dot2tex - A Graphviz to LaTeX converter

From source

Download a zip or a tarball from the download page. It is also available on CTAN. Unpack the file to a directory and run python on the setup.py file:

$ python setup.py install

This will create a dot2tex module in your Python module directory and a wrapper script in your SCRIPTS directory. Note that a few warnings will be displayed. You can safely ignore them. The warnings are shown because there is some extra information in the setup.py file that distutils does not understand.

Using easy_install

The easiest way to install dot2tex is to use easy_install:

$ easy_install dot2tex

The command will locate dot2tex and download it automatically. Note that documentation and examples are not installed by default. Easy_install will also create a wrapper script or EXE file for you and install dependencies if necessary.

Binary packages

Binary packages are available for Debian and OpenSUSE.

Development version

The development version of dot2tex is available from a Subversion repository hosted at Google code. To get the code you can use the following command:

svn checkout http://dot2tex.googlecode.com/svn/trunk/ dot2tex

Options

The following options are available:

$ dot2tex --preproc ex1.dot | dot2tex

\rput[alignstr] ...

The following options are used by the output templates.

PGF

This is the default output format. Generates code for the Portable Graphics Format (PGF) package . Mixes both PGF and TikZ commands.

PSTricks

Generates code for the PSTricks package.

TikZ

The tikz output format also uses the PGF and TikZ package. However, it relies on TikZ node and edge mechanisms to draw nodes and edges, instead of using the drawing information provided by Graphviz. This allows much tighter integration with TikZ and in some cases prettier results.

Advantages of the tikz format:

• The generated code is very compact and clean.
• Easy to modify the output.
• Labels will fit inside nodes without preprocessing.
• Full access to the power of PGF and TikZ.

You can find more details in the section: Use the tikz output format for maximum flexibility.

Label examples

Consider the following graph:

digraph G {
    a_1-> a_2 -> a_3 -> a_1;
}

Converting the graph using:

$ dot2tex -tmath ex1.dot > ex1.tex

gives the result shown in the left hand side of the figure below. The default rendering is shown to the right. Using the raw mode will result in a compilation error because of the underscore character.

Example of using texlbl:

digraph G {
    a_1 [texlbl="$\frac{\gamma}{x^2}$"];
    a_1-> a_2 -> a_3 -> a_1;
}

Example of using the texmode attribute:

digraph G {
    a_1 [texlbl="$\frac{\gamma}{2x^2+y^3}$"];
    a_1 -> a_2 -> a_3 -> a_1
    node [texmode="math"];
    a_1 -> b_1 -> b_2 -> a_3;
    b_1 [label="\\frac{\\gamma}{x^2}"];
    node [texmode="verbatim"]
    b_4 [label="\\beta"]
    a_3 -> b_4 -> a_1;
}

The above example shows two important things:

• The backslash \ character needs to be written as \\ in the label attribute.
• Using LaTeX markup in the label attribute gives oversized nodes. A workaround is to use the texlbl attribute, and manually pad the label attribute to an appropriate length. A much better solution is to use the --preproc option.

Preprocessing the above graph with:

$ dot2tex --preproc ex4.dot | dot2tex > ex4.tex

gives correctly sized nodes:

Read more about preprocessing in the Preprocessing graphs section.

Vertical label alignment

Dot2tex relies on the xdot format for drawing nodes and placing node labels. The fonts that Graphviz and LaTeX use are different, so using the label coordinates from Graphviz does not always give good results. Dot2tex's default behavior is to place node labels in the middle of the node. However, you can change this behavior by setting the valignmode option to dot. Labels will then be placed using the coordinates supplied by Graphviz.

Here is an example graph where it is necessary to use the valignmode option:

digraph G {
    node0 [label="{left|right}", shape=record];
    node1 [shape=rectangle, label="node 1"];
    node0 -> node1;
    rankdir=LR;
}

For record nodes dot2tex has to use Graphviz coordinates. This is shown in the following figure rendered with:

$ dot2tex valign.dot

To get the same vertical alignment for both nodes, you can use:

$ dot2tex --valignmode=dot valign.dot

Now the labels are aligned, but the labels are still placed too low. The reason for this is that both PSTricks and PGF by default centers text vertically on the current coordinate. The alignment point should in this case be set to the baseline. For PGF/TikZ you can use the --graphstyle option like this:

$ dot2tex --valignmode=dot --graphstyle="anchor=base" valign.dot

PSTricks users have to use the --alingstr option:

$ dot2tex --valignmode=dot --alignstr=B valign.dot

The result is better, but to get even better alignment you have to change the node font size. Graphviz' default font size is 14pt, which is larger than the typical 10pt or 11pt used in LaTeX documents. By changing the node font size to 10pt we can trick Graphviz to give us a better alignment:

digraph G {
    node [fontsize=10];
    node0 [label="{left|right}", shape=record];
    node1 [shape=rectangle, label="node 1"];
    node0 -> node1;
    rankdir=LR;
}

Examples

Consider the following graph:

digraph G {
    node [shape=circle];
    a_1 [texlbl="$x^2+\frac{\sin y}{y^2+\cos \beta}+\gamma_3$"];
    a_1 -> a_2 [label=" ", texlbl="$x_1+x_3^2+z+c+v~~$"];
    a_2 -> a_1;
}

Rendered with:

$ dot2tex -tmath example.dot > example.tex

the graph will look like this:

You could improve the result by adding a longer label attribute or setting a fixed width. A better solution is to preprocess the graph like this:

$ dot2tex -tmath --preproc example.dot > exampletmp.dot
$ dot2tex exampletmp.dot > example.tex

You can also chain the commands:

$ dot2tex -tmath --preproc example.dot | dot2tex > example.tex

A shorter alternative is:

$ dot2tex -tmath --autosize example.dot > example.tex

The resulting graph now has correctly sized nodes and edge labels:

Modifying node sizes using the widht/height and margin attributes can be a bit counterintuitive. A few examples will hopefully make it clearer:

digraph G {
    node [shape=rectangle];
    a_1 [margin="0"];
    a_2 [margin="0.7,0.4"];
    a_3 [width="2",height="1"];
    a_1-> a_2 -> a_3 -> a_1;
}

Processing the graph with:

$ dot2tex -tmath --preproc example.dot | dot2tex > example.tex

gives

Setting the margin of a_1 to 0 has no effect because of the minimum node width. Processing the graph with:

$ dot2tex -tmath --preproc --nominsize example.dot | dot2tex > example.tex

gives a different graph, where only label widths and margins affect the node sizes:

Using styles

The dot language defines the style attribute that can be used to modify the appearance of graphs, nodes, and edges. The style attribute is passed to the rendering backend, and is a powerful and flexible way of customizing the look and feel of your graphs. Using styles requires detailed knowledge of the output format.

The following example shows how interesting visual results can be achieved with the PGF/TikZ output format. The styles are PGF/TikZ specific. See the user guide for details:

graph G {
    node [shape=circle, fixedsize=True, width="0.2",
          style="ball color =green", label=""];
    edge [style="snake=zigzag, green"];
    a_1 -- c -- a_2;
    c [style="ball color=black"];
    edge [style="snake=snake, blue"];
    node [style="ball color = red", label=""];
    a_3 -- c -- a_4 --a_3;
}

The snake styles only work on straight lines. We therefore have to use the -s option. fdp is used to lay out the graph:

$ fdp -TXdot ball.dot | dot2tex -fpgf -s > balls.tex

The resulting graph is shown below.

Changing arrow types

The style attribute can be used to change arrow types. A PGF/TikZ example:

digraph G {
    graph [mindist=0.5];
    node [fixedsize=true, shape=circle, width=0.4, style="fill=green!20"];
    c -> n_1 [style="-stealth"];
    c -> n_2 [style="-to"];
    c -> n_3 [style="-latex"];
    c -> n_4 [style="-diamond"];
    c -> n_5 [style="-o"];
    c -> n_6 [style="{-]}"];
    c -> n_7 [style="-triangle 90"];
    c -> n_8 [style="-hooks"];
    c -> n_9 [style="->>"];
    c [style="fill=red!80"];
}

Rendered with:

$ circo -Txdot pgfarrows.dot | dot2tex -tmath > pgfarrows.tex

You can also set the default arrow style by using the --graphstyle option or d2tgraphstyle attribute:

$ dot2tex -tmath --graphstyle=">=diamond" ex1.dot > ex1gstyle.tex

A PSTricks example:

digraph G {
    d2tdocpreamble="\usepackage{pstricks-add}";
    graph [mindist=0.5];
    node [texmode="math", fixedsize=true, shape=circle, width=0.4];
    c -> n_1 [style="arrows=->"];
    c -> n_2 [style="arrows=->>"];
    c -> n_3 [style="arrows=-<"];
    c -> n_4 [style="arrows=-*"];
    c -> n_5 [style="arrows=-{]}"];
    c -> n_6 [style="arrows=-o"];
    c -> n_7 [style="arrows=-H"];
}

Rendered with:

$ circo -Txdot pstarrows.dot | dot2tex -fpst > pstarrows.tex

The above example shows how the d2tdocpreamble attribute can be used to load additional LaTeX packages. You could also use the `--docpreamble option:

$ ... | dot2tex -fpst --docpreamble="\usepackage{pstricks-add}" ...

Label styles

Node, edge and graph labels can be styled using the special lblstyle attribute. However, this only works for the pgf and tikz output formats.

Labels are drawn using code like:

\draw (157bp,52bp) node {label};

When you specify a lblstyle attribute, the style will be given as a parameter to the node like this:

\draw (157bp,52bp) node[lblstyle] {label};

Example:

digraph G {
    node [shape=circle];
    a -> b [label="label",lblstyle="draw=red,cross out"];
    b -> c [label="test",lblstyle="below=0.5cm,rotate=20,fill=blue!20"];
    a [label="aa",lblstyle="blue"];
    b [lblstyle="font=\Huge"];
    c [label="ccc", lblstyle="red,rotate=90"];
    label="Graph label";
    lblstyle="draw,fill=red!20";
    rankdir=LR;
}

See the PGF and TikZ documentation for more information about styles.

Note

You can use the exstyle attribute in addition to lblstyle. The difference is that exstyle is ignored in preprocessing mode. Useful when using TikZ' pin and label options and you do not want them to influence the graph layout.

...
\begin{scope}[nodeoptions]
% code for drawing nodes
\end{scope}
\begin{scope}[edgeoptions]
% code for drawing edges
\end{scope}
...

TikZ edge labels

With the --tikzedgelabel option you can bypass the XDOT edge label placement and let PGF and TikZ do the job instead. This can be useful in some cases. However, this only works properly for straight edges and to paths.

Example:

graph G {
    mindist = 0.5;
    node [shape="circle"];
    edge [lblstyle="mystyle"];
    a -- b [label="ab"];
    b -- c [label="bc"];
    c -- a [label="ca"];
}

Without the --tikzedgelabel option the code for placing edges will look something like this:

% Edge: a -- b
\draw [] (28bp,55bp) -- (28bp,75bp);
\draw (40bp,65bp) node[mystyle] {ab};
% Edge: b -- c
\draw [] (51bp,88bp) -- (68bp,78bp);
\draw (66bp,96bp) node[mystyle] {bc};
% Edge: c -- a
\draw [] (69bp,51bp) -- (52bp,41bp);
\draw (53bp,57bp) node[mystyle] {ca};

With the tikzedgelabels option the output is simply:

\draw [] (a) -- node[mystyle] {ab} (b);
\draw [] (b) -- node[mystyle] {bc} (c);
\draw [] (c) -- node[mystyle] {ca} (a);

The placement of the edge labels depends on the options passed to the edge label node (in this case mystyle), and the curve used to connect the nodes. Some examples of mystyle values are shown in the figure below. The leftmost graph is rendered without the tikzedgelabels option.

Limitations:

• Works best with straight edges and to paths
• The headlabel and taillabel attributes are currently not affected by the tikzedgelabels option.

To paths

The topath edge attribute offers a way to override the edges drawn by Graphviz. When a topath attribute is encountered, dot2tex inserts a so called to path operation to connect the nodes. A number of predefined to paths are defined by TikZ, and you can create your own.

Example:

digraph G {
    mindist = 0.5;
    node [shape="circle"];
    a -> b [topath="bend right"];
    c -> b [topath="bend left"];
    c -> a [topath="out=10,in=-90"];
    b -> b [topath="loop above"];
}

Generating the graph with:

$ circo -Txdot topaths1.dot | dot2tex -ftikz > topaths1.tex

yields:

The generated edge drawing code is:

\draw [->] (a) to[bend right] (b);
\draw [->] (c) to[bend left] (b);
\draw [->] (c) to[out=10,in=-90] (a);
\draw [->] (b) to[loop above] (b);

Here is a larger example that uses the automata library:

digraph G {
    d2tdocpreamble = "\usetikzlibrary{automata}";
    d2tfigpreamble = "\tikzstyle{every state}= \
    [draw=blue!50,very thick,fill=blue!20]";
    node [style="state"];
    edge [lblstyle="auto",topath="bend left"];
    A [style="state, initial"];
    A -> B [label=2];
    A -> D [label=7];
    B -> A [label=1];
    B -> B [label=3,topath="loop above"];
    B -> C [label=4];
    C -> F [label=5];
    F -> B [label=8];
    F -> D [label=7];
    D -> E [label=2];
    E -> A [label="1,6"];
    F [style="state,accepting"];
}

Generated with:

neato -Txdot fsm1.dot | dot2tex -ftikz --tikzedgelabels --styleonly

Default PGF/TikZ template

\documentclass{article}
\usepackage[x11names, rgb]{xcolor}
\usepackage[<<textencoding>>]{inputenc}
\usepackage{tikz}
\usetikzlibrary{snakes,arrows,shapes}
\usepackage{amsmath}
<<startpreprocsection>>%
\usepackage[active,auctex]{preview}
<<endpreprocsection>>%
<<gvcols>>%
<<cropcode>>%
<<docpreamble>>%

\begin{document}
\pagestyle{empty}
%
<<startpreprocsection>>%
<<preproccode>>
<<endpreprocsection>>%
%
<<startoutputsection>>
\enlargethispage{100cm}
% Start of code
% \begin{tikzpicture}[anchor=mid,>=latex',join=bevel,<<graphstyle>>]
\begin{tikzpicture}[>=latex',join=bevel,<<graphstyle>>]
\pgfsetlinewidth{1bp}
<<figpreamble>>%
<<drawcommands>>
<<figpostamble>>%
\end{tikzpicture}
% End of code
<<endoutputsection>>
%
\end{document}
%
<<startfigonlysection>>
\begin{tikzpicture}[>=latex,join=bevel,<<graphstyle>>]
\pgfsetlinewidth{1bp}
<<figpreamble>>%
<<drawcommands>>
<<figpostamble>>%
\end{tikzpicture}
<<endfigonlysection>>

The <<cropcode>> template tag is available when the --preview option is used. The contents will then be:

\usepackage[active,tightpage]{preview}
\PreviewEnvironment{tikzpicture}
\setlength\PreviewBorder{<<margin>>}

Default pstricks template

\documentclass{article}
% <<bbox>>
\usepackage[x11names]{xcolor}
\usepackage[<<textencoding>>]{inputenc}
\usepackage{graphicx}
\usepackage{pstricks}
\usepackage{amsmath}
<<startpreprocsection>>%
\usepackage[active,auctex]{preview}
<<endpreprocsection>>%
<<gvcols>>%
<<docpreamble>>%


\begin{document}
\pagestyle{empty}
<<startpreprocsection>>%
<<preproccode>>%
<<endpreprocsection>>%
<<startoutputsection>>%
\enlargethispage{100cm}

% Start of code
\begin{pspicture}[linewidth=1bp<<graphstyle>>]<<bbox>>
\pstVerb{2 setlinejoin} % set line join style to 'mitre'
<<figpreamble>>%
<<drawcommands>>
<<figpostamble>>%
\end{pspicture}
% End of code
<<endoutputsection>>%
\end{document}
%
<<startfigonlysection>>
\begin{pspicture}[linewidth=1bp<<graphstyle>>]<<bbox>>
\pstVerb{2 setlinejoin} % set line join style to 'mitre'
<<figpreamble>>%
<<drawcommands>>
<<figpostamble>>%
\end{pspicture}
<<endfigonlysection>>

Debugging

You can set debug=True to create a detailed log useful for debugging. To retrieve the content of the log you can use the get_logstream function. It will return a StringIO instance. You can then use the getvalue() class method to get the actual text. Example:

import dot2tex
texcode = dot2tex.dot2tex(testgraph, debug=True)
logstream = dot2tex.get_logstream()
print logstream.getvalue()

Text encoding

Graphviz's default text encoding is utf8. The latin1 encoding can also be used. Utf8 is an unicode encoding and can in theory handle any international characters. However, LaTeX's unicode support is somewhat limited.

Fonts

No font information in the DOT file is preserved by dot2tex. However, there are several ways of modifying the generated LaTeX code to achieve some control of fonts and font sizes.

• Modifying the templates.
• Using the d2tdocpreamble and d2tfigpreamble attributes or command line options.
• Using the lblstyle attribute.

To increase the font size you can for instance insert a \Huge command in the figure preamble:

$ dot2tex -tmath --figpreamble="\Huge" ex1.dot > ex1huge.tex

Debugging

When making your own templates it is easy to make mistakes, and LaTeX markup in graphs may fail to compile. To make it easier to find errors, invoke dot2tex with the --debug option:

$ dot2tex --preproc --debug test.dot

A dot2tex.log file will then be generated with detailed information. In the log file you will find the generated LaTeX code, as well as well as the compilation log.

Be consistent

Be aware of differences between the template you use for preprocessing and code used to generate final output. This is especially important if you use the --figonly option and include the code in a master document. If a 10pt font is used during preprocessing, the result may not be optimal if a 12pt font is used in the final output.

Example. A graph is generated with:

$ dot2tex --preproc -tmath --nominsize ex1.dot > ex1tmp.dot

Running through dot2tex again with:

$ dot2tex --figpreamble="\Huge" ex1tmp.dot > ex1huge.tex

gives labels that do not fit properly inside the nodes.

Postprocessing

The output from Graphviz and dot2tex is not perfect. Manual adjustments are sometimes necessary to get the right results for use in a publication. For final and cosmetic adjustments, it is often easier to edit the generated code than to hack the dot source. This is especially true when using the tikz output format.

Use the special graph attributes

Dot2tex has many options for customizing the output. Sometimes is is impractical or boring to type the various options at the command line each time you want to create the graph. To avoid this, you can use the special graph attributes. The d2toptions attribute is handy because it is interpreted as command line options.

So instead of typing:

$ dot2tex -tikz -tmath --tikzedgelabels ex1.dot

each time, use d2toptions like this:

digraph G {
    d2toptions ="-tikz -tmath --tikzedgelabels";
    ...
}

Use the tikz output format for maximum flexibility

The difference between the pgf and tikz output formats is best shown with an example. Consider the following graph:

graph G {
    mindist = 0.5;
    node [shape=circle];
    a -- b -- c -- a;
}

Rendering the graph using circo and the pgf and tikz output formats:

$ circo -Txdot simple.dot | dot2tex -tmath -fpgf -s
$ circo -Txdot simple.dot | dot2tex -tmath -ftikz -s

gives visually different graphs:

However, the main difference is in the generated code. Here is the pgf output:

% Edge: a -- b
\draw [] (19bp,38bp) -- (19bp,60bp);
% Edge: b -- c
\draw [] (35bp,70bp) -- (55bp,58bp);
% Edge: c -- a
\draw [] (55bp,40bp) -- (35bp,28bp);
% Node: a
\begin{scope}
\pgfsetstrokecolor{black}
\draw (19bp,19bp) ellipse (18bp and 19bp);
\draw (19bp,19bp) node {$a$};
\end{scope}
% Node: b
\begin{scope}
\pgfsetstrokecolor{black}
\draw (19bp,79bp) ellipse (18bp and 19bp);
\draw (19bp,79bp) node {$b$};
\end{scope}
% Node: c
\begin{scope}
\pgfsetstrokecolor{black}
\draw (71bp,49bp) ellipse (18bp and 19bp);
\draw (71bp,49bp) node {$c$};
\end{scope}

Compare the above code with the tikz output:

\node (a) at (19bp,19bp) [draw,circle,] {$a$};
\node (b) at (19bp,79bp) [draw,circle,] {$b$};
\node (c) at (71bp,49bp) [draw,circle,] {$c$};
\draw [] (a) -- (b);
\draw [] (b) -- (c);
\draw [] (c) -- (a);

The code is much more compact and it is quite easy to modify the graph.

The dot2texi LaTeX package

The dot2texi package allows you to embed DOT graphs directly in you LaTeX document. The package will automatically run dot2tex for you and include the generated code. Example:

\documentclass{article}
\usepackage{dot2texi}

\usepackage{tikz}
\usetikzlibrary{shapes,arrows}

\begin{document}
\begin{dot2tex}[neato,options=-tmath]
digraph G {
    node [shape="circle"];
    a_1 -> a_2 -> a_3 -> a_4 -> a_1;
    }
\end{dot2tex}

\end{document}

When the above code is run through LaTeX, the following will happen is shell escape is enabled:

• The graph is written to file.
• dot2tex is run on the DOT file.
• The generated code is included in the document.

The whole process is completely automated. The generated graph will look like this:

The codeonly option is useful in conjunction with dot2texi, especially when used with the tikz output format. Here is an example that shows how to annotate a graph:

\documentclass{article}
\usepackage{tikz}
\usetikzlibrary{arrows,shapes}
\usepackage{dot2texi}
\begin{document}
% Define layers
\pgfdeclarelayer{background}
\pgfdeclarelayer{foreground}
\pgfsetlayers{background,main,foreground}

% The scale option is useful for adjusting spacing between nodes.
% Note that this works best when straight lines are used to connect
% the nodes.
\begin{tikzpicture}[>=latex',scale=0.8]
    % set node style
    \tikzstyle{n} = [draw,shape=circle,minimum size=2em,
                        inner sep=0pt,fill=red!20]
    \begin{dot2tex}[dot,tikz,codeonly,styleonly,options=-s -tmath]
        digraph G  {
            node [style="n"];
            A_1 -> B_1; A_1 -> B_2; A_1 -> B_3;
            B_1 -> C_1; B_1 -> C_2;
            B_2 -> C_2; B_2 -> C_3;
            B_3 -> C_3; B_3 -> C_4;
        }
    \end{dot2tex}
    % annotations
    \node[left=1em] at (C_1.west)  (l3) {Level 3};
    \node at (l3 |- B_1) (l2){Level 2};
    \node at (l3 |- A_1) (l1) {Level 1};
    % Draw lines to separate the levels. First we need to calculate
    % where the middle is.
    \path (l3) -- coordinate (l32) (l2) -- coordinate (l21) (l1);
    \draw[dashed] (C_1 |- l32) -- (l32 -| C_4);
    \draw[dashed] (C_1 |- l21) -- (l21 -| C_4);
    \draw[<->,red] (A_1) to[out=-120,in=90] (C_2);
    % Highlight the A_1 -> B_1 -> C_2 path. Use layers to draw
    % behind everything.
    \begin{pgfonlayer}{background}
        \draw[rounded corners=2em,line width=3em,blue!20,cap=round]
                (A_1.center) -- (B_1.west) -- (C_2.center);
    \end{pgfonlayer}
\end{tikzpicture}
\end{document}