🐛 Changed the indices sampling logic to use lists to prevent duplicate clauses.

lib/kcnf.cpp

@@ -6,6 +6,7 @@
 #include <iostream>
 #include <random>
 #include <stdexcept>
+#include <unordered_set>
 #include <vector>
 
 #ifdef _OPENMP
@@ -50,18 +51,18 @@ std::vector<int> unrank_combination(long n, long k, long long idx) {
 // sample c indices in [0,total] for our c = # of clauses
 // this is the heart of the program as we are flattening the problem with our sampled indices and the unrank function!
 std::vector<long long> sample_indices(long long total, long c) {
-    std::vector<long long> idx;
+    std::unordered_set<long long> set;
-    idx.reserve(c);
+    set.reserve(c);
 
     std::random_device rd;
     std::mt19937 gen(rd());
     std::uniform_int_distribution<long long> dist(0, total - 1);
 
-    for (long i = 0; i < c; i++) {
+    while ((long)set.size() < c) {
-        idx.push_back(dist(gen));
+        set.insert(dist(gen));
     }
 
-    return idx;
+    return std::vector<long long>(set.begin(), set.end());
 }
 
 // decode one clause from global index (think of it like a hash function for better understanding!)