void Intersect(const INT_VECT &r1, const INT_VECT &r2, INT_VECT &res) {
res.resize(0);
INT_VECT_CI ri;
for (ri = r1.begin(); ri != r1.end(); ri++) {
if (std::find(r2.begin(), r2.end(), (*ri)) != r2.end()){
res.push_back(*ri);
}
}
}
void Union(const INT_VECT &r1, const INT_VECT &r2, INT_VECT &res) {
res.resize(0);
res = r1;
INT_VECT_CI ri;
for (ri = r2.begin(); ri != r2.end(); ri++) {
if (std::find(res.begin(), res.end(), (*ri)) == res.end()){
res.push_back(*ri);
}
}
}
ROMol *prepareMol(const ROMol &mol, const FragCatParams *fparams,
MatchVectType &aToFmap) {
PRECONDITION(fparams,"");
// get a mapping of the functional groups onto the molecule
INT_VECT fgBonds;
MatchVectType aidToFid = findFuncGroupsOnMol(mol, fparams, fgBonds);
// get the core piece of molecule (i.e. the part of the molecule
// without the functional groups). This basically the part of the molecule
// that does not contain the function group bonds given by "fgBonds"
INT_VECT cBonds;
int bid, nbds = mol.getNumBonds();
for (bid = 0; bid < nbds; bid++) {
if (std::find(fgBonds.begin(), fgBonds.end(), bid) == fgBonds.end()) {
cBonds.push_back(bid);
}
}
INT_MAP_INT aIdxMap; // a map from atom id in mol to the new atoms id in coreMol
ROMol *coreMol = Subgraphs::pathToSubmol(mol, cBonds, false, aIdxMap);
// now map the functional groups on mol to coreMol using aIdxMap
MatchVectType::iterator mati;
int newID;
for (mati = aidToFid.begin(); mati != aidToFid.end(); mati++) {
newID = aIdxMap[mati->first];
aToFmap.push_back(std::pair<int, int>(newID, mati->second));
}
return coreMol;
}
void FragCatalogEntry::setDescription(const FragCatParams *params) {
PRECONDITION(params, "");
INT_INT_VECT_MAP::const_iterator fMapIt;
for (fMapIt = d_aToFmap.begin(); fMapIt != d_aToFmap.end(); fMapIt++) {
int atIdx = fMapIt->first;
INT_VECT fGroups = fMapIt->second;
std::string label = "", temp;
INT_VECT::const_iterator fGroupIdx = fGroups.begin();
const ROMol *fGroup;
for (unsigned int i = 0; i < fGroups.size() - 1; i++) {
fGroup = params->getFuncGroup(*fGroupIdx);
fGroup->getProp(common_properties::_Name, temp);
label += "(<" + temp + ">)";
fGroupIdx++;
}
fGroup = params->getFuncGroup(*fGroupIdx);
fGroup->getProp(common_properties::_Name, temp);
label += "<" + temp + ">";
dp_mol->getAtomWithIdx(atIdx)
->setProp(common_properties::_supplementalSmilesLabel, label);
}
std::string smi = MolToSmiles(*dp_mol);
// std::cerr << "----" << smi << "----" << std::endl;
d_descrip = smi;
};
void FragCatalogEntry::toStream(std::ostream &ss) const {
MolPickler::pickleMol(*dp_mol, ss);
std::int32_t tmpInt;
tmpInt = getBitId();
streamWrite(ss, tmpInt);
tmpInt = d_descrip.size();
streamWrite(ss, tmpInt);
ss.write(d_descrip.c_str(), tmpInt * sizeof(char));
tmpInt = d_order;
streamWrite(ss, tmpInt);
tmpInt = d_aToFmap.size();
streamWrite(ss, tmpInt);
for (const auto &iivmci : d_aToFmap) {
tmpInt = iivmci.first;
streamWrite(ss, tmpInt);
INT_VECT tmpVect = iivmci.second;
tmpInt = tmpVect.size();
streamWrite(ss, tmpInt);
for (INT_VECT_CI ivci = tmpVect.begin(); ivci != tmpVect.end(); ivci++) {
tmpInt = *ivci;
streamWrite(ss, tmpInt);
}
}
}
Subgraphs::DiscrimTuple FragCatalogEntry::getDiscrims() const {
Subgraphs::DiscrimTuple res;
if (this->hasProp(common_properties::Discrims)) {
this->getProp(common_properties::Discrims, res);
} else {
PATH_TYPE path;
for (unsigned int i = 0; i < dp_mol->getNumBonds(); ++i) path.push_back(i);
// create invariant additions to reflect the functional groups attached to
// the atoms
std::vector<std::uint32_t> funcGpInvars;
gboost::hash<INT_VECT> vectHasher;
for (ROMol::AtomIterator atomIt = dp_mol->beginAtoms();
atomIt != dp_mol->endAtoms(); ++atomIt) {
unsigned int aid = (*atomIt)->getIdx();
std::uint32_t invar = 0;
auto mapPos = d_aToFmap.find(aid);
if (mapPos != d_aToFmap.end()) {
INT_VECT fGroups = mapPos->second;
std::sort(fGroups.begin(), fGroups.end());
invar = vectHasher(fGroups);
}
funcGpInvars.push_back(invar);
}
res = Subgraphs::calcPathDiscriminators(*dp_mol, path, true, &funcGpInvars);
this->setProp(common_properties::Discrims, res);
}
// std::cout << "DISCRIMS: " << d_descrip << " ";
// std::cout << res.get<0>() << " " << res.get<1>() << " " << res.get<2>();
// std::cout << std::endl;
return res;
}
void Union(const VECT_INT_VECT &rings, INT_VECT &res, const INT_VECT *exclude) {
res.resize(0);
INT_VECT ring;
unsigned int id;
unsigned int nrings = static_cast<unsigned int>(rings.size());
INT_VECT_CI ri;
for (id = 0; id < nrings; id++) {
if (exclude) {
if (std::find(exclude->begin(), exclude->end(), static_cast<int>(id)) != exclude->end()) {
continue;
}
}
ring = rings[id];
for (ri = ring.begin(); ri != ring.end(); ri++) {
if (std::find(res.begin(), res.end(), (*ri)) == res.end()) {
res.push_back(*ri);
}
}
}
}
/*! \brief split the formal charge across atoms of same type if we have a conjugated system
*
* This function is called before the charge equivalization iteration is started for the
* Gasteiger charges. If any of the atom involved in conjugated system have formal charges
* set on them, this charges is equally distributed across atoms of the same type in that
* conjugated system. So for example the two nitrogens in the benzamidine system start the iteration
* with equal charges of 0.5
*/
void splitChargeConjugated(const ROMol &mol, DOUBLE_VECT &charges) {
int aix;
int natms = mol.getNumAtoms();
INT_VECT marker;
INT_VECT_CI mci;
int aax, yax;
double formal;
const Atom *at, *aat, *yat;
for (aix = 0; aix < natms; aix++) {
at = mol.getAtomWithIdx(aix);
formal = at->getFormalCharge();
//std::cout << aix << " formal charges:" << formal << "\n";
marker.resize(0);
if ((fabs(formal) > EPS_DOUBLE) && (fabs(charges[aix]) < EPS_DOUBLE) ) {
marker.push_back(aix);
ROMol::OEDGE_ITER bnd1, end1, bnd2, end2;
boost::tie(bnd1,end1) = mol.getAtomBonds(at);
while (bnd1 != end1) {
if (mol[*bnd1]->getIsConjugated()) {
aax = mol[*bnd1]->getOtherAtomIdx(aix);
aat = mol.getAtomWithIdx(aax);
boost::tie(bnd2,end2) = mol.getAtomBonds(aat);
while (bnd2 != end2) {
if ((*bnd1) != (*bnd2)) {
if (mol[*bnd2]->getIsConjugated()) {
yax = mol[*bnd2]->getOtherAtomIdx(aax);
yat = mol.getAtomWithIdx(yax);
if (at->getAtomicNum() == yat->getAtomicNum()) {
formal += yat->getFormalCharge();
marker.push_back(yax);
}
}
}
bnd2++;
}
}
bnd1++;
}
for (mci = marker.begin(); mci != marker.end(); mci++) {
charges[*mci] = (formal/marker.size());
}
}
}
/*
for (aix = 0; aix < natms; aix++) {
std::cout << "In splitter: " << " charges:" << charges[aix] << "\n";
}*/
}
bool checkFused(const INT_VECT &rids, INT_INT_VECT_MAP &ringNeighs) {
INT_INT_VECT_MAP_CI nci;
int nrings = rdcast<int>(ringNeighs.size());
boost::dynamic_bitset<> done(nrings);
int rid;
INT_VECT fused;
// mark all rings in the system other than those in rids as done
for (nci = ringNeighs.begin(); nci != ringNeighs.end(); nci++) {
rid = (*nci).first;
if (std::find(rids.begin(), rids.end(), rid) == rids.end()) {
done[rid] = 1;
}
}
// then pick a fused system from the remaining (i.e. rids)
// If the rings in rids are fused we should get back all of them
// in fused
// if we get a smaller number in fused then rids are not fused
pickFusedRings(rids.front(), ringNeighs, fused, done);
CHECK_INVARIANT(fused.size() <= rids.size(), "");
return (fused.size() == rids.size());
}
void dfsBuildStack(ROMol &mol,int atomIdx,int inBondIdx,
std::vector<AtomColors> &colors,
VECT_INT_VECT &cycles,
const UINT_VECT &ranks,
INT_VECT &cyclesAvailable,
MolStack &molStack,
INT_VECT &atomOrders,
INT_VECT &bondVisitOrders,
VECT_INT_VECT &atomRingClosures,
std::vector<INT_LIST> &atomTraversalBondOrder,
const boost::dynamic_bitset<> *bondsInPlay,
const std::vector<std::string> *bondSymbols
){
#if 0
std::cerr<<"traverse from atom: "<<atomIdx<<" via bond "<<inBondIdx<<" num cycles available: "
<<std::count(cyclesAvailable.begin(),cyclesAvailable.end(),1)<<std::endl;
#endif
Atom *atom = mol.getAtomWithIdx(atomIdx);
INT_LIST directTravList,cycleEndList;
boost::dynamic_bitset<> seenFromHere(mol.getNumAtoms());
seenFromHere.set(atomIdx);
molStack.push_back(MolStackElem(atom));
atomOrders[atom->getIdx()] = molStack.size();
colors[atomIdx] = GREY_NODE;
INT_LIST travList;
if(inBondIdx>=0) travList.push_back(inBondIdx);
// ---------------------
//
// Add any ring closures
//
// ---------------------
if(atomRingClosures[atomIdx].size()){
std::vector<unsigned int> ringsClosed;
BOOST_FOREACH(int bIdx,atomRingClosures[atomIdx]){
travList.push_back(bIdx);
Bond *bond = mol.getBondWithIdx(bIdx);
seenFromHere.set(bond->getOtherAtomIdx(atomIdx));
unsigned int ringIdx;
if(bond->getPropIfPresent(common_properties::_TraversalRingClosureBond, ringIdx)){
// this is end of the ring closure
// we can just pull the ring index from the bond itself:
molStack.push_back(MolStackElem(bond,atomIdx));
bondVisitOrders[bIdx]=molStack.size();
molStack.push_back(MolStackElem(ringIdx));
// don't make the ring digit immediately available again: we don't want to have the same
// ring digit opening and closing rings on an atom.
ringsClosed.push_back(ringIdx-1);
} else {
// this is the beginning of the ring closure, we need to come up with a ring index:
INT_VECT::const_iterator cAIt=std::find(cyclesAvailable.begin(),
cyclesAvailable.end(),1);
if(cAIt==cyclesAvailable.end()){
throw ValueErrorException("Too many rings open at once. SMILES cannot be generated.");
}
unsigned int lowestRingIdx = cAIt-cyclesAvailable.begin();
cyclesAvailable[lowestRingIdx] = 0;
++lowestRingIdx;
bond->setProp(common_properties::_TraversalRingClosureBond,lowestRingIdx);
molStack.push_back(MolStackElem(lowestRingIdx));
}
}
MatchVectType findFuncGroupsOnMol(const ROMol &mol,
const FragCatParams *params,
INT_VECT &fgBonds) {
PRECONDITION(params,"bad params");
fgBonds.clear();
std::pair<int, int> amat;
MatchVectType aidFgrps;
std::vector<MatchVectType> fgpMatches;
std::vector<MatchVectType>::const_iterator mati;
MatchVectType::const_iterator mi;
int aid;
//const ROMol *fgrp;
INT_VECT_CI bi;
aidFgrps.clear();
int fid = 0;
const MOL_SPTR_VECT &fgrps = params->getFuncGroups();
MOL_SPTR_VECT::const_iterator fgci;
for (fgci = fgrps.begin(); fgci != fgrps.end(); fgci++) {
const ROMol *fgrp = fgci->get();
std::string fname;
(*fgci)->getProp(common_properties::_Name, fname);
//std::cout << "Groups number: " << fname << "\n";
//(*fgci)->debugMol(std::cout);
//mol->debugMol(std::cout);
// match this functional group onto the molecule
SubstructMatch(mol, *fgrp, fgpMatches);
// loop over all the matches we get for this fgroup
for (mati = fgpMatches.begin(); mati != fgpMatches.end(); mati++) {
//FIX: we will assume that the first atom in fgrp is always the connection
// atom
amat = mati->front();
aid = amat.second; //FIX: is this correct - the second entry in the pair is the atom ID from mol
// grab the list of atom Ids from mol that match the functional group
INT_VECT bondIds, maids;
for (mi = mati->begin(); mi != mati->end(); mi++) {
maids.push_back(mi->second);
}
// create a list of bond IDs from these atom ID
// these are the bond in mol that are part of portion that matches the
// functional group
bondIds = Subgraphs::bondListFromAtomList(mol, maids);
// now check if all these bonds have been covered as part of larger
// functional group that was dealt with earlier
// FIX: obviously we assume here that the function groups in params
// come in decreasing order of size.
bool allDone = true;
for (bi = bondIds.begin(); bi != bondIds.end(); bi++) {
if (std::find(fgBonds.begin(), fgBonds.end(), (*bi)) == fgBonds.end()) {
allDone = false;
fgBonds.push_back(*bi);
}
}
if (!allDone) {
// this functional group mapping onto mol is not part of a larger func
// group mapping so record it
aidFgrps.push_back(std::pair<int, int>(aid, fid));
}
}
fid++;
}
return aidFgrps;
}
bool kekulizeWorker(RWMol &mol, const INT_VECT &allAtms,
boost::dynamic_bitset<> dBndCands,
boost::dynamic_bitset<> dBndAdds, INT_VECT done,
unsigned int maxBackTracks) {
INT_DEQUE astack;
INT_INT_DEQ_MAP options;
int lastOpt = -1;
boost::dynamic_bitset<> localBondsAdded(mol.getNumBonds());
// ok the algorithm goes something like this
// - start with an atom that has been marked aromatic before
// - check if it can have a double bond
// - add its neighbors to the stack
// - check if one of its neighbors can also have a double bond
// - if yes add a double bond.
// - if multiple neighbors can have double bonds - add them to a
// options stack we may have to retrace out path if we chose the
// wrong neighbor to add the double bond
// - if double bond added update the candidates for double bond
// - move to the next atom on the stack and repeat the process
// - if an atom that can have multiple a double bond has no
// neighbors that can take double bond - we made a mistake
// earlier by picking a wrong candidate for double bond
// - in this case back track to where we made the mistake
int curr;
INT_DEQUE btmoves;
unsigned int numBT = 0; // number of back tracks so far
while ((done.size() < allAtms.size()) || (astack.size() > 0)) {
// pick a curr atom to work with
if (astack.size() > 0) {
curr = astack.front();
astack.pop_front();
} else {
for (int allAtm : allAtms) {
if (std::find(done.begin(), done.end(), allAtm) == done.end()) {
curr = allAtm;
break;
}
}
}
done.push_back(curr);
// loop over the neighbors if we can add double bonds or
// simply push them onto the stack
INT_DEQUE opts;
bool cCand = false;
if (dBndCands[curr]) {
cCand = true;
}
int ncnd;
// if we are here because of backtracking
if (options.find(curr) != options.end()) {
opts = options[curr];
CHECK_INVARIANT(opts.size() > 0, "");
} else {
RWMol::ADJ_ITER nbrIdx, endNbrs;
boost::tie(nbrIdx, endNbrs) =
mol.getAtomNeighbors(mol.getAtomWithIdx(curr));
while (nbrIdx != endNbrs) {
// ignore if the neighbor has already been dealt with before
if (std::find(done.begin(), done.end(), static_cast<int>(*nbrIdx)) !=
done.end()) {
++nbrIdx;
continue;
}
// ignore if the neighbor is not part of the fused system
if (std::find(allAtms.begin(), allAtms.end(),
static_cast<int>(*nbrIdx)) == allAtms.end()) {
++nbrIdx;
continue;
}
// if the neighbor is not on the stack add it
if (std::find(astack.begin(), astack.end(),
static_cast<int>(*nbrIdx)) == astack.end()) {
astack.push_back(rdcast<int>(*nbrIdx));
}
// check if the neighbor is also a candidate for a double bond
// the refinement that we'll make to the candidate check we've already
// done is to make sure that the bond is either flagged as aromatic
// or involves a dummy atom. This was Issue 3525076.
// This fix is not really 100% of the way there: a situation like
// that for Issue 3525076 but involving a dummy atom in the cage
// could lead to the same failure. The full fix would require
// a fairly detailed analysis of all bonds in the molecule to determine
// which of them is eligible to be converted.
if (cCand && dBndCands[*nbrIdx] &&
(mol.getBondBetweenAtoms(curr, *nbrIdx)->getIsAromatic() ||
mol.getAtomWithIdx(curr)->getAtomicNum() == 0 ||
mol.getAtomWithIdx(*nbrIdx)->getAtomicNum() == 0)) {
opts.push_back(rdcast<int>(*nbrIdx));
} // end of curr atoms can have a double bond
++nbrIdx;
} // end of looping over neighbors
}
// now add a double bond from current to one of the neighbors if we can
if (cCand) {
if (opts.size() > 0) {
ncnd = opts.front();
opts.pop_front();
Bond *bnd = mol.getBondBetweenAtoms(curr, ncnd);
bnd->setBondType(Bond::DOUBLE);
// remove current and the neighbor from the dBndCands list
dBndCands[curr] = 0;
dBndCands[ncnd] = 0;
// add them to the list of bonds to which have been made double
dBndAdds[bnd->getIdx()] = 1;
localBondsAdded[bnd->getIdx()] = 1;
// if this is an atom we previously visted and picked we
// simply tried a different option now, overwrite the options
// stored for this atoms
if (options.find(curr) != options.end()) {
if (opts.size() == 0) {
options.erase(curr);
btmoves.pop_back();
if (btmoves.size() > 0) {
lastOpt = btmoves.back();
} else {
lastOpt = -1;
}
} else {
options[curr] = opts;
}
} else {
// this is new atoms we are trying and have other
// neighbors as options to add double bond store this to
// the options stack, we may have made a mistake in
// which one we chose and have to return here
if (opts.size() > 0) {
lastOpt = curr;
btmoves.push_back(lastOpt);
options[curr] = opts;
}
}
} // end of adding a double bond
else {
// we have an atom that should be getting a double bond
// but none of the neighbors can take one. Most likely
// because of a wrong choice earlier so back track
if ((lastOpt >= 0) && (numBT < maxBackTracks)) {
// std::cerr << "PRE BACKTRACK" << std::endl;
// mol.debugMol(std::cerr);
backTrack(mol, options, lastOpt, done, astack, dBndCands, dBndAdds);
// std::cerr << "POST BACKTRACK" << std::endl;
// mol.debugMol(std::cerr);
numBT++;
} else {
// undo any remaining changes we made while here
// this was github #962
for (unsigned int bidx = 0; bidx < mol.getNumBonds(); ++bidx) {
if (localBondsAdded[bidx]) {
mol.getBondWithIdx(bidx)->setBondType(Bond::SINGLE);
}
}
return false;
}
} // end of else try to backtrack
} // end of curr atom atom being a cand for double bond
} // end of while we are not done with all atoms
return true;
}
void canonicalDFSTraversal(ROMol &mol,int atomIdx,int inBondIdx,
std::vector<AtomColors> &colors,
VECT_INT_VECT &cycles,
INT_VECT &ranks,
INT_VECT &cyclesAvailable,
MolStack &molStack,
INT_VECT &atomOrders,
INT_VECT &bondVisitOrders,
VECT_INT_VECT &atomRingClosures,
std::vector<INT_LIST> &atomTraversalBondOrder,
const boost::dynamic_bitset<> *bondsInPlay,
const std::vector<std::string> *bondSymbols
){
PRECONDITION(colors.size()>=mol.getNumAtoms(),"vector too small");
PRECONDITION(ranks.size()>=mol.getNumAtoms(),"vector too small");
PRECONDITION(atomOrders.size()>=mol.getNumAtoms(),"vector too small");
PRECONDITION(bondVisitOrders.size()>=mol.getNumBonds(),"vector too small");
PRECONDITION(atomRingClosures.size()>=mol.getNumAtoms(),"vector too small");
PRECONDITION(atomTraversalBondOrder.size()>=mol.getNumAtoms(),"vector too small");
PRECONDITION(!bondsInPlay || bondsInPlay->size()>=mol.getNumBonds(),"bondsInPlay too small");
PRECONDITION(!bondSymbols || bondSymbols->size()>=mol.getNumBonds(),"bondSymbols too small");
int nAttached=0;
Atom *atom = mol.getAtomWithIdx(atomIdx);
INT_LIST directTravList,cycleEndList;
molStack.push_back(MolStackElem(atom));
atomOrders[atom->getIdx()] = molStack.size();
colors[atomIdx] = GREY_NODE;
// ---------------------
//
// Build the list of possible destinations from here
//
// ---------------------
std::vector< PossibleType > possibles;
possibles.resize(0);
ROMol::OBOND_ITER_PAIR bondsPair = mol.getAtomBonds(atom);
possibles.reserve(bondsPair.second-bondsPair.first);
while(bondsPair.first != bondsPair.second){
BOND_SPTR theBond = mol[*(bondsPair.first)];
bondsPair.first++;
if(bondsInPlay && !(*bondsInPlay)[theBond->getIdx()]) continue;
if(inBondIdx<0 || theBond->getIdx() != static_cast<unsigned int>(inBondIdx)){
int otherIdx = theBond->getOtherAtomIdx(atomIdx);
long rank=ranks[otherIdx];
// ---------------------
//
// things are a bit more complicated if we are sitting on a
// ring atom we would like to traverse first to the
// ring-closure atoms, then to atoms outside the ring first,
// then to atoms in the ring that haven't already been visited
// (non-ring-closure atoms).
//
// Here's how the black magic works:
// - non-ring atom neighbors have their original ranks
// - ring atom neighbors have this added to their ranks:
// (Bond::OTHER - bondOrder)*MAX_NATOMS*MAX_NATOMS
// - ring-closure neighbors lose a factor of:
// (Bond::OTHER+1)*MAX_NATOMS*MAX_NATOMS
//
// This tactic biases us to traverse to non-ring neighbors first,
// original ordering if bond orders are all equal... crafty, neh?
//
// ---------------------
if( colors[otherIdx] == GREY_NODE ) {
rank -= static_cast<int>(Bond::OTHER+1) *
MAX_NATOMS*MAX_NATOMS;
if(!bondSymbols){
rank += static_cast<int>(Bond::OTHER - theBond->getBondType()) *
MAX_NATOMS;
} else {
const std::string &symb=(*bondSymbols)[theBond->getIdx()];
boost::uint32_t hsh=gboost::hash_range(symb.begin(),symb.end());
rank += (hsh%MAX_NATOMS) * MAX_NATOMS;
}
} else if( theBond->getOwningMol().getRingInfo()->numBondRings(theBond->getIdx()) ){
if(!bondSymbols){
rank += static_cast<int>(Bond::OTHER - theBond->getBondType()) *
MAX_NATOMS*MAX_NATOMS;
} else {
const std::string &symb=(*bondSymbols)[theBond->getIdx()];
boost::uint32_t hsh=gboost::hash_range(symb.begin(),symb.end());
rank += (hsh%MAX_NATOMS)*MAX_NATOMS*MAX_NATOMS;
}
}
possibles.push_back(PossibleType(rank,otherIdx,theBond.get()));
}
}
// ---------------------
//
// Sort on ranks
//
// ---------------------
std::sort(possibles.begin(),possibles.end(),_possibleComp);
// ---------------------
//
// Now work the children
//
// ---------------------
std::vector<MolStack> subStacks;
for(std::vector<PossibleType>::iterator possiblesIt=possibles.begin();
possiblesIt!=possibles.end();
possiblesIt++){
MolStack subStack;
#if 0
int possibleIdx = possiblesIt->second.first;
Bond *bond = possiblesIt->second.second;
#endif
int possibleIdx = possiblesIt->get<1>();
Bond *bond = possiblesIt->get<2>();
Atom *otherAtom=mol.getAtomWithIdx(possibleIdx);
unsigned int lowestRingIdx;
INT_VECT::const_iterator cAIt;
switch(colors[possibleIdx]){
case WHITE_NODE:
// -----
// we haven't seen this node at all before
// -----
// it might have some residual data from earlier calls, clean that up:
if(otherAtom->hasProp("_TraversalBondIndexOrder")){
otherAtom->clearProp("_TraversalBondIndexOrder");
}
directTravList.push_back(bond->getIdx());
subStack.push_back(MolStackElem(bond,atomIdx));
canonicalDFSTraversal(mol,possibleIdx,bond->getIdx(),colors,
cycles,ranks,cyclesAvailable,subStack,
atomOrders,bondVisitOrders,atomRingClosures,atomTraversalBondOrder,
bondsInPlay,bondSymbols);
subStacks.push_back(subStack);
nAttached += 1;
break;
case GREY_NODE:
// -----
// we've seen this, but haven't finished it (we're finishing a ring)
// -----
cycleEndList.push_back(bond->getIdx());
cAIt=std::find(cyclesAvailable.begin(),
cyclesAvailable.end(),1);
if(cAIt==cyclesAvailable.end()){
throw ValueErrorException("Too many rings open at once. SMILES cannot be generated.");
}
lowestRingIdx = cAIt-cyclesAvailable.begin();
cyclesAvailable[lowestRingIdx] = 0;
cycles[possibleIdx].push_back(lowestRingIdx);
++lowestRingIdx;
bond->setProp("_TraversalRingClosureBond",lowestRingIdx);
molStack.push_back(MolStackElem(bond,
atom->getIdx()));
molStack.push_back(MolStackElem(lowestRingIdx));
// we need to add this bond (which closes the ring) to the traversal list for the
// other atom as well:
atomTraversalBondOrder[otherAtom->getIdx()].push_back(bond->getIdx());
atomRingClosures[otherAtom->getIdx()].push_back(bond->getIdx());
break;
default:
// -----
// this node has been finished. don't do anything.
// -----
break;
}
}
INT_VECT &ringClosures=atomRingClosures[atom->getIdx()];
CHECK_INVARIANT(ringClosures.size()==cycles[atomIdx].size(),
"ring closure mismatch");
for(unsigned int i=0;i<ringClosures.size();i++){
int ringIdx=cycles[atomIdx][i];
ringIdx += 1;
molStack.push_back(MolStackElem(ringIdx));
}
cycles[atomIdx].resize(0);
MolStack::const_iterator ciMS;
for(int i=0;i<nAttached;i++){
if(i<nAttached-1){
int branchIdx=0;
if(subStacks[i].begin()->type==MOL_STACK_ATOM){
branchIdx=subStacks[i].begin()->obj.atom->getIdx();
} else if(subStacks[i].begin()->type==MOL_STACK_BOND){
branchIdx=-1*subStacks[i].begin()->obj.bond->getIdx();
} else {
ASSERT_INVARIANT(0,"branch started with something other than an atom or bond");
}
molStack.push_back(MolStackElem("(",branchIdx));
for(ciMS=subStacks[i].begin();ciMS!=subStacks[i].end();ciMS++){
molStack.push_back(*ciMS);
switch(ciMS->type){
case MOL_STACK_ATOM:
atomOrders[ciMS->obj.atom->getIdx()] = molStack.size();
break;
case MOL_STACK_BOND:
bondVisitOrders[ciMS->obj.bond->getIdx()] = molStack.size();
break;
default:
break;
}
}
molStack.push_back(MolStackElem(")",branchIdx));
} else {
for(ciMS=subStacks[i].begin();ciMS!=subStacks[i].end();ciMS++){
molStack.push_back(*ciMS);
switch(ciMS->type){
case MOL_STACK_ATOM:
atomOrders[ciMS->obj.atom->getIdx()] = molStack.size();
break;
case MOL_STACK_BOND:
bondVisitOrders[ciMS->obj.bond->getIdx()] = molStack.size();
break;
default:
break;
}
}
}
}
//std::cerr<<"*****>>>>>> Traversal results for atom: "<<atom->getIdx()<<"> ";
INT_LIST travList;
// first push on the incoming bond:
if(inBondIdx >= 0){
//std::cerr<<" "<<inBondIdx;
travList.push_back(inBondIdx);
}
// ... ring closures that end here:
for(INT_LIST_CI ilci=cycleEndList.begin();ilci!=cycleEndList.end();++ilci){
//std::cerr<<" ["<<*ilci<<"]";
travList.push_back(*ilci);
}
// ... ring closures that start here:
// if(atom->hasProp("_TraversalBondIndexOrder")){
// INT_LIST indirectTravList;
// atom->getProp("_TraversalBondIndexOrder",indirectTravList);
// for(INT_LIST_CI ilci=indirectTravList.begin();ilci!=indirectTravList.end();++ilci){
// //std::cerr<<" ("<<*ilci<<")";
// travList.push_back(*ilci);
// }
// }
BOOST_FOREACH(int ili,atomTraversalBondOrder[atom->getIdx()]){
travList.push_back(ili);
}