vec CoSaMP(const mat & Phi, const vec & u, int K, int max_iter, double tol1, int D){
assert(K<= 2*D);
assert(K>=1);
assert(Phi.rows() == Phi.cols());
assert(Phi.rows() == D);
assert(u.size() == D);
vec Sest = zeros(D);
vec utrue = Sest;
vec v = u;
int t=1;
ivec T2;
while (t<max_iter){
ivec z = sort_index(fabs(Phi.transpose() * v));
z = reverse(z);
ivec Omega = head(z,2*K);
ivec T=sort_union(Omega,T2);
mat phit=get_cols(Phi, T);
vec b;
bool ret = backslash(phit, u, b);
assert(ret);
ret = false;//avoid warning
b= fabs(b);
ivec z3 = sort_index(b);
z3 = reverse(z3);
Sest=zeros(D);
for (int i=0; i< K; i++)
set_val(Sest, z3[i], b[z3[i]]);
ivec z2 = sort_index(fabs(Sest));
z2 = reverse(z2);
T2 = head(z2,K-1);
v=u-Phi*Sest;
double n2 = max(fabs(v));
if (n2 < tol1)
break;
t++;
}
return Sest;
}
RelationSpec* sortmerge_join(RelationSpec* rSpec, RelationSpec* sSpec,
const string &mainJoinAttr, const vector<string> &joinAttrOrder,
size_t& recordCount, bool saveResult, bool printProcess) {
// sort relations based on join attributes
int roffset = rSpec->get_attr_idx(mainJoinAttr);
int soffset = sSpec->get_attr_idx(mainJoinAttr);
assert(roffset != -1);
assert(soffset != -1);
sort(rSpec->memDB.begin(), rSpec->memDB.end(), sorter(roffset));
sort(sSpec->memDB.begin(), sSpec->memDB.end(), sorter(soffset));
// Check whether there are other attributes (beside the mainJoinAttr) need to be joined at the same time
vector<string> commonAttrs = sort_intersect(rSpec->attrNames,
sSpec->attrNames);
assert(commonAttrs.size() > 0);
vector<string> otherAttrsToJoin;
if (commonAttrs.size() == 1) {
assert(commonAttrs.at(0) == mainJoinAttr);
} else {
// cerr << "Additional attributes: ";
// make sure the mainJoinAttr is in the commonAttrs
int mainIdx = find_offset(commonAttrs, mainJoinAttr);
assert(mainIdx != -1);
// get other attributes need to be joined
for (int i = 0; i != commonAttrs.size(); i++) {
string curCommonAttr = commonAttrs.at(i);
if (i == mainIdx)
continue;
if (find_offset(joinAttrOrder, curCommonAttr) != -1) {
// cerr << curCommonAttr << " ";
otherAttrsToJoin.push_back(curCommonAttr);
}
}
// cerr << ". ";
}
int otherSize = otherAttrsToJoin.size();
int *rOffs = new int[otherSize];
int *sOffs = new int[otherSize];
for (int i = 0; i != otherAttrsToJoin.size(); i++) {
string curAttr = otherAttrsToJoin.at(i);
rOffs[i] = rSpec->get_attr_idx(curAttr);
sOffs[i] = sSpec->get_attr_idx(curAttr);
}
// Create the joint relation specification file
// The attributes of the joined relation is a distinct union of the attributes of the two input relations
vector<string> nAttrNames = sort_union(rSpec->attrNames, sSpec->attrNames);
int nAttrSize = nAttrNames.size();
string nRelName = rSpec->relName + sSpec->relName;
RelationSpec* nSpec = new RelationSpec(nRelName, "", nAttrNames, false);
int* rAttrIdx = new int[nAttrSize];
int* sAttrIdx = new int[nAttrSize];
for (int i = 0; i != nAttrSize; i++) {
string curAttr = nAttrNames.at(i);
rAttrIdx[i] = find_offset(rSpec->attrNames, curAttr);
sAttrIdx[i] = find_offset(sSpec->attrNames, curAttr);
}
// cerr << "Joined schema: ";
// for (int i = 0; i != nAttrNames.size(); i++)
// cerr << nAttrNames.at(i) << " ";
// cerr << endl;
// Here start the core part of the sort merge join algorithm
bool firstRun = true;
int rKey, sKey, rLastKey = -1, sLastIdx = 0;
size_t rSize = rSpec->memDB.size();
size_t sSize = sSpec->memDB.size();
size_t rIdx = -1, sIdx = -1;
int *rRecPtr, *sRecPtr, *nRecPtr;
for (rIdx = 0; rIdx != rSize; rIdx++) {
rRecPtr = rSpec->memDB[rIdx];
rKey = *(rRecPtr + roffset);
if (!firstRun && rKey != rLastKey) {
sLastIdx = sIdx;
}
for (sIdx = sLastIdx; sIdx != sSize; sIdx++) {
sRecPtr = sSpec->memDB[sIdx];
sKey = *(sRecPtr + soffset);
if (rKey == sKey) {
// Before we move on, we need to check other common attributes are the same, too.
if (!check_other_attrs(rRecPtr, sRecPtr, rOffs, sOffs,
otherSize)) {
continue;
} else {
// Making new records.
recordCount++;
if (saveResult) {
nRecPtr = new int[nAttrSize];
make_record(rRecPtr, sRecPtr, rAttrIdx, sAttrIdx,
nAttrSize, nRecPtr);
nSpec->memDB.push_back(nRecPtr);
}
if (printProcess) {
if (recordCount % PRINT_NUM == 1)
cerr << ".";
if (recordCount % (PRINT_NUM * 20) == 1)
cerr << endl;
}
}
} else {
if (rKey > sKey) {
continue;
} else { // rKey < sKey
break;
}
}
}
rLastKey = rKey;
firstRun = false;
}
// cerr << endl;
// cerr << "Joining done. Records: " << recordCount << endl;
//clean up
delete[] rOffs;
delete[] sOffs;
delete[] rAttrIdx;
delete[] sAttrIdx;
return nSpec;
}
