klips/cpp/algorithms/graphs/templated/lib-graph.hpp

469 lines
15 KiB
C++
Raw Permalink Normal View History

2021-07-22 15:15:44 +00:00
/*##############################################################################
## Author: Shaun Reed ##
2022-04-14 01:15:03 +00:00
## Legal: All Content (c) 2022 Shaun Reed, all rights reserved ##
## About: An example of a templated object graph implementation ##
2021-07-22 15:15:44 +00:00
## Algorithms in this example are found in MIT Intro to Algorithms ##
## ##
## Contact: shaunrd0@gmail.com | URL: www.shaunreed.com | GitHub: shaunrd0 ##
################################################################################
*/
#ifndef LIB_GRAPH_HPP
#define LIB_GRAPH_HPP
#include <algorithm>
2022-04-14 01:15:03 +00:00
#include <iostream>
2021-07-22 15:15:44 +00:00
#include <map>
#include <queue>
#include <unordered_map>
2022-04-14 01:15:03 +00:00
#include <unordered_set>
#include <utility>
#include <vector>
2021-07-22 15:15:44 +00:00
/******************************************************************************/
2022-04-14 01:15:03 +00:00
// Node structure for representing a graph
template <typename T>
struct Node {
public:
template <typename> friend class Graph;
template <typename> friend class InfoMST;
// Constructors
Node(const Node &rhs) = default;
Node & operator=(Node rhs)
{
if (this == &rhs) return *this;
swap(*this, rhs);
return *this;
}
Node(T data, const std::vector<std::pair<T, int>> &adj) : data_(data)
{
// Place each adjacent node in vector into our unordered_map of edges
for (const auto &i : adj) adjacent_.emplace(i.first, i.second);
}
2021-07-22 15:15:44 +00:00
2022-04-14 01:15:03 +00:00
friend void swap(Node &a, Node &b)
{
std::swap(a.data_, b.data_);
std::swap(a.adjacent_, b.adjacent_);
}
// Operators
// Define operator== for std::find; And comparisons between nodes
bool operator==(const Node<T> &b) const { return this->data_ == b.data_;}
// Define an operator!= for comparing nodes for inequality
bool operator!=(const Node<T> &b) const { return this->data_ != b.data_;}
// Accessors
inline T GetData() const { return data_;}
inline std::unordered_map<int, int> GetAdjacent() const { return adjacent_;}
private:
T data_;
// Adjacent stored in an unordered_map<adj.number, edgeWeight>
std::unordered_map<T, int> adjacent_;
};
/******************************************************************************/
// Base struct for storing traversal information on all nodes
2021-07-22 15:15:44 +00:00
// Color represents the discovery status of any given node
2022-04-14 01:15:03 +00:00
enum Color {
// Node is marked as undiscovered
White,
// Node discovery is in progress; Some adjacent nodes have not been checked
Gray,
// Node has been discovered; All adjacent nodes have been checked
Black
};
2021-07-22 15:15:44 +00:00
// Information used in all searches tracked for each node
struct NodeInfo {
2021-07-22 15:15:44 +00:00
// Coloring of the nodes is used in both DFS and BFS
Color discovered = White;
};
/******************************************************************************/
2022-04-14 01:15:03 +00:00
// BFS search information struct
2021-07-22 15:15:44 +00:00
// Node information that is only used in BFS
2021-07-22 15:15:44 +00:00
template <typename T>
struct BFS : NodeInfo {
2021-07-22 15:15:44 +00:00
// Used to represent distance from start node
int distance = 0;
// Used to represent the parent node that discovered this node
// + If we use this node as the starting point, this will remain a nullptr
const Node<T> *predecessor = nullptr;
};
2022-04-14 01:15:03 +00:00
// Store search information in unordered_maps so we can pass it around easily
// + Allows each node to store relative information on the traversal
template <typename T> using InfoBFS = std::unordered_map<T, struct BFS<T>>;
2021-07-22 15:15:44 +00:00
/******************************************************************************/
2022-04-14 01:15:03 +00:00
// DFS search information struct
2021-07-22 15:15:44 +00:00
// Node information that is only used in DFS
struct DFS : NodeInfo {
2021-07-22 15:15:44 +00:00
// Create a pair to track discovery / finish time
// + Discovery time is the iteration the node is first discovered
// + Finish time is the iteration the node has been checked completely
// ++ A finished node has considered all adjacent nodes
std::pair<int, int> discoveryFinish;
};
2022-04-14 01:15:03 +00:00
template <typename T> using InfoDFS = std::unordered_map<T, struct DFS>;
2021-07-22 15:15:44 +00:00
/******************************************************************************/
2022-04-14 01:15:03 +00:00
// MST search information struct
2021-07-22 15:15:44 +00:00
// Edges stored as multimap<weight, pair<nodeA.data_, nodeB.data_>>
template <typename T> using Edges = std::multimap<int, std::pair<T, T>>;
struct MST : NodeInfo {
2021-07-22 15:15:44 +00:00
int32_t parent = INT32_MIN;
int rank = 0;
};
template <typename T>
struct InfoMST {
template <typename> friend class Graph;
2022-04-14 01:15:03 +00:00
explicit InfoMST(const std::vector<Node<T>> &nodes)
{
for (const auto &node : nodes) {
2021-07-22 15:15:44 +00:00
// Initialize the default values for forest tracked by this struct
// + This data is used in KruskalMST() to find the MST
MakeSet(node.data_);
for (const auto adj : node.adjacent_) {
// node.number is the number that represents this node
// adj.first is the node number that is connected to this node
// adj.second is the weight of the connected edge
edges_.emplace(adj.second, std::make_pair(node.data_, adj.first));
// So we initialize the multimap<weight, <nodeA.number, nodeB.number>>
// + Since a multimap sorts by key, we have sorted our edges by weight
}
}
}
void Print()
{
std::cout << "MST result: \n";
for (const auto &edge : edgesMST_) {
std::cout << "Connected nodes: " << edge.second.first << "->"
<< edge.second.second << " with weight of " << edge.first << "\n";
}
std::cout << "Total MST weight: " << weightMST_ << std::endl;
}
void MakeSet(T x)
{
searchInfo[x].parent = x;
searchInfo[x].rank = 0;
}
void Union(T x, T y)
{
Link(FindSet(x), FindSet(y));
}
void Link(T x, T y)
{
if (searchInfo[x].rank > searchInfo[y].rank) {
searchInfo[y].parent = x;
}
else {
searchInfo[x].parent = y;
if (searchInfo[x].rank == searchInfo[y].rank) {
searchInfo[y].rank += 1;
}
}
}
T FindSet(T x)
{
if (x != searchInfo[x].parent) {
searchInfo[x].parent = FindSet(searchInfo[x].parent);
}
return searchInfo[x].parent;
}
private:
std::unordered_map<T, struct MST> searchInfo;
// All of the edges within our graph
// + Since each node stores its own edges, this is initialized in InfoMST ctor
Edges<T> edges_;
// A multimap of the edges found for our MST
Edges<T> edgesMST_;
// The total weight of our resulting MST
int weightMST_ = 0;
};
/******************************************************************************/
// Templated graph class
template <class T>
class Graph {
public:
// Constructor
2022-04-14 01:15:03 +00:00
explicit Graph(std::vector<Node<T>> nodes) : nodes_(std::move(nodes)) {}
2021-07-22 15:15:44 +00:00
// Breadth First Search
InfoBFS<T> BFS(const Node<T>& startNode) const;
std::deque<Node<T>> PathBFS(const Node<T> &start, const Node<T> &finish) const;
// Depth First Search
InfoDFS<T> DFS() const;
// An alternate DFS that checks each node of the graph beginning at startNode
InfoDFS<T> DFS(const Node<T> &startNode) const;
// Visit function is used in both versions of DFS
void DFSVisit(int &time, const Node<T>& startNode, InfoDFS<T> &searchInfo) const;
// Topological sort, using DFS
std::vector<Node<T>> TopologicalSort(const Node<T> &startNode) const;
// Kruskal's MST
InfoMST<T> KruskalMST() const;
// Returns a copy of a node with the number i within the graph
// + This uses the private, non-const accessor GetNode() and returns a copy
inline Node<T> GetNodeCopy(T i) { return GetNode(i);}
// Return a constant iterator for reading node values
inline typename std::vector<Node<T>>::const_iterator NodeBegin()
{ return nodes_.cbegin();}
private:
// A non-const accessor for direct access to a node with the number value i
inline Node<T> & GetNode(T i)
{ return *std::find(nodes_.begin(), nodes_.end(), Node<T>(i, {}));}
// For grabbing a const qualified node
inline const Node<T> & GetNode(T i) const
{ return *std::find(nodes_.begin(), nodes_.end(), Node<T>(i, {}));}
std::vector<Node<T>> nodes_;
};
/******************************************************************************/
// Graph class member function definitions
template <typename T>
InfoBFS<T> Graph<T>::BFS(const Node<T> &startNode) const
{
// Create local object to track the information gathered during traversal
InfoBFS<T> searchInfo;
// Create a queue to visit discovered nodes in FIFO order
std::queue<const Node<T> *> visitQueue;
// Mark the startNode as in progress until we finish checking adjacent nodes
searchInfo[startNode.data_].discovered = Gray;
// Visit the startNode
visitQueue.push(&startNode);
// Continue to visit nodes until there are none left in the graph
while (!visitQueue.empty()) {
// Remove thisNode from the visitQueue, storing its vertex locally
const Node<T> * thisNode = visitQueue.front();
visitQueue.pop();
std::cout << "Visiting node " << thisNode->data_ << std::endl;
// Check if we have already discovered all the adjacentNodes to thisNode
for (const auto &adjacent : thisNode->adjacent_) {
if (searchInfo[adjacent.first].discovered == White) {
std::cout << "Found undiscovered adjacentNode: " << adjacent.first
<< "\n";
// Mark the adjacent node as in progress
searchInfo[adjacent.first].discovered = Gray;
searchInfo[adjacent.first].distance =
searchInfo[thisNode->data_].distance + 1;
searchInfo[adjacent.first].predecessor =
&GetNode(thisNode->data_);
// Add the discovered node the the visitQueue
visitQueue.push(&GetNode(adjacent.first));
}
}
// We are finished with this node and the adjacent nodes; Mark it discovered
searchInfo[thisNode->data_].discovered = Black;
}
// Return the information gathered from this search, JIC caller needs it
return searchInfo;
}
template <typename T>
std::deque<Node<T>> Graph<T>::PathBFS(const Node<T> &start,
const Node<T> &finish) const
{
// Store the path as copies of each node
// + If the caller modifies these, it will not impact the graph's data
std::deque<Node<T>> path;
InfoBFS<T> searchInfo = BFS(start);
const Node<T> * next = searchInfo[finish.data_].predecessor;
bool isValid = false;
do {
// If we have reached the start node, we have found a valid path
if (*next == Node<T>(start)) isValid = true;
// Add the node to the path as we check each node
// + Use emplace_front to call the Node copy constructor
path.emplace_front(*next);
// Move to the next node
next = searchInfo[next->data_].predecessor;
} while (next != nullptr);
// Use emplace_back to call Node copy constructor
path.emplace_back(finish);
// If we never found a valid path, erase all contents of the path
if (!isValid) path.erase(path.begin(), path.end());
// Return the path, the caller should handle empty paths accordingly
return path;
}
template <typename T>
InfoDFS<T> Graph<T>::DFS() const
{
// Track the nodes we have discovered
InfoDFS<T> searchInfo;
int time = 0;
// Visit each node in the graph
for (const auto& node : nodes_) {
std::cout << "Visiting node " << node.data_ << std::endl;
// If the node is undiscovered, visit it
if (searchInfo[node.data_].discovered == White) {
std::cout << "Found undiscovered node: " << node.data_ << std::endl;
// Visiting the undiscovered node will check it's adjacent nodes
DFSVisit(time, node, searchInfo);
}
}
return searchInfo;
}
template <typename T>
InfoDFS<T> Graph<T>::DFS(const Node<T> &startNode) const
{
// Track the nodes we have discovered
InfoDFS<T> searchInfo;
int time = 0;
auto startIter = std::find(nodes_.begin(), nodes_.end(),
Node<T>(startNode.data_, {})
);
// beginning at startNode, visit each node in the graph until we reach the end
while (startIter != nodes_.end()) {
std::cout << "Visiting node " << startIter->data_ << std::endl;
// If the startIter is undiscovered, visit it
if (searchInfo[startIter->data_].discovered == White) {
std::cout << "Found undiscovered node: " << startIter->data_ << std::endl;
// Visiting the undiscovered node will check it's adjacent nodes
DFSVisit(time, *startIter, searchInfo);
}
startIter++;
}
// Once we reach the last node, check the beginning for unchecked nodes
startIter = nodes_.begin();
// Once we reach the initial startNode, we have checked all nodes
while (*startIter != startNode) {
std::cout << "Visiting node " << startIter->data_ << std::endl;
// If the startIter is undiscovered, visit it
if (searchInfo[startIter->data_].discovered == White) {
std::cout << "Found undiscovered node: " << startIter->data_ << std::endl;
// Visiting the undiscovered node will check it's adjacent nodes
DFSVisit(time, *startIter, searchInfo);
}
startIter++;
}
return searchInfo;
}
template <typename T>
void Graph<T>::DFSVisit(int &time, const Node<T>& startNode,
InfoDFS<T> &searchInfo) const
{
searchInfo[startNode.data_].discovered = Gray;
time++;
searchInfo[startNode.data_].discoveryFinish.first = time;
// Check the adjacent nodes of the startNode
for (const auto &adjacent : startNode.adjacent_) {
auto iter = std::find(nodes_.begin(), nodes_.end(),
Node<T>(adjacent.first, {}));
// If the adjacentNode is undiscovered, visit it
// + Offset by 1 to account for 0 index of discovered vector
if (searchInfo[iter->data_].discovered == White) {
std::cout << "Found undiscovered adjacentNode: "
<< GetNode(adjacent.first).data_ << std::endl;
// Visiting the undiscovered node will check it's adjacent nodes
DFSVisit(time, *iter, searchInfo);
}
}
searchInfo[startNode.data_].discovered = Black;
time++;
searchInfo[startNode.data_].discoveryFinish.second = time;
}
template <typename T>
std::vector<Node<T>> Graph<T>::TopologicalSort(const Node<T> &startNode) const
{
InfoDFS<T> topological = DFS(GetNode(startNode.data_));
std::vector<Node<T>> order(nodes_);
auto comp = [&topological](const Node<T> &a, const Node<T> &b) {
return (topological[a.data_].discoveryFinish.second <
topological[b.data_].discoveryFinish.second);
};
std::sort(order.begin(), order.end(), comp);
// The topologicalOrder is read right-to-left in the final result
// + Output is handled in main as FILO, similar to a stack
return order;
}
template <typename T>
InfoMST<T> Graph<T>::KruskalMST() const
{
InfoMST<T> searchInfo(nodes_);
// The ctor for InfoMST initializes all edges within the graph into a multimap
// + Key for multimap is edge weight, so they're already sorted in ascending
// For each edge in the graph, check if they are part of the same tree
// + Since we do not want to create a cycle in the MST forest -
// + we don't connect nodes that are part of the same tree
for (const auto &edge : searchInfo.edges_) {
// Two integers representing the node.number for the connected nodes
const int u = edge.second.first;
const int v = edge.second.second;
// Check if the nodes are of the same tree
if (searchInfo.FindSet(u) != searchInfo.FindSet(v)) {
// If they are not, add the edge to our MST
searchInfo.edgesMST_.emplace(edge);
searchInfo.weightMST_ += edge.first;
// Update the forest to reflect this change
searchInfo.Union(u, v);
}
}
return searchInfo;
}
#endif // LIB_GRAPH_HPP