graphviz-2999.12.0.3: Bindings to Graphviz for graph visualisation.




Four different representations of Dot graphs are available, all of which are based loosely upon the specifications at: The DotRepr class provides a common interface for them (the PrintDotRepr, ParseDotRepr and PPDotRepr classes are used until class aliases are implemented).

Every representation takes in a type parameter: this indicates the node type (e.g. DotGraph Int is a Dot graph with integer nodes). Sum types are allowed, though care must be taken when specifying their ParseDot instances if there is the possibility of overlapping definitions. The GraphID type is an existing sum type that allows textual and numeric values.

If you require using more than one Dot representation, you will most likely need to import at least one of them qualified, as they typically all use the same names.

As a comparison, all four representations provide how you would define the following Dot graph (or at least one isomorphic to it) (the original of which can be found at Note that in all the examples, they are not necessarily done the best way (variables rather than repeated constants, etc.); they are just there to provide a comparison on the structure of each representation.

 digraph G {

 	subgraph cluster_0 {
 		node [style=filled,color=white];
 		a0 -> a1 -> a2 -> a3;
 		label = "process #1";

 	subgraph cluster_1 {
 		node [style=filled];
 		b0 -> b1 -> b2 -> b3;
 		label = "process #2";
 	start -> a0;
 	start -> b0;
 	a1 -> b3;
 	b2 -> a3;
 	a3 -> a0;
 	a3 -> end;
 	b3 -> end;

 	start [shape=Mdiamond];
 	end [shape=Msquare];

Each representation is suited for different things:

is ideal for converting other graph-like data structures into Dot graphs (the Data.GraphViz module provides some functions for this). It is a structured representation of Dot code.
matches the actual structure of Dot code. As such, it is suited for parsing in existing Dot code.
provides graph operations for manipulating Dot graphs; this is suited when you want to edit existing Dot code. It uses generalised Dot graphs for parsing and canonical Dot graphs for printing.
is a much easier representation to use when defining relatively static Dot graphs in Haskell code, and looks vaguely like actual Dot code if you squint a bit.

Please also read the limitations section at the end for advice on how to properly use these Dot representations.



class Ord n => DotRepr dg n whereSource

This class is used to provide a common interface to different ways of representing a graph in Dot form.

You will most probably not need to create your own instances of this class.

The type variable represents the current node type of the Dot graph, and the Ord restriction is there because in practice most implementations of some of these methods require it.


fromCanonical :: DotGraph n -> dg nSource

Convert from a graph in canonical form. This is especially useful when using the functions from Data.GraphViz.Algorithms.

getID :: dg n -> Maybe GraphIDSource

Return the ID of the graph.

setID :: GraphID -> dg n -> dg nSource

Set the ID of the graph.

graphIsDirected :: dg n -> BoolSource

Is this graph directed?

setIsDirected :: Bool -> dg n -> dg nSource

Set whether a graph is directed or not.

graphIsStrict :: dg n -> BoolSource

Is this graph strict? Strict graphs disallow multiple edges.

setStrictness :: Bool -> dg n -> dg nSource

A strict graph disallows multiple edges.

mapDotGraph :: (Ord n', DotRepr dg n') => (n -> n') -> dg n -> dg n'Source

Change the node values. This function is assumed to be injective, otherwise the resulting graph will not be identical to the original (modulo labels).

graphStructureInformation :: dg n -> (GlobalAttributes, ClusterLookup)Source

Return information on all the clusters contained within this DotRepr, as well as the top-level GraphAttrs for the overall graph.

nodeInformation :: Bool -> dg n -> NodeLookup nSource

Return information on the DotNodes contained within this DotRepr. The Bool parameter indicates if applicable NodeAttrs should be included.

edgeInformation :: Bool -> dg n -> [DotEdge n]Source

Return information on the DotEdges contained within this DotRepr. The Bool parameter indicates if applicable EdgeAttrs should be included.

unAnonymise :: dg n -> dg nSource

Give any anonymous sub-graphs or clusters a unique identifier (i.e. there will be no Nothing key in the ClusterLookup from graphStructureInformation).


class (DotRepr dg n, PrintDot (dg n)) => PrintDotRepr dg n Source

This class exists just to make type signatures nicer; all instances of DotRepr should also be an instance of PrintDotRepr.

class (DotRepr dg n, ParseDot (dg n)) => ParseDotRepr dg n Source

This class exists just to make type signatures nicer; all instances of DotRepr should also be an instance of ParseDotRepr.

class (PrintDotRepr dg n, ParseDotRepr dg n) => PPDotRepr dg n Source

This class exists just to make type signatures nicer; all instances of DotRepr should also be an instance of PPDotRepr.


Common sub-types

data GraphID Source

A polymorphic type that covers all possible ID values allowed by Dot syntax. Note that whilst the ParseDot and PrintDot instances for String will properly take care of the special cases for numbers, they are treated differently here.


Str Text 
Int Int 
Dbl Double 

data GlobalAttributes Source

Represents a list of top-level list of Attributes for the entire graph/sub-graph. Note that GraphAttrs also applies to DotSubGraphs.

Note that Dot allows a single Attribute to be listen on a line; if this is the case then when parsing, the type of Attribute it is determined and that type of GlobalAttribute is created.




attrs :: Attributes


attrs :: Attributes


attrs :: Attributes

data DotNode n Source

A node in DotGraph.




Functor DotNode 
Eq n => Eq (DotNode n) 
Ord n => Ord (DotNode n) 
Read n => Read (DotNode n) 
Show n => Show (DotNode n) 
ParseDot n => ParseDot (DotNode n) 
PrintDot n => PrintDot (DotNode n) 

data DotEdge n Source

An edge in DotGraph.




fromNode :: n
toNode :: n
edgeAttributes :: Attributes


Functor DotEdge 
Eq n => Eq (DotEdge n) 
Ord n => Ord (DotEdge n) 
Read n => Read (DotEdge n) 
Show n => Show (DotEdge n) 
ParseDot n => ParseDot (DotEdge n) 
PrintDot n => PrintDot (DotEdge n) 

Helper types for looking up information within a DotRepr.

type ClusterLookup = Map (Maybe GraphID) ([Path], GlobalAttributes)Source

The available information for each cluster; the [Path] denotes all locations where that particular cluster is located (more than one location can indicate possible problems).

type NodeLookup n = Map n (Path, Attributes)Source

The available information on each DotNode (both explicit and implicit).

type Path = Seq (Maybe GraphID)Source

The path of clusters that must be traversed to reach this spot.

Obtaining the DotNodes and DotEdges.

graphNodes :: DotRepr dg n => dg n -> [DotNode n]Source

Returns all resultant DotNodes in the DotRepr (not including NodeAttrs).

graphEdges :: DotRepr dg n => dg n -> [DotEdge n]Source

Returns all resultant DotEdges in the DotRepr (not including EdgeAttrs).

Printing and parsing a DotRepr.

printDotGraph :: PrintDotRepr dg n => dg n -> TextSource

The actual Dot code for an instance of DotRepr. Note that it is expected that parseDotGraph . printDotGraph == id (this might not be true the other way around due to un-parseable components).

parseDotGraph :: ParseDotRepr dg n => Text -> dg nSource

Parse a limited subset of the Dot language to form an instance of DotRepr. Each instance may have its own limitations on what may or may not be parseable Dot code.

Also removes any comments, etc. before parsing.

Limitations and documentation

Printing of Dot code is done as strictly as possible, whilst parsing is as permissive as possible. For example, if the types allow it then "2" will be parsed as an Int value. Note that quoting and escaping of textual values is done automagically.

A summary of known limitations/differences:

  • When creating GraphID values for graphs and sub-graphs, you should ensure that none of them have the same printed value as one of the node identifiers values to avoid any possible problems.
  • If you want any GlobalAttributes in a sub-graph and want them to only apply to that sub-graph, then you must ensure it does indeed have a valid GraphID.
  • All sub-graphs which represent clusters should have unique identifiers (well, only if you want them to be generated sensibly).
  • If eventually outputting to a format such as SVG, then you should make sure to specify an identifier for the overall graph, as that is used as the title of the resulting image.
  • Whilst the graphs, etc. are polymorphic in their node type, you should ensure that you use a relatively simple node type (that is, it only covers a single line, etc.).
  • Also, whilst Graphviz allows you to mix the types used for nodes, this library requires/assumes that they are all the same type (but you can use a sum-type).
  • DotEdge defines an edge (a, b) (with an edge going from a to b); in Dot parlance the edge has a head at a and a tail at b. Care must be taken when using the related Head* and Tail* Attributes. See the differences section in Data.GraphViz.Attributes for more information.
  • It is common to see multiple edges defined on the one line in Dot (e.g. n1 -> n2 -> n3 means to create a directed edge from n1 to n2 and from n2 to n3). These types of edge definitions are parseable; however, they are converted to singleton edges.
  • It is not yet possible to create or parse edges with subgraphs/clusters as one of the end points.
  • The parser will strip out comments and pre-processor lines, join together multiline statements and concatenate split strings together. However, pre-processing within HTML-like labels is currently not supported.
  • Graphviz allows a node to be "defined" twice (e.g. the actual node definition, and then in a subgraph with extra global attributes applied to it). This actually represents the same node, but when parsing they will be considered as separate DotNodes (such that graphNodes will return both "definitions"). canonicalise from Data.GraphViz.Algorithms can be used to fix this.

See Data.GraphViz.Attributes for more limitations.