| 1 | /** \file |
| 2 | * \brief Tests for Max Flow Algorithms |
| 3 | * |
| 4 | * \author Ivo Hedtke, Tilo Wiedera |
| 5 | * |
| 6 | * \par License: |
| 7 | * This file is part of the Open Graph Drawing Framework (OGDF). |
| 8 | * |
| 9 | * \par |
| 10 | * Copyright (C)<br> |
| 11 | * See README.md in the OGDF root directory for details. |
| 12 | * |
| 13 | * \par |
| 14 | * This program is free software; you can redistribute it and/or |
| 15 | * modify it under the terms of the GNU General Public License |
| 16 | * Version 2 or 3 as published by the Free Software Foundation; |
| 17 | * see the file LICENSE.txt included in the packaging of this file |
| 18 | * for details. |
| 19 | * |
| 20 | * \par |
| 21 | * This program is distributed in the hope that it will be useful, |
| 22 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 23 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 24 | * GNU General Public License for more details. |
| 25 | * |
| 26 | * \par |
| 27 | * You should have received a copy of the GNU General Public |
| 28 | * License along with this program; if not, see |
| 29 | * http://www.gnu.org/copyleft/gpl.html |
| 30 | */ |
| 31 | |
| 32 | #include <ogdf/graphalg/MaxFlowEdmondsKarp.h> |
| 33 | #include <ogdf/graphalg/MaxFlowGoldbergTarjan.h> |
| 34 | #include <ogdf/graphalg/MaxFlowSTPlanarDigraph.h> |
| 35 | #include <ogdf/graphalg/MaxFlowSTPlanarItaiShiloach.h> |
| 36 | #include <ogdf/graphalg/ConnectivityTester.h> |
| 37 | #include <ogdf/basic/graph_generators.h> |
| 38 | |
| 39 | #include <resources.h> |
| 40 | |
| 41 | // if defined will print the first failed instance |
| 42 | #define PRINT_FIRST_FAIL |
| 43 | |
| 44 | using std::string; |
| 45 | using std::endl; |
| 46 | |
| 47 | #ifdef PRINT_FIRST_FAIL |
| 48 | bool printedFailedInstance = false; |
| 49 | |
| 50 | /** |
| 51 | * Defines which properties a graph fullfills |
| 52 | * or an algorithm requires. |
| 53 | */ |
| 54 | enum MaxFlowRequirement { |
| 55 | MFR_NONE = 0, |
| 56 | MFR_PLANAR = 1, |
| 57 | MFR_ST_PLANAR = 2, |
| 58 | MFR_CONNECTED = 4, |
| 59 | MFR_ST_NON_INCIDENT_FACE = 8, |
| 60 | }; |
| 61 | |
| 62 | /** |
| 63 | * Used to combine requirements or properties. |
| 64 | */ |
| 65 | MaxFlowRequirement operator|(MaxFlowRequirement a, MaxFlowRequirement b) |
| 66 | { |
| 67 | return MaxFlowRequirement(int(a) | int(b)); |
| 68 | } |
| 69 | |
| 70 | /** |
| 71 | * Establishes the properties of the given graph. |
| 72 | * |
| 73 | * @param graph The graph to be investigated |
| 74 | * @param s source node of the instance |
| 75 | * @param t target node of the instance |
| 76 | * @return all properties except for \c MFR_ST_NON_INCIDENT_FACE |
| 77 | */ |
| 78 | MaxFlowRequirement determineProperties(const Graph &graph, node s, node t) |
| 79 | { |
| 80 | MaxFlowRequirement result = MFR_NONE; |
| 81 | |
| 82 | if(isPlanar(graph)) { |
| 83 | result = result | MFR_PLANAR; |
| 84 | |
| 85 | if(isSTPlanar(graph, s, t)) { |
| 86 | result = result | MFR_ST_PLANAR; |
| 87 | } |
| 88 | } |
| 89 | |
| 90 | if(isConnected(graph)) { |
| 91 | result = result | MFR_CONNECTED; |
| 92 | } |
| 93 | |
| 94 | return result; |
| 95 | } |
| 96 | |
| 97 | /** |
| 98 | * Used to print the first encountered failed instance. |
| 99 | * Always returns false and can thus be used in an bandit assertion. |
| 100 | * |
| 101 | * @param graph the graph to be printed |
| 102 | * @param caps the capacities |
| 103 | * @param node s the source node |
| 104 | * @param node t the sink node |
| 105 | * @param flow the calculated flow |
| 106 | */ |
| 107 | template<typename T> |
| 108 | bool printInstance(const Graph &graph, const EdgeArray<T> caps, const node s, const node t, const EdgeArray<T> &flows) |
| 109 | { |
| 110 | if(!printedFailedInstance) { |
| 111 | printedFailedInstance = true; |
| 112 | |
| 113 | std::cout << std::endl << "Graph consists of " << graph.numberOfNodes() << " nodes:" << std::endl; |
| 114 | for(node v : graph.nodes) { |
| 115 | std::cout << " " << v; |
| 116 | if(v == s) { std::cout << " [source]" ; } |
| 117 | if(v == t) { std::cout << " [sink]" ; } |
| 118 | std::cout << std::endl; |
| 119 | } |
| 120 | std::cout << "Graph has " << graph.numberOfEdges() << " edges:" << std::endl; |
| 121 | for(edge e : graph.edges) { |
| 122 | std::cout << " " << e << " " << flows[e] << " / " << caps[e] << std::endl; |
| 123 | } |
| 124 | } |
| 125 | return false; |
| 126 | } |
| 127 | #endif |
| 128 | |
| 129 | /** |
| 130 | * Asserts the provided flow is valid. |
| 131 | * Utilizes EpsilonTest to check the flow values. |
| 132 | * |
| 133 | * @param graph the problem instance |
| 134 | * @param caps the capacity of each edge |
| 135 | * @param s the source node |
| 136 | * @param t the sink node |
| 137 | * @param flows the flow to be validated |
| 138 | * @param the total flow from source to sink |
| 139 | * @param computeFlow if true the reference algorithms result is |
| 140 | * compared to the given flow |
| 141 | */ |
| 142 | template<typename VALUE_TYPE> |
| 143 | void validateFlow( |
| 144 | const Graph &graph, |
| 145 | const EdgeArray<VALUE_TYPE> &caps, |
| 146 | const node s, |
| 147 | const node t, |
| 148 | const EdgeArray<VALUE_TYPE> &flows, |
| 149 | const VALUE_TYPE flow, |
| 150 | bool computeFlow = false) |
| 151 | { |
| 152 | EpsilonTest et; |
| 153 | const VALUE_TYPE ZERO(0); |
| 154 | |
| 155 | for(edge e : graph.edges) { |
| 156 | AssertThat(et.leq(flows[e], caps[e]) || printInstance(graph, caps, s, t, flows), IsTrue()); |
| 157 | } |
| 158 | |
| 159 | for(node v : graph.nodes) { |
| 160 | VALUE_TYPE income(ZERO); |
| 161 | VALUE_TYPE output(ZERO); |
| 162 | for(adjEntry adj : v->adjEntries) { |
| 163 | edge e = adj->theEdge(); |
| 164 | if(e->isSelfLoop()) { |
| 165 | // self-loop, flow must be 0 |
| 166 | AssertThat(et.equal(flows[e], ZERO) || printInstance(graph, caps, s, t, flows), IsTrue()); |
| 167 | } else { |
| 168 | if(e->source() == v) { |
| 169 | output += flows[e]; |
| 170 | } else { |
| 171 | OGDF_ASSERT(e->target() == v); |
| 172 | income += flows[e]; |
| 173 | } |
| 174 | } |
| 175 | } |
| 176 | if(v == s) { |
| 177 | // there are Max-Flow algorithms that allow incoming flow in s |
| 178 | AssertThat(et.equal(output, flow+income) || printInstance(graph, caps, s, t, flows), IsTrue()); |
| 179 | } else if(v == t) { |
| 180 | // there are Max-Flow algorithms that allow outgoing flow from t |
| 181 | AssertThat(et.equal(income, flow+output) || printInstance(graph, caps, s, t, flows), IsTrue()); |
| 182 | } else { |
| 183 | AssertThat(et.equal(income, output) || printInstance(graph, caps, s, t, flows), IsTrue()); |
| 184 | } |
| 185 | } |
| 186 | |
| 187 | // using Edmonds & Karp algorithm for reference |
| 188 | if(computeFlow) { |
| 189 | MaxFlowEdmondsKarp<VALUE_TYPE> mfek(graph); |
| 190 | VALUE_TYPE refFlow = mfek.computeValue(caps, s, t); |
| 191 | AssertThat(et.equal(flow, refFlow) || printInstance(graph, caps, s, t, flows), IsTrue()); |
| 192 | } |
| 193 | } |
| 194 | |
| 195 | /** |
| 196 | * Tests a given maximum flow algorithm. |
| 197 | * |
| 198 | * @param name the human-readable description of this algorithm |
| 199 | * @param the requiremets for this algorithm |
| 200 | */ |
| 201 | template<typename MAX_FLOW_ALGO, typename VALUE_TYPE> |
| 202 | void describeMaxFlowModule(const string &name, const MaxFlowRequirement reqs = MFR_NONE) |
| 203 | { |
| 204 | const int maxCapacity = 100; |
| 205 | const int maxNodes = 50; |
| 206 | |
| 207 | describe(name, [&](){ |
| 208 | // test predefined instances |
| 209 | for_each_file("maxflow" , [&](const ResourceFile* file){ |
| 210 | it("works on " + file->fullPath(), [&] { |
| 211 | // optimal solution value is extracted from the filename |
| 212 | std::string filename = file->name(); |
| 213 | string tmp = filename.substr(0, filename.length() - 4); |
| 214 | tmp = tmp.substr(tmp.find_last_of('.') + 1); |
| 215 | std::stringstream ss(tmp); |
| 216 | VALUE_TYPE opt = -1; |
| 217 | ss >> opt; |
| 218 | |
| 219 | Graph graph; |
| 220 | EdgeArray<VALUE_TYPE> caps(graph, 0); |
| 221 | node s; |
| 222 | node t; |
| 223 | std::stringstream is{file->data()}; |
| 224 | AssertThat(GraphIO::readDMF(graph, caps, s, t, is), IsTrue()); |
| 225 | MaxFlowRequirement props = determineProperties(graph, s, t); |
| 226 | |
| 227 | // create non s-t-incident face if required |
| 228 | if(reqs & MFR_ST_NON_INCIDENT_FACE) { |
| 229 | props = props | MFR_ST_NON_INCIDENT_FACE; |
| 230 | node v = graph.newNode(); |
| 231 | graph.newEdge(v, t); |
| 232 | graph.newEdge(t, v); |
| 233 | } |
| 234 | |
| 235 | if(props & MFR_PLANAR) { |
| 236 | planarEmbed(graph); |
| 237 | } |
| 238 | |
| 239 | if((reqs & props) == reqs) { |
| 240 | MAX_FLOW_ALGO alg(graph); |
| 241 | |
| 242 | VALUE_TYPE value = alg.computeValue(caps, s, t); |
| 243 | AssertThat(value, Equals(opt)); |
| 244 | |
| 245 | EdgeArray<VALUE_TYPE> flow(graph); |
| 246 | alg.computeFlowAfterValue(flow); |
| 247 | validateFlow(graph, caps, s, t, flow, value); |
| 248 | } |
| 249 | }); |
| 250 | }); |
| 251 | |
| 252 | // test random instances |
| 253 | for(int n = 2; n < maxNodes; n++) { |
| 254 | it("works on a random graph of approximate size " + to_string(n), [&] { |
| 255 | Graph graph; |
| 256 | EdgeArray<VALUE_TYPE> caps(graph); |
| 257 | node s = nullptr; |
| 258 | node t = nullptr; |
| 259 | |
| 260 | // generate a connected graph based on the requirements of this algorithm |
| 261 | if(reqs & MFR_ST_PLANAR) { |
| 262 | if(n % 2) { |
| 263 | int r = 1 + (int)sqrt(n); |
| 264 | gridGraph(graph, r, r, false, false); |
| 265 | List<node> nodes; |
| 266 | graph.allNodes(nodes); |
| 267 | s = *nodes.get(randomNumber(0, r-1)); |
| 268 | t = *nodes.get(randomNumber(r*(r-1), r*r-1)); |
| 269 | } else { |
| 270 | int m = randomNumber(n-1, max(n-1, 3*n-6)); |
| 271 | randomPlanarConnectedGraph(graph, n, m); |
| 272 | s = graph.chooseNode(); |
| 273 | CombinatorialEmbedding ce(graph); |
| 274 | |
| 275 | // select sink with common face |
| 276 | for(adjEntry adj = s->firstAdj(); |
| 277 | t == nullptr || randomNumber(0, s->degree()); |
| 278 | adj = adj->faceCycleSucc()) { |
| 279 | node v = adj->theNode(); |
| 280 | |
| 281 | if(s != v) { |
| 282 | t = v; |
| 283 | } |
| 284 | } |
| 285 | } |
| 286 | } else if(reqs & MFR_PLANAR) { |
| 287 | int m = randomNumber(n-1, max(n-1, 3*n-6)); |
| 288 | randomPlanarConnectedGraph(graph, n, m); |
| 289 | } else { |
| 290 | int m = randomNumber(n*2, max(n*2, n*(n-1)/2)); |
| 291 | randomBiconnectedGraph(graph, n, m); |
| 292 | } |
| 293 | |
| 294 | // generate capacities |
| 295 | caps.init(graph); |
| 296 | for (edge e: graph.edges) { |
| 297 | caps[e] = (VALUE_TYPE) randomDouble(1, maxCapacity); |
| 298 | } |
| 299 | |
| 300 | // choose source and sink |
| 301 | if(s == nullptr || s == t) { |
| 302 | s = graph.chooseNode([&](node v) { return v != t; }); |
| 303 | } |
| 304 | while(t == nullptr || t == s) { |
| 305 | t = graph.chooseNode([&](node v) { return v != s; }); |
| 306 | } |
| 307 | |
| 308 | // create non s-t-incident face if required |
| 309 | if(reqs & MFR_ST_NON_INCIDENT_FACE) { |
| 310 | node v = graph.newNode(); |
| 311 | caps[graph.newEdge(v, t)] = 0; |
| 312 | caps[graph.newEdge(t, v)] = 0; |
| 313 | } |
| 314 | |
| 315 | // compute flow and validate it |
| 316 | MAX_FLOW_ALGO alg(graph); |
| 317 | EdgeArray<VALUE_TYPE> algFlows(graph); |
| 318 | |
| 319 | VALUE_TYPE algFlow = alg.computeValue(caps, s, t); |
| 320 | alg.computeFlowAfterValue(algFlows); |
| 321 | |
| 322 | validateFlow(graph, caps, s, t, algFlows, algFlow, true); |
| 323 | }); |
| 324 | } |
| 325 | }); |
| 326 | } |
| 327 | |
| 328 | template<typename T> |
| 329 | void registerTestSuite(const string typeName) |
| 330 | { |
| 331 | const string suffix = "<" + typeName + ">" ; |
| 332 | |
| 333 | describeMaxFlowModule<MaxFlowSTPlanarItaiShiloach<T>, T>("MaxFlowSTPlanarItaiShiloach" + suffix, |
| 334 | MFR_CONNECTED | MFR_ST_PLANAR); |
| 335 | describeMaxFlowModule<MaxFlowSTPlanarDigraph<T>, T>("MaxFlowSTPlanarDigraph" + suffix, |
| 336 | MFR_CONNECTED | MFR_ST_PLANAR); |
| 337 | describeMaxFlowModule<MaxFlowEdmondsKarp<T>, T>("MaxFlowEdmondsKarp" + suffix); |
| 338 | describeMaxFlowModule<MaxFlowGoldbergTarjan<T>, T>("MaxFlowGoldbergTarjan" + suffix); |
| 339 | } |
| 340 | |
| 341 | /** |
| 342 | * Tests the ConnectivityTester on the given graphs. |
| 343 | * Node and edge connectivity is computed for both directed and undirected (i.e. bi-directed) graphs. |
| 344 | * The resulting values are tested for consistency. |
| 345 | * |
| 346 | * @param title A description of the graphs |
| 347 | * @param expected The minimal expected node connectivity |
| 348 | * @param initializer A lambda expression used to initialize the test instances. |
| 349 | * Each call is provided with the graph to be initialized and a counter. |
| 350 | */ |
| 351 | void ( |
| 352 | string title, |
| 353 | int expected, |
| 354 | std::function<void (Graph&, int)> initializer) |
| 355 | { |
| 356 | describe(title, [&]() |
| 357 | { |
| 358 | const int maxNodes = 50; |
| 359 | |
| 360 | // undirected node connectivity |
| 361 | ConnectivityTester nodeAlgo; |
| 362 | |
| 363 | // undirected edge connectivity |
| 364 | ConnectivityTester edgeAlgo(false); |
| 365 | |
| 366 | // directed node connectivity |
| 367 | ConnectivityTester nodeDirAlgo(true, true); |
| 368 | |
| 369 | // directed edge connectivity |
| 370 | ConnectivityTester edgeDirAlgo(false, true); |
| 371 | |
| 372 | for (int n = 3; n < maxNodes / 2; n++) { |
| 373 | it("works for " + to_string(n) + " nodes" , [&] { |
| 374 | Graph graph; |
| 375 | initializer(graph, n); |
| 376 | |
| 377 | NodeArray<NodeArray<int>> edgeCon(graph, NodeArray<int>(graph)); |
| 378 | NodeArray<NodeArray<int>> nodeCon(graph, NodeArray<int>(graph)); |
| 379 | NodeArray<NodeArray<int>> edgeDirCon(graph, NodeArray<int>(graph)); |
| 380 | NodeArray<NodeArray<int>> nodeDirCon(graph, NodeArray<int>(graph)); |
| 381 | |
| 382 | // compute the connectivity |
| 383 | edgeAlgo.computeConnectivity(graph, edgeCon); |
| 384 | int minConnectivity = nodeAlgo.computeConnectivity(graph, nodeCon); |
| 385 | nodeDirAlgo.computeConnectivity(graph, nodeDirCon); |
| 386 | edgeDirAlgo.computeConnectivity(graph, edgeDirCon); |
| 387 | |
| 388 | AssertThat(minConnectivity, IsGreaterThan(expected - 1)); |
| 389 | |
| 390 | // assert consistency with existing tests |
| 391 | if(n > 3 && isTriconnected(graph)) { |
| 392 | AssertThat(minConnectivity, IsGreaterThan(2)); |
| 393 | } else if(n > 2 && isBiconnected(graph)) { |
| 394 | AssertThat(minConnectivity, IsGreaterThan(1)); |
| 395 | } else if(n > 1 && isConnected(graph)) { |
| 396 | AssertThat(minConnectivity, IsGreaterThan(0)); |
| 397 | } |
| 398 | |
| 399 | // check consistency of connectivity variants |
| 400 | for (node v : graph.nodes) { |
| 401 | for (node w : graph.nodes) { |
| 402 | if (v == w) { |
| 403 | AssertThat(nodeCon[v][w], Equals(0)); |
| 404 | } else { |
| 405 | // compare with expected values |
| 406 | AssertThat(nodeCon[v][w], IsGreaterThan(expected - 1)); |
| 407 | AssertThat(nodeCon[v][w], IsGreaterThan(minConnectivity - 1)); |
| 408 | |
| 409 | // edge connectivity is least restrictive |
| 410 | AssertThat(edgeCon[v][w], IsGreaterThan(nodeCon[v][w] - 1)); |
| 411 | AssertThat(edgeCon[v][w], IsGreaterThan(edgeDirCon[v][w] - 1)); |
| 412 | |
| 413 | // (node) connectivity might never be greater than edge connectivity |
| 414 | AssertThat(edgeCon[v][w], IsGreaterThan(edgeDirCon[v][w] - 1)); |
| 415 | |
| 416 | // directed connectivity is most restrictive |
| 417 | AssertThat(nodeCon[v][w], IsGreaterThan(nodeDirCon[v][w] - 1)); |
| 418 | AssertThat(edgeDirCon[v][w], IsGreaterThan(nodeDirCon[v][w] - 1)); |
| 419 | } |
| 420 | } |
| 421 | } |
| 422 | |
| 423 | // create a new node with some edges |
| 424 | // thus reducing the overall connectivity |
| 425 | if(minConnectivity > 0) { |
| 426 | node w = graph.firstNode(); |
| 427 | node v = graph.newNode(); |
| 428 | int modifiedExpected = randomNumber(0, minConnectivity - 1); |
| 429 | for (int i = 0; i < modifiedExpected; i++) { |
| 430 | OGDF_ASSERT(w != v); |
| 431 | graph.newEdge(w, v); |
| 432 | w = w->succ(); |
| 433 | } |
| 434 | |
| 435 | AssertThat(nodeAlgo.computeConnectivity(graph, nodeCon), Equals(modifiedExpected)); |
| 436 | } |
| 437 | }); |
| 438 | } |
| 439 | }); |
| 440 | } |
| 441 | |
| 442 | go_bandit([]() { |
| 443 | describe("Maximum flow algorithms" , [](){ |
| 444 | registerTestSuite<int>("int" ); |
| 445 | registerTestSuite<double>("double" ); |
| 446 | registerTestSuite<unsigned long long int>("unsigned long long int" ); |
| 447 | }); |
| 448 | |
| 449 | describe("Connectivity Tester" , [](){ |
| 450 | describeConnectivityTester("random graphs" , 0, [](Graph &graph, int n) { |
| 451 | randomGraph(graph, n, randomNumber(n, (n*(n-1))/2)); |
| 452 | }); |
| 453 | |
| 454 | describeConnectivityTester("planar connected graphs" , 1, [](Graph &graph, int n) { |
| 455 | randomPlanarConnectedGraph(graph, n, randomNumber(n, (n*(n-1))/2)); |
| 456 | }); |
| 457 | |
| 458 | describeConnectivityTester("biconnected graphs" , 2, [](Graph &graph, int n) { |
| 459 | randomBiconnectedGraph(graph, n, randomNumber(n, (n*(n-1))/2)); |
| 460 | }); |
| 461 | |
| 462 | describeConnectivityTester("triconnected graphs" , 3, [](Graph &graph, int n) { |
| 463 | randomTriconnectedGraph(graph, n, randomDouble(0, 1), randomDouble(0, 1)); |
| 464 | }); |
| 465 | }); |
| 466 | }); |
| 467 | |