// REVIEW: the poolSize can be pulled from the numeric array
RDKit::INT_VECT MaxMinPicks(MaxMinPicker *picker, python::object distMat,
int poolSize, int pickSize,
python::object firstPicks, int seed) {
if (pickSize >= poolSize) {
throw ValueErrorException("pickSize must be less than poolSize");
}
if (!PyArray_Check(distMat.ptr())) {
throw ValueErrorException("distance mat argument must be a numpy matrix");
}
PyArrayObject *copy;
copy = (PyArrayObject *)PyArray_ContiguousFromObject(distMat.ptr(),
PyArray_DOUBLE, 1, 1);
double *dMat = (double *)copy->data;
RDKit::INT_VECT firstPickVect;
for (unsigned int i = 0;
i < python::extract<unsigned int>(firstPicks.attr("__len__")()); ++i) {
firstPickVect.push_back(python::extract<int>(firstPicks[i]));
}
RDKit::INT_VECT res =
picker->pick(dMat, poolSize, pickSize, firstPickVect, seed);
Py_DECREF(copy);
return res;
}
RDKit::INT_VECT HierarchicalClusterPicker::pick(const double *distMat,
unsigned int poolSize,
unsigned int pickSize) const {
PRECONDITION(distMat,"bad distance matrix");
RDKit::VECT_INT_VECT clusters = this->cluster(distMat, poolSize, pickSize);
CHECK_INVARIANT(clusters.size() == pickSize, "");
// the last step: find a representative element from each of the
// remaining clusters
RDKit::INT_VECT picks;
for (unsigned int i = 0; i < pickSize; i++) {
int pick;
double minSumD2 = RDKit::MAX_DOUBLE;
for (RDKit::INT_VECT_CI cxi1 = clusters[i].begin();
cxi1 != clusters[i].end(); ++cxi1 ) {
int curPick = (*cxi1);
double d2sum = 0.0;
for (RDKit::INT_VECT_CI cxi2 = clusters[i].begin();
cxi2 != clusters[i].end(); ++cxi2) {
if (cxi1 == cxi2) {
continue;
}
double d = getDistFromLTM(distMat, curPick, (*cxi2));
d2sum += (d*d);
}
if (d2sum < minSumD2) {
pick = curPick;
minSumD2 = d2sum;
}
}
picks.push_back(pick);
}
return picks;
}
void SetMaskBits(InfoBitRanker *ranker, python::object maskBits) {
RDKit::INT_VECT cList;
PySequenceHolder<int> bList(maskBits);
cList.reserve(bList.size());
for (unsigned int i = 0; i < bList.size(); i++) {
cList.push_back(bList[i]);
}
ranker->setMaskBits(cList);
}
void setBitList(BitCorrMatGenerator *cmGen, python::object bitList) {
PySequenceHolder<int> blist(bitList);
unsigned int nb = blist.size();
RDKit::INT_VECT res;
res.reserve(nb);
for (unsigned int i = 0; i < nb; i++) {
res.push_back(blist[i]);
}
cmGen->setBitIdList(res);
}
RDKit::INT_VECT LazyMaxMinPicks(MaxMinPicker *picker, python::object distFunc,
int poolSize, int pickSize,
python::object firstPicks, int seed,
bool useCache) {
RDKit::INT_VECT firstPickVect;
for (unsigned int i = 0;
i < python::extract<unsigned int>(firstPicks.attr("__len__")()); ++i) {
firstPickVect.push_back(python::extract<int>(firstPicks[i]));
}
RDKit::INT_VECT res;
pyobjFunctor functor(distFunc, useCache);
res = picker->lazyPick(functor, poolSize, pickSize, firstPickVect, seed);
return res;
}
RDKit::INT_VECT LazyVectorMaxMinPicks(MaxMinPicker *picker,
python::object objs,
int poolSize,
int pickSize,
python::object firstPicks,
int seed,
DistanceMethod method
) {
pyBVFunctor functor(objs,method);
RDKit::INT_VECT firstPickVect;
for(unsigned int i=0;i<python::extract<unsigned int>(firstPicks.attr("__len__")());++i){
firstPickVect.push_back(python::extract<int>(firstPicks[i]));
}
RDKit::INT_VECT res=picker->lazyPick(functor, poolSize, pickSize,firstPickVect,seed);
return res;
}
RDKit::INT_VECT LazyVectorMaxMinPicks(MaxMinPicker *picker, python::object objs,
int poolSize, int pickSize,
python::object firstPicks, int seed,
bool useCache) {
std::vector<const ExplicitBitVect *> bvs(poolSize);
for (int i = 0; i < poolSize; ++i) {
bvs[i] = python::extract<const ExplicitBitVect *>(objs[i]);
}
pyBVFunctor<ExplicitBitVect> functor(bvs, TANIMOTO, useCache);
RDKit::INT_VECT firstPickVect;
for (unsigned int i = 0;
i < python::extract<unsigned int>(firstPicks.attr("__len__")()); ++i) {
firstPickVect.push_back(python::extract<int>(firstPicks[i]));
}
RDKit::INT_VECT res =
picker->lazyPick(functor, poolSize, pickSize, firstPickVect, seed);
return res;
}
RDKit::VECT_INT_VECT HierarchicalClusterPicker::cluster(const double *distMat,
unsigned int poolSize,
unsigned int pickSize) const {
PRECONDITION(distMat, "Invalid Distance Matrix");
PRECONDITION((poolSize >= pickSize),
"pickSize cannot be larger than the poolSize");
// Do the clustering
long int method = (long int)d_method;
long int len = poolSize*(poolSize-1);
long int *ia = (long int *)calloc(poolSize, sizeof(long int));
long int *ib = (long int *)calloc(poolSize, sizeof(long int));
real *crit = (real *)calloc(poolSize,sizeof(real));
CHECK_INVARIANT(ia,"failed to allocate memory");
CHECK_INVARIANT(ib,"failed to allocate memory");
CHECK_INVARIANT(crit,"failed to allocate memory");
long int poolSize2=static_cast<long int>(poolSize);
distdriver_(&poolSize2, // number of items in the pool
&len, // number of entries in the distance matrix
(real *)distMat, // distance matrix
&method, // the clustering method (ward, slink etc.)
ia, // int vector with clustering history
ib, // one more clustering history matrix
crit // I believe this is a vector the difference in heights of two clusters
);
// we have the clusters now merge then until the number of clusters is same
// as the number of picks we need
// before we do that a bit of explanation on the vectors "ia" and "ib"
// - We with each item in the pool as an individual cluster
// - then we use the vectors ia and ib to merge them.
// ia and ib provides the ids of the clusters that need to be merged
// it is assumed that when a cluster ia[j] is merged with ib[j]
// ia[j] is replaced by the new cluster in the cluster list
//
RDKit::VECT_INT_VECT clusters;
for (unsigned int i = 0; i < poolSize; i++) {
RDKit::INT_VECT cls;
cls.push_back(i);
clusters.push_back(cls);
}
// do the merging, each round of of this loop eleminates one cluster
RDKit::INT_VECT removed;
for (unsigned int i = 0; i < (poolSize - pickSize); i++) {
int cx1 = ia[i] - 1;
int cx2 = ib[i] - 1;
// add the items from cluster cx2 to cx1
// REVIEW: merge function???
for (RDKit::INT_VECT_CI cx2i = clusters[cx2].begin(); cx2i != clusters[cx2].end(); cx2i++) {
clusters[cx1].push_back(*cx2i);
}
// mark the second cluster as removed
removed.push_back(cx2);
}
free(ia);
free(ib);
free(crit);
// sort removed so that looping will be easier later
std::sort(removed.begin(), removed.end());
//some error checking here, uniqueify removed and the vector should not changed
// REVIEW can we put this inside a #ifdef DEBUG?
RDKit::INT_VECT_CI nEnd = std::unique(removed.begin(), removed.end());
CHECK_INVARIANT(nEnd == removed.end(), "Somehow there are duplicates in the list of removed clusters");
RDKit::VECT_INT_VECT res;
unsigned int j = 0;
for (unsigned int i = 0; i < poolSize; i++) {
if (static_cast<int>(i) == removed[j]) {
j++;
continue;
}
res.push_back(clusters[i]);
}
return res;
}