refactor: use Eigen vector type to store basic vars

src/main.cpp

@@ -1,10 +1,7 @@
-#include <iostream>
-
 #include "suboptimal/solvers/linear/SimplexPivotRule.h"
 #include "suboptimal/solvers/linear/SimplexSolverConfig.h"
 #include "suboptimal/solvers/linear/simplex.h"
 
-// using namespace std;
 using namespace Eigen;
 
 void solveBasicProblem() {

src/solvers/linear/simplex.cpp

@@ -2,10 +2,8 @@
 
 #include <Eigen/Core>
 #include <algorithm>
-#include <gsl/narrow>
 #include <iostream>
 #include <limits>
-#include <span>
 #include <stdexcept>
 #include <vector>
 
@@ -19,7 +17,7 @@ using namespace suboptimal;
 using namespace Eigen;
 using namespace std;
 
-int findPivotPosition(const MatrixXd& tableau, const span<Index> basic_vars, const SimplexPivotRule pivot_rule,
+int findPivotPosition(const MatrixXd& tableau, const VectorX<Index>& basic_vars, const SimplexPivotRule pivot_rule,
                       Index& pivot_row, Index& pivot_col) {
   pivot_col = -1;
   pivot_row = -1;
@@ -63,7 +61,7 @@ int findPivotPosition(const MatrixXd& tableau, const span<Index> basic_vars, con
         }
       } else if (ratio == min_ratio && pivot_rule == SimplexPivotRule::kBland) {
         // Bland's rule: select the basic variable with the smallest index
-        if (basic_vars[i] < basic_vars[pivot_row]) pivot_row = i;
+        if (basic_vars(i) < basic_vars(pivot_row)) pivot_row = i;
       }
     }
   }
@@ -75,7 +73,7 @@ int findPivotPosition(const MatrixXd& tableau, const span<Index> basic_vars, con
   return 0;
 }
 
-void pivot(MatrixXd& tableau, const span<Index> basic_vars, const Index pivot_row, const Index pivot_col) {
+void pivot(MatrixXd& tableau, VectorX<Index>& basic_vars, const Index pivot_row, const Index pivot_col) {
   const double pivot_element = tableau(pivot_row, pivot_col);
   tableau.row(pivot_row) /= pivot_element;
   tableau(pivot_row, pivot_col) = 1;  // Account for floating point errors
@@ -86,20 +84,20 @@ void pivot(MatrixXd& tableau, const span<Index> basic_vars, const Index pivot_ro
   }
 
   // Update basic variables
-  basic_vars[pivot_row] = pivot_col;
+  basic_vars(pivot_row) = pivot_col;
 }
 
-vector<Index> findBasicVars(const MatrixXd& tableau) {
-  vector<Index> basic_vars(tableau.rows() - 1);
+VectorX<Index> findBasicVars(const MatrixXd& tableau) {
+  VectorX<Index> basic_vars = VectorX<Index>::Zero(tableau.rows() - 1);
   for (Index i = 0; i < tableau.cols(); i++) {
     const auto col = tableau.col(i);
     Index max_index;
-    if (col.lpNorm<1>() == 1 && col.maxCoeff(&max_index) == 1) basic_vars[max_index] = i;
+    if (col.lpNorm<1>() == 1 && col.maxCoeff(&max_index) == 1) basic_vars(max_index) = i;
   }
   return basic_vars;
 }
 
-SolverExitStatus solveTableau(MatrixXd& tableau, const span<Index> basic_vars, SolverProfiler& profiler,
+SolverExitStatus solveTableau(MatrixXd& tableau, VectorX<Index>& basic_vars, SolverProfiler& profiler,
                               const SimplexSolverConfig& config) {
   int num_iterations = 0;
   auto exit_status = SolverExitStatus::kSuccess;
@@ -181,7 +179,7 @@ SolverExitStatus solveSimplex(const LinearProblem& problem, VectorXd& solution,
     tableau.row(tableau.rows() - 1) = auxiliary_objective;
 
     // Find basic variables
-    vector<Index> basic_vars = findBasicVars(tableau);
+    VectorX<Index> basic_vars = findBasicVars(tableau);
 
     // Perform simplex iterations to find initial BFS
     SolverProfiler aux_profiler{};
@@ -206,13 +204,13 @@ SolverExitStatus solveSimplex(const LinearProblem& problem, VectorXd& solution,
 
     auto objective_row = tableau.row(tableau.rows() - 1);
     objective_row.head(problem.numDecisionVars()) = -problem.getObjectiveCoeffs().transpose();
-    for (size_t i = 0; i < basic_vars.size(); i++) {
-      objective_row -= tableau.row(gsl::narrow<Index>(i)) * objective_row(basic_vars[i]);
+    for (Index i = 0; i < basic_vars.size(); i++) {
+      objective_row -= tableau.row(i) * objective_row(basic_vars(i));
     }
   }
 
   // Initialize basic variables
-  vector<Index> basic_vars = findBasicVars(tableau);
+  VectorX<Index> basic_vars = findBasicVars(tableau);
 
   // Perform simplex iterations
   SolverProfiler profiler{};
@@ -230,9 +228,9 @@ SolverExitStatus solveSimplex(const LinearProblem& problem, VectorXd& solution,
     // Extract solution and objective value
     solution = VectorXd::Zero(problem.numDecisionVars());
     const VectorXd rhs = tableau.col(tableau.cols() - 1);
-    for (size_t i = 0; i < basic_vars.size(); i++) {
-      if (const Index var_index = basic_vars[i]; var_index < problem.numDecisionVars()) {
-        solution(var_index) = rhs(gsl::narrow<Index>(i));
+    for (Index i = 0; i < basic_vars.size(); i++) {
+      if (const Index var_index = basic_vars(i); var_index < problem.numDecisionVars()) {
+        solution(var_index) = rhs(i);
       }
     }
     objective_value = tableau(tableau.rows() - 1, tableau.cols() - 1);