Creating this tree structure with LaTeX

Question

I would like to easily create tree-like structures like

D─┬───────────┬───A
  R─┬───A─┬─A R─┐  
    R─┐   R     R  
      R 

D is the top root node. It starts a horizontal line to the right, which stands for the "same level". This horizontal line can have more letters A threaded at any position, which do not have additional semantic meaning. These letters A always have a line incoming from the left.

The more interesting feature is that a horizontal line can split downwards to a child (base) node R which always have a line incoming from the top. This split can happen at multiple positions on the horizontal baseline. The child nodes recursively start a new virtual block and hence are usually continued by an outgoing horizontal line to the right.- Hence, the R next to the D effectively starts the whole thing again in its own "bounding box" (here marked with +).

D─┬───────────────┬───A
  +++++++++++     R─┐
 +R─┬───A─┬─A+      R 
 +  R─┐   R  +     
 +    R      + 
  +++++++++++

If for a node there are no more nodes connected to the right, then the horizontal outgoing line is not shown. And of course tree/text overlap needs to be avoided, i.e. some horizontal lines need to stretch far enough before starting a new child node. A type A node also should only appear after all previous child blocks have completed to the left of them.

What would be the best way to create such a tree-like drawing in LaTeX?

I've found \usetikzlibrary{trees} which goes in this direction, but I'm not sure how to flip the example, how to make multiple letters on a horizontal line, how to control text overlap, etc...

Any good suggestion how to recreate that diagram?

There is a working answer with manually positioning everything, but I hope there is another solution like tikz-tree which can calculate the positioning automatically for me, so that I only specify the child-node relations.

With arbitrary and longer labels:

D─┬───────────────────────G─┬─────C
  Cc─┬─────────Aa─┬──Aaaa   Eee─┐  
     Bbb─┐        Dd            A  
         Aaaaa 

Answer 1

Variant with parameter for label

Adaptations

• see original variant below for more description
• \nodeD, \nodeR and \nodeA each have now one parameter to set the text.
• Here I used relative positioning instead of absolute. Therefore, different text lengths are considered.
• I set base node/.style={draw, dashed} just for clarification

Result

Code

\documentclass[border=2mm]{standalone}

\usepackage{etoolbox}
\usepackage{tikz}
\usetikzlibrary{positioning}

\def\nodedistance{5mm}
\def\treeY{0}

\newcounter{treeRnode}
\setcounter{treeRnode}{0}
\newcounter{treeAnode}
\setcounter{treeAnode}{0}

\def\lastXnode{}

\newcommand{\nodeD}[1]{%
    \node[D] (D0) at (0,0) {#1};
    \csdef{treeLastY\treeY}{D0}
    \xdef\lastXnode{D0}
}

\newcommand{\nodeR}[1]{%
    \pgfmathtruncatemacro\lastY{\treeY}
    \pgfmathtruncatemacro\treeY{\treeY+1}
    \ifcsdef{treeLastY\treeY}{%
        \node[R, anchor=west] (R\arabic{treeRnode}) at ([xshift=\nodedistance] \csuse{treeLastY\treeY}.east-|\lastXnode.east) {#1};
    }{%
        \node[R, below right=of \csuse{treeLastY\lastY}] (R\arabic{treeRnode}) {#1};
    }
    \draw (\csuse{treeLastY\lastY}) -| (R\arabic{treeRnode});
    \csxdef{treeLastY\treeY}{R\arabic{treeRnode}}
    \xdef\lastXnode{R\arabic{treeRnode}}
    \stepcounter{treeRnode}
}

\newcommand{\nodeA}[1]{%
    \node[A, anchor=west] (A) at ([xshift=\nodedistance] \csuse{treeLastY\treeY}.east-|\lastXnode.east) (A\arabic{treeAnode}) {#1};
    \draw (\csuse{treeLastY\treeY}) -- (A\arabic{treeAnode});
    \csxdef{treeLastY\treeY}{A\arabic{treeAnode}}
    \xdef\lastXnode{A\arabic{treeAnode}}
    \stepcounter{treeAnode}
}

\newcommand{\branchend}{%
    \pgfmathtruncatemacro\treeY{\treeY-1}
}

\begin{document}

\begin{tikzpicture}[
    node distance=\nodedistance,
    base node/.style={draw, dashed},
    D/.style={base node},
    R/.style={base node},
    A/.style={base node},
]
    \nodeD{$D_1$ with longer text}
        \nodeR{$R_1$}
            \nodeR{$R_2$ with longer text}
                \nodeR{$R_3$}
                \branchend
            \branchend
        \nodeA{$A_1$}
            \nodeR{$R_4$}
            \branchend
        \nodeA{$A_2$ with longer text}
        \branchend
        %
        \nodeR{$R_5$}
            \nodeR{$R_6$}
            \branchend
        \branchend
    \nodeA{$A_3$}
\end{tikzpicture}

\end{document}

Variant without parameters (original)

Description

I defined some commands that can be used within tikz:

• \nodeD, \nodeR and \nodeA for the different node types
• \branchend to stop a branch and go up one level.

I used counters treeRnode and treeAnode to name the nodes and variables \treeX and \treeY to save the current position. Futhermore, \treeLastY<y> saves the last node in layer <y>.

So the code for the given tree is in my syntax:

\nodeD
    \nodeR
        \nodeR
            \nodeR
            \branchend
        \branchend
    \nodeA
        \nodeR
        \branchend
    \nodeA
    \branchend
    %
    \nodeR
        \nodeR
        \branchend
    \branchend
\nodeA

The indentation in the source code is just for readability. So, this would be the same:

\nodeD\nodeR\nodeR\nodeR\branchend\branchend\nodeA\nodeR\branchend\nodeA\branchend\nodeR\nodeR\branchend\branchend\nodeA

Result

Code

\documentclass{article}

\usepackage{etoolbox}
\usepackage{tikz}

\begin{document}

\def\treeX{0}
\def\treeY{0}

\newcounter{treeRnode}
\setcounter{treeRnode}{0}
\newcounter{treeAnode}
\setcounter{treeAnode}{0}

\newcommand{\nodeD}{%
    \node[D] (D0) at (0,0) {D};
    \csdef{treeLastY\treeY}{D0}
}

\newcommand{\nodeR}{%
    \pgfmathtruncatemacro\treeX{\treeX+1}
    \pgfmathtruncatemacro\lastY{\treeY}
    \pgfmathtruncatemacro\treeY{\treeY+1}
    \node[R] (R\arabic{treeRnode}) at (\treeX, -\treeY) {R};
    \draw (\csuse{treeLastY\lastY}) -| (R\arabic{treeRnode});
    \csxdef{treeLastY\treeY}{R\arabic{treeRnode}}
    \stepcounter{treeRnode}
}

\newcommand{\nodeA}{%
    \pgfmathtruncatemacro\treeX{\treeX+1}
    \node[A] (A) (A\arabic{treeAnode}) at (\treeX, -\treeY) {A};
    \draw (\csuse{treeLastY\treeY}) -- (A\arabic{treeAnode});
    \csxdef{treeLastY\treeY}{A\arabic{treeAnode}}
    \stepcounter{treeAnode}
}

\newcommand{\branchend}{%
    \pgfmathtruncatemacro\treeY{\treeY-1}
}

\begin{tikzpicture}[
    base node/.style={},
    D/.style={base node},
    R/.style={base node},
    A/.style={base node},
]
    \nodeD
        \nodeR
            \nodeR
                \nodeR
                \branchend
            \branchend
        \nodeA
            \nodeR
            \branchend
        \nodeA
        \branchend
        %
        \nodeR
            \nodeR
            \branchend
        \branchend
    \nodeA
\end{tikzpicture}

\end{document}

Answer 2

TikZ is a multi-purpose drawing system with \draw and \node (a coordinate is a special node). Here is my suggestion. The code is self-explained (a bit care on + and ++).

\documentclass[tikz,border=5mm]{standalone}
\begin{document}
\begin{tikzpicture}
\path 
(0,0) node[left]  (D) {D}
(5,0) node[right] (A) {A}
(4,0)  coordinate (P1) 
++(0,-.5)    node (P2) {R}
++(.5,-.5)   node (P3) {R}
(.5,0) coordinate (T1)
++(0,-.5)    node (T2) {R}
++(.5,-.5)   node (T3) {R}
++(.5,-.5)   node (T4) {R}
(T2)
+(2.5,0)  node (A2) {A}
+(1.5,0)  node (A1) {A}
+(2,-.5)  node (A3) {R}
;
\draw[thick,magenta] (D)--(A) 
(P1)--(P2)-|(P3)
(T1)--(T2)-|(T3)-|(T4)
(T2)--(A1)--(A2)-|(A3)
;
\end{tikzpicture}
\end{document}

Answer 3

Here's my approach with forest.

Rules:

1. One empty root.

2. The nodes on the same visual level are siblings.

3. Should nodes connect to their previous sibling use connect left.

   If you have to customize all edges or want to place nodes along these edges more work might be necessary.

4. Note: Some line segments are drawn multiple times when connect left connects to a sibling that has children on its own.

   In a proper PDF viewer this shouldn't be noticable, less so on paper. (Don't trust the picture I added to the answer.)

The gere tree style install a few things:

1. The parent is a coordinate not a node (not really necessary but why bother).

   The edges to its children are hidden and its children have no l meaning there's no vertical space above the first visual row.

2. The first children (except the one in the first row) get a previous sibling that gets hidden. This shifts the first child to the right without messing with the ls coordinate system.

3. fit = rectangle implements your

   And of course tree/text overlap needs to be avoided, i.e. some horizontal lines need to stretch far enough before starting a new child node. A type A node also should only appear after all previous child blocks have completed to the left of them.

4. The edge path' in for tree

   edge path'={
     ([xshift={\pgfkeysvalueof{/pgf/inner xsep}+.4em}].north west) |- (!u)}
   

   sets all edges to a (child) |- (parent) path (meaning up from the children and then orthogonally to the parent which will always be to the left).

   It is now setup in such a way that it connects the children roughly above the first (capital) letter. For hitting the true horizontal center of the first letter much more work is needed.

   If the line should hit the children centered across the whole width, simply use

   edge path' = {() |- (!u)}
   

Code

\documentclass[tikz,convert]{standalone}
\usepackage{forest}
\forestset{
  connect left/.style={% don't connect to parent but to previous sibling
    tikz+={\draw()--(!p);}, no edge},
  gere tree/.style={
    shape=coordinate,               % parent hidden
    for tree={
      calign=first, fit=rectangle,
      edge path'={
        ([xshift={\pgfkeysvalueof{/pgf/inner xsep}+.4em}].north west)|-(!u)}},
    for children={
      no edge,                      % parent hidden
      before computing xy={l=+0pt}, % parent hidden
      where n children>={0}{prepend={[,phantom]}}{},% invisible sibling at first
    }
  }
}
\begin{document}
\begin{forest} gere tree
[
  [D
    [Cc
      [Bbb
        [Aaaaa]]]
    [Aa, connect left
      [Dd]]
    [Aaaa, connect left]
  ]
  [G, connect left
    [Eee
      [A]]
  ]
  [C, connect left]
]
\end{forest}
\end{document}

Output