// Update the overlap block list with a righthand extension to b, removing ranges that become invalid
void OverlapAlgorithm::updateOverlapBlockRangesRight(const BWT* pBWT, const BWT* pRevBWT,
OverlapBlockList& obList, char canonical_base) const
{
OverlapBlockList::iterator iter = obList.begin();
while(iter != obList.end())
{
char relative_base = iter->flags.isQueryComp() ? complement(canonical_base) : canonical_base;
BWTAlgorithms::updateBothR(iter->ranges, relative_base, iter->getExtensionBWT(pBWT, pRevBWT));
// remove the block from the list if its no longer valid
if(!iter->ranges.isValid())
{
iter = obList.erase(iter);
}
else
{
// Add the base to the extension history
int currExtension = iter->forwardHistory.size();
iter->forwardHistory.add(currExtension, canonical_base);
++iter;
}
}
}
// Extend all the blocks in activeList by one base to the right
// Move all right-terminal blocks to the termainl list. If a block
// is terminal and potentially contained by another block, add it to
// containedList
void OverlapAlgorithm::extendActiveBlocksRight(const BWT* pBWT, const BWT* pRevBWT,
OverlapBlockList& activeList,
OverlapBlockList& terminalList,
OverlapBlockList& /*containedList*/) const
{
OverlapBlockList::iterator iter = activeList.begin();
OverlapBlockList::iterator next;
while(iter != activeList.end())
{
next = iter;
++next;
// Check if block is terminal
AlphaCount64 ext_count = iter->getCanonicalExtCount(pBWT, pRevBWT);
if(ext_count.get('$') > 0)
{
// Only consider this block to be terminal irreducible if it has at least one extension
// or else it is a substring block
if(iter->forwardHistory.size() > 0)
{
OverlapBlock branched = *iter;
BWTAlgorithms::updateBothR(branched.ranges, '$', branched.getExtensionBWT(pBWT, pRevBWT));
terminalList.push_back(branched);
#ifdef DEBUGOVERLAP_2
std::cout << "Block of length " << iter->overlapLen << " moved to terminal\n";
#endif
}
}
int curr_extension = iter->forwardHistory.size();
// Perform the right extensions
// Best case, there is only a single extension character
// Handle this case specially so we don't need to copy the potentially
// large OverlapBlock structure and its full history
if(ext_count.hasUniqueDNAChar())
{
// Get the extension character with respect to the queried sequence
char canonical_base = ext_count.getUniqueDNAChar();
// Flip the base into the frame of reference for the block
char block_base = iter->flags.isQueryComp() ? complement(canonical_base) : canonical_base;
// Update the block using the base in its frame of reference
BWTAlgorithms::updateBothR(iter->ranges, block_base, iter->getExtensionBWT(pBWT, pRevBWT));
// Add the base to the history in the frame of reference of the query read
// This is so the history is consistent when comparing between blocks from different strands
iter->forwardHistory.add(curr_extension, canonical_base);
}
else
{
for(size_t idx = 0; idx < DNA_ALPHABET_SIZE; ++idx)
{
char canonical_base = ALPHABET[idx];
char block_base = iter->flags.isQueryComp() ? complement(canonical_base) : canonical_base;
if(ext_count.get(canonical_base) == 0)
continue;
// Branch the sequence. This involves copying the entire history which can be large
// if the input sequences are very long. This could be avoided by using the SearchHistoyNode/Link
// structure but branches are infrequent enough to not have a large impact
OverlapBlock branched = *iter;
BWTAlgorithms::updateBothR(branched.ranges, block_base, branched.getExtensionBWT(pBWT, pRevBWT));
assert(branched.ranges.isValid());
// Add the base in the canonical frame
branched.forwardHistory.add(curr_extension, canonical_base);
// Insert the new block after the iterator
activeList.insert(iter, branched);
}
// Remove the original block, which has been superceded by the branches
activeList.erase(iter);
}
iter = next; // this skips the newly-inserted blocks
}
}