net_apps_test.cc

Go to the documentation of this file.

/* Tests for net_apps.H - Network flow applications
 *
 * Tests cover:
 * - Circulation with demands
 * - Project selection
 * - Baseball elimination
 * - Image segmentation
 * - Survey design
 */
 
 
#include <array>
 
#include <gtest/gtest.h>
#include <map>
#include <net_apps.H>
 
using namespace Aleph;
 
 
//==============================================================================
// Project Selection Tests
//==============================================================================
 
class ProjectSelectionTest : public ::testing::Test
{
protected:
};
 
TEST_F(ProjectSelectionTest, SimpleProjects)
{
  std::vector<Project<double>> projects = {
    {0, 100, {}, "Profitable A"},   // Pure profit
    {1, 50, {}, "Profitable B"},    // Pure profit
  };
 
  auto result = solve_project_selection(projects);
 
  EXPECT_DOUBLE_EQ(result.max_profit, 150.0);  // Select both
  EXPECT_EQ(result.selected.size(), 2);
}
 
TEST_F(ProjectSelectionTest, WithDependency)
{
  std::vector<Project<double>> projects = {
    {0, 100, {}, "Project A"},      // Profit 100
    {1, -30, {}, "Infra"},          // Cost 30
    {2, 50, {1}, "Project B"},      // Profit 50, needs Infra
  };
 
  auto result = solve_project_selection(projects);
 
  // Should select A (100) + Infra (-30) + B (50) = 120
  // Or just A (100) if B's dependency makes it not worth it
  // 50 - 30 = 20 > 0, so B + Infra is profitable
  // Total: 100 + 50 - 30 = 120
  EXPECT_GE(result.max_profit, 120.0);
}
 
TEST_F(ProjectSelectionTest, NoProfitableProjects)
{
  std::vector<Project<double>> projects = {
    {0, -50, {}, "Cost A"},
    {1, -30, {}, "Cost B"},
  };
 
  auto result = solve_project_selection(projects);
 
  EXPECT_DOUBLE_EQ(result.max_profit, 0.0);
  EXPECT_TRUE(result.selected.empty());
}
 
TEST_F(ProjectSelectionTest, EmptyProjects)
{
  std::vector<Project<double>> projects;
 
  auto result = solve_project_selection(projects);
 
  EXPECT_DOUBLE_EQ(result.max_profit, 0.0);
}
 
TEST_F(ProjectSelectionTest, CircularDependency)
{
  // Note: Circular dependencies are handled but don't make practical sense
  std::vector<Project<double>> projects = {
    {0, 100, {1}, "A needs B"},
    {1, 50, {0}, "B needs A"},
  };
 
  auto result = solve_project_selection(projects);
 
  // Both are selected together due to mutual dependency
  EXPECT_GE(result.max_profit, 150.0);
}
 
TEST_F(ProjectSelectionTest, ChainDependency)
{
  std::vector<Project<double>> projects = {
    {0, -10, {}, "Foundation"},
    {1, -10, {0}, "Level 1"},
    {2, -10, {1}, "Level 2"},
    {3, 100, {2}, "Payoff"},  // Profit 100, needs all previous
  };
 
  auto result = solve_project_selection(projects);
 
  // Net profit: 100 - 10 - 10 - 10 = 70
  EXPECT_GE(result.max_profit, 70.0);
}
 
 
//==============================================================================
// Baseball Elimination Tests
//==============================================================================
 
class BaseballEliminationTest : public ::testing::Test
{
protected:
  std::vector<Team> create_simple_division()
  {
    // 4-team division
    std::vector<Team> teams(4);
 
    teams[0].name = "Atlanta";
    teams[0].wins = 83;
    teams[0].losses = 71;
    teams[0].remaining = 8;
    teams[0].against = {0, 1, 6, 1};
 
    teams[1].name = "Philly";
    teams[1].wins = 80;
    teams[1].losses = 79;
    teams[1].remaining = 3;
    teams[1].against = {1, 0, 0, 2};
 
    teams[2].name = "New York";
    teams[2].wins = 78;
    teams[2].losses = 78;
    teams[2].remaining = 6;
    teams[2].against = {6, 0, 0, 0};
 
    teams[3].name = "Montreal";
    teams[3].wins = 77;
    teams[3].losses = 82;
    teams[3].remaining = 3;
    teams[3].against = {1, 2, 0, 0};
 
    return teams;
  }
};
 
TEST_F(BaseballEliminationTest, NotEliminated)
{
  auto teams = create_simple_division();
 
  // Atlanta (team 0) is not eliminated
  auto result = check_baseball_elimination(teams, 0);
 
  EXPECT_FALSE(result.eliminated);
  EXPECT_EQ(result.max_possible_wins, 91);
}
 
TEST_F(BaseballEliminationTest, TrivialElimination)
{
  std::vector<Team> teams(3);
 
  teams[0].name = "Leader";
  teams[0].wins = 100;
  teams[0].remaining = 0;
  teams[0].against = {0, 0, 0};
 
  teams[1].name = "Middle";
  teams[1].wins = 50;
  teams[1].remaining = 40;
  teams[1].against = {0, 0, 40};
 
  teams[2].name = "Loser";
  teams[2].wins = 10;
  teams[2].remaining = 50;
  teams[2].against = {0, 40, 0};
 
  // Team 2 max wins = 60, Leader already has 100
  auto result = check_baseball_elimination(teams, 2);
 
  EXPECT_TRUE(result.eliminated);
}
 
TEST_F(BaseballEliminationTest, NonTrivialElimination)
{
  auto teams = create_simple_division();
 
  // Montreal (team 3) might be eliminated non-trivially
  auto result = check_baseball_elimination(teams, 3);
 
  // Montreal max wins = 77 + 3 = 80
  // Atlanta can reach 83 + 8 = 91
  // But we need to check if there's any scenario where Montreal can win
}
 
TEST_F(BaseballEliminationTest, InvalidTeamIndex)
{
  auto teams = create_simple_division();
 
  auto result = check_baseball_elimination(teams, 100);
 
  // Should handle gracefully
  EXPECT_FALSE(result.eliminated);
}
 
 
//==============================================================================
// Image Segmentation Tests
//==============================================================================
 
class ImageSegmentationTest : public ::testing::Test
{
protected:
};
 
TEST_F(ImageSegmentationTest, SimpleSegmentation)
{
  // 2x2 image
  std::vector<std::vector<std::array<double, 2>>> data(2,
    std::vector<std::array<double, 2>>(2));
 
  // Top-left strongly prefers foreground (label 1)
  data[0][0] = {100, 10};  // cost[0]=100 (background), cost[1]=10 (foreground)
 
  // Others prefer background
  data[0][1] = {10, 100};
  data[1][0] = {10, 100};
  data[1][1] = {10, 100};
 
  auto result = segment_image(2, 2, data, 50.0);
 
  EXPECT_EQ(result.labels.size(), 2);
  EXPECT_EQ(result.labels[0].size(), 2);
 
  // Top-left should be foreground if smoothness cost allows
  // With smoothness=50, need to evaluate trade-off
}
 
TEST_F(ImageSegmentationTest, UniformPreference)
{
  // 3x3 image all preferring foreground
  std::vector<std::vector<std::array<double, 2>>> data(3,
    std::vector<std::array<double, 2>>(3));
 
  for (int i = 0; i < 3; ++i)
    for (int j = 0; j < 3; ++j)
      data[i][j] = {100, 10};  // All prefer foreground
 
  auto result = segment_image(3, 3, data, 10.0);
 
  // All should be foreground
  for (int i = 0; i < 3; ++i)
    for (int j = 0; j < 3; ++j)
      EXPECT_EQ(result.labels[i][j], 1);
}
 
TEST_F(ImageSegmentationTest, EmptyImage)
{
  std::vector<std::vector<std::array<double, 2>>> data;
 
  auto result = segment_image(0, 0, data, 10.0);
 
  EXPECT_TRUE(result.labels.empty());
}
 
TEST_F(ImageSegmentationTest, SinglePixel)
{
  std::vector<std::vector<std::array<double, 2>>> data(1,
    std::vector<std::array<double, 2>>(1));
 
  data[0][0] = {5, 10};  // Prefers background (lower cost)
 
  auto result = segment_image(1, 1, data, 100.0);
 
  EXPECT_EQ(result.labels[0][0], 0);  // Background
}
 
 
//==============================================================================
// Survey Design Tests
//==============================================================================
 
class SurveyDesignTest : public ::testing::Test
{
protected:
};
 
TEST_F(SurveyDesignTest, SimpleFeasible)
{
  std::vector<SurveyQuestion> questions = {
    {0, 1, 2},  // Question 0: needs 1-2 responses
    {1, 1, 2},  // Question 1: needs 1-2 responses
  };
 
  std::vector<SurveyRespondent> respondents = {
    {0, 1, 2, {0, 1}},  // Respondent 0: answers 1-2 questions, eligible for both
    {1, 1, 2, {0, 1}},  // Respondent 1: answers 1-2 questions, eligible for both
  };
 
  auto result = design_survey(questions, respondents);
 
  EXPECT_TRUE(result.feasible);
  EXPECT_GE(result.assignments.size(), 2);  // At least 2 assignments
}
 
TEST_F(SurveyDesignTest, Infeasible)
{
  std::vector<SurveyQuestion> questions = {
    {0, 5, 10},  // Needs at least 5 responses
  };
 
  std::vector<SurveyRespondent> respondents = {
    {0, 1, 1, {0}},  // Only 1 respondent who can answer 1 question
  };
 
  auto result = design_survey(questions, respondents);
 
  EXPECT_FALSE(result.feasible);
}
 
TEST_F(SurveyDesignTest, EligibilityConstraints)
{
  std::vector<SurveyQuestion> questions = {
    {0, 1, 3},  // Question 0
    {1, 1, 3},  // Question 1
  };
 
  std::vector<SurveyRespondent> respondents = {
    {0, 1, 2, {0}},     // Only eligible for Q0
    {1, 1, 2, {1}},     // Only eligible for Q1
    {2, 1, 2, {0, 1}},  // Eligible for both
  };
 
  auto result = design_survey(questions, respondents);
 
  EXPECT_TRUE(result.feasible);
 
  // Check assignments respect eligibility
  for (const auto& [r, q] : result.assignments)
    {
      bool eligible = false;
      for (size_t eq : respondents[r].eligible_questions)
        if (eq == q)
          eligible = true;
      EXPECT_TRUE(eligible);
    }
}
 
TEST_F(SurveyDesignTest, EmptySurvey)
{
  std::vector<SurveyQuestion> questions;
  std::vector<SurveyRespondent> respondents;
 
  auto result = design_survey(questions, respondents);
 
  // Empty survey is trivially feasible but not considered feasible here
  EXPECT_FALSE(result.feasible);
}
 
 
//==============================================================================
// Circulation Tests (Basic)
//==============================================================================
 
class CirculationTest : public ::testing::Test
{
protected:
  using TestNet = Net_Graph<Net_Node<Empty_Class>,
                            Net_Arc<Empty_Class, double>>;
};
 
TEST_F(CirculationTest, NoDemands)
{
  TestNet net;
  auto a = net.insert_node();
  auto b = net.insert_node();
  net.insert_arc(a, b, 10);
 
  auto result = solve_circulation(net,
    [](auto*) { return 0.0; },
    [](auto*) { return 0.0; });
 
  EXPECT_TRUE(result.feasible);
}
 
TEST_F(CirculationTest, SimpleFeasible)
{
  // Simple network: source produces 5 units, sink consumes 5 units
  //   s ---(10)---> t
  // demand(s) = -5 (produces), demand(t) = +5 (consumes)
  TestNet net;
  auto s = net.insert_node();
  auto t = net.insert_node();
  auto arc = net.insert_arc(s, t, 10.0);  // capacity 10
 
  std::map<TestNet::Node*, double> demands;
  demands[s] = -5.0;  // produces 5
  demands[t] = 5.0;   // consumes 5
 
  auto result = solve_circulation(net,
    [&](auto* n) { return demands.count(n) ? demands[n] : 0.0; },
    [](auto*) { return 0.0; });
 
  EXPECT_TRUE(result.feasible);
  EXPECT_DOUBLE_EQ(result.excess_flow, 5.0);
 
  // Check flow on the arc
  ASSERT_TRUE(result.flow.contains(arc));
  EXPECT_DOUBLE_EQ(result.flow[arc], 5.0);
}
 
TEST_F(CirculationTest, InfeasibleDemands)
{
  // Demand exceeds capacity
  //   s ---(5)---> t
  // demand(t) = +10 (needs 10), but capacity is only 5
  TestNet net;
  auto s = net.insert_node();
  auto t = net.insert_node();
  net.insert_arc(s, t, 5.0);
 
  std::map<TestNet::Node*, double> demands;
  demands[s] = -10.0;  // wants to produce 10
  demands[t] = 10.0;   // needs 10
 
  auto result = solve_circulation(net,
    [&](auto* n) { return demands.count(n) ? demands[n] : 0.0; },
    [](auto*) { return 0.0; });
 
  EXPECT_FALSE(result.feasible);
}
 
TEST_F(CirculationTest, WithLowerBounds)
{
  // Network with lower bounds on arcs
  //   s ---(cap=10, lower=3)---> t
  // Flow must be at least 3, at most 10
  TestNet net;
  auto s = net.insert_node();
  auto t = net.insert_node();
  auto arc = net.insert_arc(s, t, 10.0);
 
  std::map<TestNet::Node*, double> demands;
  demands[s] = -5.0;  // produces 5
  demands[t] = 5.0;   // consumes 5
 
  std::map<TestNet::Arc*, double> lower_bounds;
  lower_bounds[arc] = 3.0;  // minimum flow of 3
 
  auto result = solve_circulation(net,
    [&](auto* n) { return demands.count(n) ? demands[n] : 0.0; },
    [&](auto* a) { return lower_bounds.count(a) ? lower_bounds[a] : 0.0; });
 
  EXPECT_TRUE(result.feasible);
 
  // Flow should be 5 (satisfies demand and >= lower bound of 3)
  ASSERT_TRUE(result.flow.contains(arc));
  EXPECT_GE(result.flow[arc], 3.0);  // At least lower bound
  EXPECT_LE(result.flow[arc], 10.0); // At most capacity
}
 
TEST_F(CirculationTest, LowerBoundTooHigh)
{
  // Lower bound exceeds what's possible given demands
  //   s ---(cap=10, lower=8)---> t
  // But we only need 5 flow, and lower bound requires 8
  // This should still be feasible if capacity allows
  TestNet net;
  auto s = net.insert_node();
  auto t = net.insert_node();
  auto arc = net.insert_arc(s, t, 10.0);
 
  std::map<TestNet::Node*, double> demands;
  demands[s] = -5.0;
  demands[t] = 5.0;
 
  std::map<TestNet::Arc*, double> lower_bounds;
  lower_bounds[arc] = 8.0;  // minimum 8, but only 5 demanded
 
  auto result = solve_circulation(net,
    [&](auto* n) { return demands.count(n) ? demands[n] : 0.0; },
    [&](auto* a) { return lower_bounds.count(a) ? lower_bounds[a] : 0.0; });
 
  // This might or might not be feasible depending on exact algorithm
  // Lower bound forces more flow than demand requires
  // With proper circulation, excess flow needs to go somewhere
}
 
TEST_F(CirculationTest, BalancedNetwork)
{
  // Three nodes in a triangle, balanced demands
  //   a ---> b ---> c ---> a
  // All demands are 0, so any circulation is valid
  TestNet net;
  auto a = net.insert_node();
  auto b = net.insert_node();
  auto c = net.insert_node();
 
  auto ab = net.insert_arc(a, b, 10.0);
  auto bc = net.insert_arc(b, c, 10.0);
  auto ca = net.insert_arc(c, a, 10.0);
 
  auto result = solve_circulation(net,
    [](auto*) { return 0.0; },
    [](auto*) { return 0.0; });
 
  EXPECT_TRUE(result.feasible);
 
  // With zero demands and zero lower bounds, zero flow is a valid circulation
  EXPECT_DOUBLE_EQ(result.flow[ab], 0.0);
  EXPECT_DOUBLE_EQ(result.flow[bc], 0.0);
  EXPECT_DOUBLE_EQ(result.flow[ca], 0.0);
}
 
TEST_F(CirculationTest, NetworkUnmodified)
{
  // Verify that the network is not modified after solve_circulation
  TestNet net;
  auto s = net.insert_node();
  auto t = net.insert_node();
  auto arc = net.insert_arc(s, t, 10.0);
 
  size_t num_nodes_before = net.get_num_nodes();
  size_t num_arcs_before = net.get_num_arcs();
  double cap_before = arc->cap;
 
  std::map<TestNet::Node*, double> demands;
  demands[s] = -5.0;
  demands[t] = 5.0;
 
  auto result = solve_circulation(net,
    [&](auto* n) { return demands.count(n) ? demands[n] : 0.0; },
    [](auto*) { return 0.0; });
 
  // Network should be unchanged
  EXPECT_EQ(net.get_num_nodes(), num_nodes_before);
  EXPECT_EQ(net.get_num_arcs(), num_arcs_before);
  EXPECT_DOUBLE_EQ(arc->cap, cap_before);
  EXPECT_DOUBLE_EQ(arc->flow, 0.0);  // Flow should be reset
}
 
TEST_F(CirculationTest, MultiplePathsRequired)
{
  // Network where multiple paths are needed to satisfy demand:
  //   s -> b -> t
  //   s -> c -> t
  // (Diamond shape with s as source, t as sink, b and c as intermediate)
  // s produces 10, t consumes 10
  // Each path has capacity 6, so both paths needed
  TestNet net;
  auto s = net.insert_node();
  auto b = net.insert_node();
  auto c = net.insert_node();
  auto t = net.insert_node();
 
  auto sb = net.insert_arc(s, b, 6.0);
  auto sc = net.insert_arc(s, c, 6.0);
  auto bt = net.insert_arc(b, t, 6.0);
  auto ct = net.insert_arc(c, t, 6.0);
 
  std::map<TestNet::Node*, double> demands;
  demands[s] = -10.0;  // produces 10
  demands[t] = 10.0;   // consumes 10
 
  auto result = solve_circulation(net,
    [&](auto* n) { return demands.count(n) ? demands[n] : 0.0; },
    [](auto*) { return 0.0; });
 
  EXPECT_TRUE(result.feasible);
  EXPECT_DOUBLE_EQ(result.excess_flow, 10.0);
 
  // Both paths should be used
  double total_to_t = result.flow[bt] + result.flow[ct];
  EXPECT_DOUBLE_EQ(total_to_t, 10.0);
 
  // Verify flow on all arcs
  double total_from_s = result.flow[sb] + result.flow[sc];
  EXPECT_DOUBLE_EQ(total_from_s, 10.0);
}
 
TEST_F(CirculationTest, EmptyNetwork)
{
  TestNet net;
 
  auto result = solve_circulation(net,
    [](auto*) { return 0.0; },
    [](auto*) { return 0.0; });
 
  EXPECT_TRUE(result.feasible);
}
 
 
int main(int argc, char** argv)
{
  ::testing::InitGoogleTest(&argc, argv);
  return RUN_ALL_TESTS();
}