Fixed missing taskflow. simple test to see if validation is being used

environment.yml

@@ -11,6 +11,7 @@ dependencies:
     - ninja
     - ceres-solver=2.1.0
     - pybind11>=2.6.2
+    - taskflow
     - pytest #=6.2.4
     - pydot
     - scikit-learn

src/engine.cpp

@@ -242,10 +242,9 @@ bool Engine<T>::update_best(const Dataset& data, bool val)
         //std::cout << ind.program.get_model() << std::endl;
 
         //std::cout << "got individual of rank" << ind.fitness.rank << std::endl;
-        if (val)
-            f = ind.fitness.loss_v;
-        else
-            f = ind.fitness.loss;
+
+        // if there is no validation, then loss_v == loss and this should work just fine
+        f = ind.fitness.loss_v;
 
         if (f*error_weight > bs*error_weight
         || (f == bs && ind.fitness.complexity < this->best_complexity)

tests/cpp/test_brush.cpp

@@ -46,10 +46,19 @@ TEST(Engine, EngineWorks)
 
     params.set_verbosity(2); // TODO: verbosity tests
 
+     // checking if validation size works
+    params.set_validation_size(0.2);
+
     std::cout << "n jobs = 1" << std::endl;
     params.set_n_jobs(1);
     Brush::RegressorEngine est5(params);
-    est5.run(data);
+    est5.run(data); // this will not use validation size from parameters
+    std::cout << "best individual using run(data)" << std::endl;
+    std::cout << est5.best_ind.program.get_model() << std::endl;
+
+    est5.fit(X, y); // this will use validation size from parameters
+    std::cout << "best individual using fit(X, y)" << std::endl;
+    std::cout << est5.best_ind.program.get_model() << std::endl;
 
     std::cout << "n jobs = 2" << std::endl;
     params.set_n_jobs(2);