Awesome
DSViz
DSViz is a header-only C++11 library for visualization node-based data structure for generating GraphViz dot files. You can copy and put it anywhere you like.
To visualize your data structure, there are two approaches: invasive or non-invasive
Basic invasive approach
The invasive approach is very simple, you need to inherit the interface DSViz::IDataStructure
and implement the dsviz_show method in the data structure that you want to visualize. The bad thing is that you need to modify your data structure.
struct Node : public DSViz::IDataStructure {
string name;
float sum;
Node* parent;
Node* left;
Node* right;
Node(string name, float sum, Node* left, Node* right)
: name(name), sum(sum), parent(nullptr), left(left), right(right) {
if (left) left->parent = this;
if (right) right->parent = this;
}
virtual void dsviz_show(DSViz::IViz& viz) {
DSViz::TableNode node(viz, 2);
viz.setName(this, node.name);
node.addPointer("parent", parent, "", "", "", "[constraint=false]");
node.add("name", name);
node.add("sum", sum);
node.addEdge(left, "left");
node.addEdge(right, "right");
}
};
TableNode
is the class used for printing table nodes in GraphViz.
addPointer
can set a pointer which is another nodeadd
can add a new fieldaddEdge
links nodes and influence the layout of the graph
Generate a *.dot file for your data structure
Once you have the data structures, create a Dot
object and load data structures using the API load_ds
. You can load any node that can used to access the whole data structure (by keep tracing pointers). Usually, a root node is the one you want to load. We will do a depth-first order traversal to load all nodes in the graph.
Then, dot.print
will print the GraphViz *.dot format in string. You can output it to a file or print it into stdout.
Node A {string("A"), 10.0f, nullptr, nullptr};
Node B {string("B"), 12.0f, nullptr, nullptr};
Node hello {string("hello"), 11.5f, &A, &B};
DSViz::Dot dot;
dot.load_ds(&hello);
std::cout << dot.print();
You can use xdot
in Linux to open the graphviz dot file or using dot
to convert it into a png image:
dot -Tpng filename.dot -o outfile.png
Non-invasive approach
The non-invasive approach is more flexible, you don't need to modify your data structure. You can use a mock class that contains the pointer of nodes and provide a dsviz_show
function to visualize the data structure.
Use binary search tree as an example, we may have some code that implements the BST:
struct bintree_node {
bintree_node *left;
bintree_node *right;
int data;
};
class bst {
bintree_node *root;
public:
bst() { root = NULL; }
int isempty() { return (root == NULL); }
bintree_node *getRoot() { return root; }
void insert(int item);
void displayBinTree();
void printBinTree(bintree_node *);
};
The individual function dsviz_show
is used to visualize the data structure. It has the following signature:
void dsviz_show(bintree_node* P, DSViz::IViz &viz);
P
is the pointer of the node, viz
is the interface for adding nodes and edges.
Then we need to define a mock class which use this function. DSViz::Mock
is a template that has two parameters: the type of the node and the function for visualization:
typedef DSViz::Mock<bintree_node, dsviz_show> mock;
Finally, we use mock::get
to map the pointer of the original node to the mock node. It should be used everywhere in the dsviz_show
function:
#include "dsv.hpp"
void dsviz_show(bintree_node* P, DSViz::IViz &viz);
typedef DSViz::Mock<bintree_node, dsviz_show> mock;
void dsviz_show(bintree_node* P, DSViz::IViz &viz) {
DSViz::TableNode node(viz);
viz.setName(mock::get(P), node.name);
node.add("data", P->data);
if (P->left) node.addEdge(mock::get(P->left), "left");
if (P->right) node.addEdge(mock::get(P->right), "right");
}
Then we can print the dot file:
DSViz::Dot dot;
dot.load_ds(mock::get(bst.getRoot()));
std::cout << dot.print();