libgvizard

libgvizard is a c++17 header-only graph visualization abstraction library that only provides a comprehensive flexible set of types and interfaces heavily inspired by graphviz, along with visualization attributes, graph and related data structures, and other utilities.

upon this library, many can make different renderers and thus one with a single graph definition can render it by different renderers of choice.

heavily inspired by graphviz.

Requirements

if vcpkg build toolchain isn’t specified, the packages will be automatically downloaded on cmake build.

Build tests and examples

if you already have vcpkg, specify it with -DCMAKE_TOOLCHAIN_FILE=/path/to/vcpkg.cmake.

mkdir build && cd build
cmake ..
make

to test:

ctest --output-on-failure

to run a specific example:

./examples/simple_dot_generator/simple_dot_generator

to generate docs add -DLIBGVIZARD_OPT_GENERATE_DOCS=1 as it’s not targeted by default:

mkdir build && cd build
cmake .. -DLIBGVIZARD_OPT_GENERATE_DOCS=1
make

now the html files will be at ./docs/sphinx/.

Usage exampe

a glare of some parts of the api:

using NodeName = typename MyRenderer::NodeName;

auto make_graph()
{
  gviz::GvizGraph<> ret{};

  auto node_a = ret.graph.create_node();
  auto node_b = ret.graph.create_node();
  auto node_c = ret.graph.create_node();
  auto node_d = ret.graph.create_node();

  ret.graph.set_entity_attr<NodeName>(node_a, NodeName{"a"});
  ret.graph.set_entity_attr<NodeName>(node_b, NodeName{"b"});
  ret.graph.set_entity_attr<NodeName>(node_c, NodeName{"c"});
  ret.graph.set_entity_attr<NodeName>(node_d, NodeName{"d"});

  ret.graph.create_edge(node_a, node_b);
  ret.graph.create_edge(node_c, node_d);

  ret.set_node_label(node_a, gviz::attrtypes::Label<>("This is Node A"));
  ret.set_node_label(node_d, gviz::attrtypes::Label<>("This is Node D"));

  ret.set_node_style(node_a, gviz::attrtypes::CommonStyle::bold);
  ret.set_node_style(node_b, gviz::attrtypes::CommonStyle::dashed);
  ret.set_node_style(node_c, gviz::attrtypes::CommonStyle::dotted);

  return ret;
}

the example above is from examples/simple_dot_generator/main.cpp and executing it after build outputs this:

graph {
    a [label="This is Node A" style=bold];
    b [style=dashed];
    c [style=dotted];
    d [label="This is Node D"];
    b -- a;
    d -- c;
}

What problem does it solve?

the problem of defining a graph to be drawn.

graphviz is considered to be the most notable graph visualization toolset, but there aren’t any c++ library bindings to it nor any alternative.

aside from graphviz there aren’t any c++ graph visualization that isn’t tied to a specific rendering framework, and they usually don’t have an easy/simple api, nor support much customization.

libgvizard provides all of that, and is very flexible by being generic.

it is not tied to any renderer, only provides a set of interfaces and utilities with at least one predefined implementation of its interfaces. thus anyone can make a renderer upon this library’s interface and utilities.

it can be used either for rendering once (like graphviz) or to be rendered in real-time, for example in game engines (e.g. visual scripting and node-based flow/graphics compositions).

How does it work?

There are 5 main components: GvizGraph, Graph, Registry, Attribute, AttributeType.

The word entity refers to node, edge or cluster and is basically just an identifier along with its holden storage for set types that is created/destroyed and its id is used to access its underlying holden storage in Registry.

GvizGraph is what puts all other components together along with helper methods, consisting of Graph data structure itself (which contains its registry) and a registry entity proxy (same registry methods that don’t take an entity id and forward the args along with entity id given at construction to the shared registry container at graph) for each entity type (node, edge, cluster) that hold global values for an Attribute, which according to graphviz is used for an entity when that Attribute isn’t set for an entity of same type.

Graph is the data structure holding the hierarchy of the graph, having nodes, edges (which connect two nodes to each other), and cluster (which can contain zero or more nodes but two clusters never share a node). the Graph is also in charge of holding its Registry and provide access/modifier methods to get/set/remove an Attribute from/to an entity.

Registry is a set of entities each with a unique id mapped to limited or unlimited set of types which are accessed/modifed by a pair of entity id and value type (thus an entity can’t contain two values for same type). this type is commonly an Attribute, but it’s not enforced and it’s up to renderer what types to accepts and how behave about it.

an Attribute supposedly defines appearance of an entity and consists of a name, a default value and the set value, the type of this value is an AttributeType. the type can be variant and accept serveral AttributeTypes. The resulting appearance after render is up to renderer and this library doesn’t guarantee that.

(this library provides all attributes defined in graphviz documentation)

AttributeType is a type that is used by an Attribute to define appearance of an entity. such as Style, Shape or Color.

(this library provides all attrtibute types defined in graphviz documentation)

General utilities:

  • Color: RGB, RGBA, HSV colors and X11Color, SGVColor color schemes. all convertible to each other using Converter utility.

  • Converter: one-way conversion interface/struct where to conversion of type T to type U can be defined by writing a template specification of it and defining convert method which can be overloaded too.

  • EnumHelper: a reflection-like utility to get an enum value by its underlying value, field name, or index. also vice versa.

  • LambdaVisitor: a simple utility that takes one or more lambda (or any struct/class based callable, not function) instances, and makes a single overloaded callable. it comes very handy when dealing with std::variant by std::visit.

  • LambdaVisit: takes a visitable and one or more callables that LambdaVisitor accepts and returns std::visit(LambdaVisitor{ callables... }, visitable).

  • OptionalRef: a std::optional<T&>, taking a T and holding a pointer to its original address or nulloptref to hold a null pointer.

  • Contract: a monad-like single container that holds a value that satisfies a given predicate. used as Contract<T, auto callable_predicate>.

Template Meta-Programming utilites:

  • all_unique<typename ...Types>: checks each type T in Types against all other type U in Types by !std::is_same_v<T, U>.

  • all_unique<typename Base, typename ...Types>:

    all T from Types must satisfy std::is_base_of<Base, T>

  • always<typename T>: having auto f = always<int>{42};, calling f by any argument returns 42, f(x) == 42.

  • Expandable<typename ...Ts>: provides expand_for_t<Consumer, ...Us> which makes the type Consumer<Ts..., Us...>. similarly rexpand_for_t makes Consumer<Us..., Ts...>.

  • TypeInfo<typename T, typename ...Ts>: contains types first = T and constexpr static value rest = TypeInfo<Ts...>{}, along with size = 1 + sizeof...(Ts).

  • MixTypeInfo<typename ...TypeInfos>: combines several TypeInfo<...Ts> into a single TypeInfo. for example MixTypeInfo<TypeInfo<A, B, C>, TypeInfo<D, E>, TypeInfo<F>>::type is same as TypeInfo<A, B, C, D, E, F>.

  • PackDrop<std::size_t I, typename ...Ts>: drops I many types from Ts... and storing the rest as type = TypeInfo<Us...>.

  • find_type_index_in<typename X, typename ...Ts>: finds index of X in Ts... plus 1, having 0 reserved as not found.

  • has_type_in<typename X, typename ...Ts>: checks if X is in Ts....