Erasing from a vector, passed by reference - c++

Question

I'm attempting to erase elements from a vector, by index, which has been passed by reference to some function. Typically, I'd do so like this:
void erase_element(vector<int> &my_vector, int index){
my_vector.erase(my_vector.begin() + index);
}
void make_vector(){
vector<int> my_vector = {1,2,3,4,5};
erase_element(my_vector, 3);
}
And, indeed, to check my sanity, I ran this and it works.
For some reason, however, this code throws a segfault. The problematic line seems to be that which erases elements from vectors a1, a2, and a3.
My code:
double maximum(double a, double b, double c)
{
return max(max(a, b), c);
}
double minimum(double a, double b, double c)
{
return min(min(a, b), c);
}
void findClosestTriplet(vector<double> &a1, vector<double> &a2, vector<double> &a3, TH2 *histo){
double res_min, res_max, res_mid;
int i = 0, j = 0, k = 0;
int a1Size = a1.size();
int a2Size = a2.size();
int a3Size = a3.size();
double diff = DBL_MAX;
while(i < a1Size && j < a2Size && k < a3Size){
double sum = a1[i] + a2[j] + a3[k];
double min = minimum(a1[i], a2[j], a3[k]);
double max = maximum(a1[i], a2[j], a3[k]);
if(min == a1[i]){
++i;
} else if (min == a2[j]){
++j;
} else {
++k;
}
if (diff > (max-min)){
diff = max - min;
res_max = max;
res_mid = sum - (max + min);
res_min = min;
}
}
a1.erase(a1.begin() + i); a2.erase(a2.begin() + j); a3.erase(a3.begin() + k);
}
void minimizer(){
TH2 *histo = new TH2D("Histo","Histo",1000,-0.1,0.1,1000,-0.1,0.1);
ROOT::RDataFrame f1("D","data1.root");
ROOT::RDataFrame f2("D","data2.root");
ROOT::RDataFrame f3("D","data3.root");
vector<double> a1,a2,a3;
a1.reserve(1E+6); a2.reserve(1E+6); a3.reserve(1E+6);
f1.Foreach([&](double tstamp){a1.push_back(tstamp);},{"UNIX"});
f2.Foreach([&](double tstamp){a2.push_back(tstamp);},{"UNIX"});
f3.Foreach([&](double tstamp){a3.push_back(tstamp);},{"UNIX"});
int maxiter = std::max(std::max(a1.size(), a2.size()), a3.size());
// std::ios_base::sync_with_stdio(false);
// std::cin.tie(NULL);
for(int i = 0; i < maxiter; ++i){
findClosestTriplet(a1,a2,a3,histo);
}
}
I attempted to make a minimal reproducible example, but have failed. This works:
void findClosestTriplet(vector<double> &a1, vector<double> &a2, vector<double> &a3, TH2 *histo){
int i = 1; int j = 2; int k = 3;
a1.erase(a1.begin() + i); a2.erase(a2.begin() + j); a3.erase(a3.begin() + k);
}
void minimizer(){
TH2 *histo = new TH2D("Histo","Histo",1000,-0.1,0.1,1000,-0.1,0.1);
ROOT::RDataFrame f1("D","data1.root");
ROOT::RDataFrame f2("D","data2.root");
ROOT::RDataFrame f3("D","data3.root");
vector<double> a1,a2,a3;
a1.reserve(1E+6); a2.reserve(1E+6); a3.reserve(1E+6);
f1.Foreach([&](double tstamp){a1.push_back(tstamp);},{"UNIX"});
f2.Foreach([&](double tstamp){a2.push_back(tstamp);},{"UNIX"});
f3.Foreach([&](double tstamp){a3.push_back(tstamp);},{"UNIX"});
for(int i = 0; i < a1.size(); ++i){
findClosestTriplet(a1,a2,a3,histo);
std::cout << i << "\n";
}
histo->Draw("colz");
}
Vectors a1,a2, and a3 are identical in both cases. Integers i,j,k are outside of the loop, so that seems unrelated. Moreover, I know that i,j,k are not larger than the size of a1,a2,a3, nor less than 0, at the time of the error.
I know I must be doing something dumb. But what?

Answer 1

One of i, j, or k will be the size of its respective container. The erase call for that element will be equivalent to a.erase(a.begin() + a.size()), or a.erase(a.end()). The iterator passed to erase must be valid and dereferenceable. The end iterator is not derefenceable and cannot be passed to erase.
You must check that the index is in range before using it in your erase call.