);
}}
{{
CRef<CSeq_id> seqid(new CSeq_id);
seqid->Set(CSeq_id::e_Local, id + ".2");
CRef<CSeq_entry> seq_entry(new CSeq_entry);
CBioseq& bioseq = seq_entry->SetSeq();
bioseq.SetInst().SetMol(CSeq_inst::eMol_na);
bioseq.SetInst().SetRepr(CSeq_inst::eRepr_raw);
bioseq.SetId().clear();
bioseq.SetId().push_back(seqid);
bioseq.SetInst().SetLength(second_len);
second[second_len] = 0;
bioseq.SetInst().SetSeq_data().SetIupacna(CIUPACna(second));
bioseq.SetDescr().Set().push_back(seqdesc_last);
// add a sequence to the batch
bioseq_set.SetSeq_set().push_back(seq_entry);
}}
m_BasesAdded += first_len + second_len;
id.clear();
}
}
}
void
CShortReadFastaInputSource::x_ReadFastaOrFastq(CBioseq_set& bioseq_set,
TSeqPos batch_size)
{
// tags to indicate paired sequences
CRef<CSeqdesc> seqdesc_first(new CSeqdesc);
seqdesc_first->SetUser().SetType().SetStr("Mapping");
seqdesc_first->SetUser().AddField("has_pair", eFirstSegment);
CRef<CSeqdesc> seqdesc_last(new CSeqdesc);
seqdesc_last->SetUser().SetType().SetStr("Mapping");
seqdesc_last->SetUser().AddField("has_pair", eLastSegment);
m_BasesAdded = 0;
while (m_BasesAdded < batch_size && !m_LineReader->AtEOF()) {
CRef<CSeq_entry> first;
CRef<CSeq_entry> second;
switch (m_Format) {
case eFasta:
first = x_ReadFastaOneSeq(m_LineReader);
break;
case eFastq:
first = x_ReadFastqOneSeq(m_LineReader);
break;
default:
NCBI_THROW(CInputException, eInvalidInput, "Invalid input file "
"format x_ReadFastaOrFastq read either FASTA or FASTQ");
}
// if paired read the next sequence and mark a pair
if (m_IsPaired) {
switch (m_Format) {
case eFasta:
second = x_ReadFastaOneSeq(m_LineReader);
break;
case eFastq:
second = x_ReadFastqOneSeq(m_LineReader);
break;
default:
NCBI_THROW(CInputException, eInvalidInput, "Invalid input file "
"format x_ReadFastaOrFastq read either FASTA or "
"FASTQ");
}
if (first.NotEmpty()) {
if (second.NotEmpty()) {
first->SetSeq().SetDescr().Set().push_back(seqdesc_first);
}
bioseq_set.SetSeq_set().push_back(first);
}
if (second.NotEmpty()) {
if (first.NotEmpty()) {
second->SetSeq().SetDescr().Set().push_back(seqdesc_last);
}
bioseq_set.SetSeq_set().push_back(second);
}
}
else {
// otherwise just add the read sequence
if (first.NotEmpty()) {
bioseq_set.SetSeq_set().push_back(first);
}
}
}
}