Joining corners of a node and matrix

Question

\usepackage{tikz,pgfplots}
\usepackage{tikz}
\usetikzlibrary{positioning,shapes,arrows,shadows,patterns}
\tikzset{
 mymat/.style={
    matrix of math nodes,
    text height=2.5ex,
    text depth=0.75ex,
    text width=3.25ex,
    align=center,
    column sep=-\pgflinewidth
},
mymats/.style={
    mymat,
    nodes={draw,fill=#1}
}  
}
\begin{document}
\begin{figure}
\centering
\begin{tikzpicture}[>=latex]
\matrix[mymat,anchor=west,row 2/.style={nodes=draw}]
at (0,0) 
(mat1)
{   
     & \\
     N1 \\
};

\matrix[mymat,right=of mat1,row 2/.style={nodes={draw,fill=gray!30}}]
(mat2)
{
     & \\
    t_2 & t_4 \\
};

\matrix[mymat,anchor=west,row 2/.style={nodes=draw}]
at (0,-1) 
(mat3)
{   
& \\
N2 \\
};

\matrix[mymat,right=of mat3,row 2/.style={nodes={draw,fill=gray!30}}]
(mat4)
{
&  &  \\
t_1 & t_3 & t_4  \\
};
\matrix[mymat,anchor=west,row 2/.style={nodes=draw}]
at (0,-2) 
(mat5)
{   
& \\
N3 \\
};

\matrix[mymat,right=of mat5,row 2/.style={nodes={draw,fill=gray!30}}]
(mat6)
{
& &\\
t_1 & t_2 & t_3 \\
};
\matrix[mymat,anchor=west,row 2/.style={nodes=draw}]
at (0,-3) 
(mat7)
{   
    & \\
    N4 \\
};

\matrix[mymat,right=of mat7,row 2/.style={nodes={draw,fill=gray!30}}]
(mat8)
{
& \\
t_3 & t_4  \\
};
\matrix[mymat,anchor=west,row 2/.style={nodes=draw}]
at (0,-4) 
(mat9)
{   
& \\
N5 \\
};

\matrix[mymat,right=of mat9,row 2/.style={nodes={draw,fill=gray!30}}]
(mat10)
{
& \\
t_1 & t_3  \\
};

\matrix[mymat, anchor=west,row 2/.style={nodes={draw,fill=gray!30}}]
at (1.65,-5.5) 
(mat11)
{   
& & & \\
t_1 & t_2 & t_3 & t_4\\
};
\node[ draw, red, dashed, line width=0.7pt, fit=(mat2)(mat4)(mat6)(mat8) 
(mat10)](node1){};
\begin{scope}[shorten <= -2pt]
\draw[*->]
(mat1-2-1.east) -- (mat2-2-1.west);
\draw[*->]
(mat3-2-1.east) -- (mat4-2-1.west);
\draw[*->]
(mat5-2-1.east) -- (mat6-2-1.west);
\draw[*->]
(mat7-2-1.east) -- (mat8-2-1.west);
\draw[*->]
(mat9-2-1.east) -- (mat10-2-1.west);

\end{scope}
\draw[green, dashed](node1.south west) -- (mat11.north west);
\draw[green, dashed](node1.south east) -- (mat11.north east);

\end{tikzpicture} 
\end{figure}
\end{document}

Unable to join the corners of node1 and mat11 with green dashed lines.

Part of the code is taken from here

Solution provided here is elegant though, didn't work for this figure.

Your example code is missing two libraries by the way, fit and matrix, as well as a \documentclass. And of course, it's not necessary to load tikz twice. You're actually loading it three times, as pgfplots loads tikz as well. — Torbjørn T., May 02 '18 at 10:19

Torbjørn T. · Accepted Answer · 2018-05-02T10:47:47.093

If I understand correctly, I think your problem comes from the fact that a \matrix is just a special type of \node, and like all nodes it has a (by default non-zero) inner sep. You can see that if you add draw to the options of mat11, so that you have

\matrix[mymat, draw, anchor=west,row 2/.style={nodes={draw,fill=gray!30}}]
at (1.65,-5.5) 
(mat11)
{   
& & & \\
t_1 & t_2 & t_3 & t_4\\
};

which gives this:

So the lines you've drawn goes to the corners of the outer border, not the corners of the nodes.

I can think of two ways around this:

Draw the lines to mat11-2-1.north west and mat11-2-4.north east instead of mat11.north west/mat11.north east. This way the lines point to the nodes inside the matrix, instead of the matrix itself.
```
\draw[green, dashed](node1.south west) -- (mat11-2-1.north west);
\draw[green, dashed](node1.south east) -- (mat11-2-4.north east);
```

Set the inner sep of mat11 to 0, but then you have to set the inner sep of the nodes back to 0.333em, which is the default value:

\matrix[mymat, inner sep=0, anchor=west,row 2/.style={nodes={draw,fill=gray!30,inner sep=0.333em}}]
at (1.65,-5.5) 
(mat11)
{   
& & & \\
t_1 & t_2 & t_3 & t_4\\
};

Complete code with both methods:

\documentclass{article}
\usepackage{tikz}
\usetikzlibrary{matrix,fit,positioning,arrows}
\tikzset{
 mymat/.style={
    matrix of math nodes,
    text height=2.5ex,
    text depth=0.75ex,
    text width=3.25ex,
    align=center,
    column sep=-\pgflinewidth,
},
mymats/.style={
    mymat,
    nodes={draw,fill=#1}
}  
}
\begin{document}
\begin{tikzpicture}[>=latex]
\matrix[mymat,anchor=west,row 2/.style={nodes=draw}]
at (0,0) 
(mat1)
{   
     & \\
     N1 \\
};

\matrix[mymat,right=of mat1,row 2/.style={nodes={draw,fill=gray!30}}]
(mat2)
{
     & \\
    t_2 & t_4 \\
};

\matrix[mymat,anchor=west,row 2/.style={nodes=draw}]
at (0,-1) 
(mat3)
{   
& \\
N2 \\
};

\matrix[mymat,right=of mat3,row 2/.style={nodes={draw,fill=gray!30}}]
(mat4)
{
&  &  \\
t_1 & t_3 & t_4  \\
};
\matrix[mymat,anchor=west,row 2/.style={nodes=draw}]
at (0,-2) 
(mat5)
{   
& \\
N3 \\
};

\matrix[mymat,right=of mat5,row 2/.style={nodes={draw,fill=gray!30}}]
(mat6)
{
& &\\
t_1 & t_2 & t_3 \\
};
\matrix[mymat,anchor=west,row 2/.style={nodes=draw}]
at (0,-3) 
(mat7)
{   
    & \\
    N4 \\
};

\matrix[mymat,right=of mat7,row 2/.style={nodes={draw,fill=gray!30}}]
(mat8)
{
& \\
t_3 & t_4  \\
};
\matrix[mymat,anchor=west,row 2/.style={nodes=draw}]
at (0,-4) 
(mat9)
{   
& \\
N5 \\
};

\matrix[mymat,right=of mat9,row 2/.style={nodes={draw,fill=gray!30}}]
(mat10)
{
& \\
t_1 & t_3  \\
};

\matrix[mymat, inner sep=0, anchor=west,row 2/.style={nodes={draw,fill=gray!30,inner sep=0.333em}}]
at (1.65,-5.5) 
(mat11)
{   
& & & \\
t_1 & t_2 & t_3 & t_4\\
};


\node[ draw, red, dashed, line width=0.7pt, fit=(mat2)(mat4)(mat6)(mat8) 
(mat10)](node1){};
\begin{scope}[shorten <= -2pt]
\draw[*->]
(mat1-2-1.east) -- (mat2-2-1.west);
\draw[*->]
(mat3-2-1.east) -- (mat4-2-1.west);
\draw[*->]
(mat5-2-1.east) -- (mat6-2-1.west);
\draw[*->]
(mat7-2-1.east) -- (mat8-2-1.west);
\draw[*->]
(mat9-2-1.east) -- (mat10-2-1.west);

\end{scope}
\draw[green, dashed](node1.south west) -- (mat11.north west);
\draw[green, dashed](node1.south east) -- (mat11.north east);

\end{tikzpicture} 

\bigskip

\begin{tikzpicture}[>=latex]
\matrix[mymat,anchor=west,row 2/.style={nodes=draw}]
at (0,0) 
(mat1)
{   
     & \\
     N1 \\
};

\matrix[mymat,right=of mat1,row 2/.style={nodes={draw,fill=gray!30}}]
(mat2)
{
     & \\
    t_2 & t_4 \\
};

\matrix[mymat,anchor=west,row 2/.style={nodes=draw}]
at (0,-1) 
(mat3)
{   
& \\
N2 \\
};

\matrix[mymat,right=of mat3,row 2/.style={nodes={draw,fill=gray!30}}]
(mat4)
{
&  &  \\
t_1 & t_3 & t_4  \\
};
\matrix[mymat,anchor=west,row 2/.style={nodes=draw}]
at (0,-2) 
(mat5)
{   
& \\
N3 \\
};

\matrix[mymat,right=of mat5,row 2/.style={nodes={draw,fill=gray!30}}]
(mat6)
{
& &\\
t_1 & t_2 & t_3 \\
};
\matrix[mymat,anchor=west,row 2/.style={nodes=draw}]
at (0,-3) 
(mat7)
{   
    & \\
    N4 \\
};

\matrix[mymat,right=of mat7,row 2/.style={nodes={draw,fill=gray!30}}]
(mat8)
{
& \\
t_3 & t_4  \\
};
\matrix[mymat,anchor=west,row 2/.style={nodes=draw}]
at (0,-4) 
(mat9)
{   
& \\
N5 \\
};

\matrix[mymat,right=of mat9,row 2/.style={nodes={draw,fill=gray!30}}]
(mat10)
{
& \\
t_1 & t_3  \\
};


\matrix[mymat, anchor=west,row 2/.style={nodes={draw,fill=gray!30}}]
at (1.65,-5.5) 
(mat11)
{   
& & & \\
t_1 & t_2 & t_3 & t_4\\
};

\node[ draw, red, dashed, line width=0.7pt, fit=(mat2)(mat4)(mat6)(mat8) 
(mat10)](node1){};
\begin{scope}[shorten <= -2pt]
\draw[*->]
(mat1-2-1.east) -- (mat2-2-1.west);
\draw[*->]
(mat3-2-1.east) -- (mat4-2-1.west);
\draw[*->]
(mat5-2-1.east) -- (mat6-2-1.west);
\draw[*->]
(mat7-2-1.east) -- (mat8-2-1.west);
\draw[*->]
(mat9-2-1.east) -- (mat10-2-1.west);

\end{scope}
\draw[green, dashed](node1.south west) -- (mat11-2-1.north west);
\draw[green, dashed](node1.south east) -- (mat11-2-4.north east);

\end{tikzpicture} 

\end{document}

Addendum

As mentioned in a comment, I don't think it's necessary to use a \matrix for the N nodes on the left. I don't understand the purpose of the empty first row in your matrices either, seems you're adding a lot of unnecessary code. Here is one alternative method for making your diagram. I make use of the chains library to create the N nodes on the left (could also have used a \matrix, actually). Everything is placed relative to other stuff, so no explicit coordinates are used. To draw the arrows, I used a loop.

I used the arrows.meta library instead of arrows, as TikZ considers the latter to be deprecated in favor of the former (see the manual for version 3.0.1a page 512).

\documentclass{article}
\usepackage{tikz}
\usetikzlibrary{matrix,fit,positioning,chains,arrows.meta}
\tikzset{
 nodestyle/.style={ 
    text height=2.5ex,
    text depth=0.75ex,
    text width=3.25ex,
    align=center,
    draw
},
 mymat/.style={
    matrix of math nodes,
    nodes={
      nodestyle,
      fill=gray!30,
    },
    column sep=-\pgflinewidth,
}  
}
\begin{document}

\begin{tikzpicture}[node distance=1.2cm]
\begin{scope}[
  start chain=N going below,
  every node/.style={
     nodestyle,
     on chain
  },
  node distance=2mm
]    
  \node {$N1$};
  \node {$N2$};
  \node {$N3$};
  \node {$N4$};
  \node {$N5$};
\end{scope}

\matrix [mymat,right=of N-1, name=m1] {t_2 & t_4 \\ };
\matrix [mymat,right=of N-2, name=m2] {t_1 & t_3 & t_4 \\ };
\matrix [mymat,right=of N-3, name=m3] {t_1 & t_2 & t_3 \\ };
\matrix [mymat,right=of N-4, name=m4] {t_3 & t_4 \\ };
\matrix [mymat,right=of N-5, name=m5] {t_1 & t_3 \\ };


\node [draw, red, dashed, fit=(m1)(m2)(m3)(m4)(m5), name=M] {};

\foreach \i in {1,...,5}
   \draw [Circle-Latex,shorten <=-2pt] (N-\i) -- (m\i-1-1);


\matrix [mymat,below=5mm of M, name=m11] {t_1 & t_2 & t_3 & t_4 \\ };

\draw[green, dashed](M.south west) -- (m11-1-1.north west);
\draw[green, dashed](M.south east) -- (m11-1-4.north east);

\end{tikzpicture}
\end{document}

I don't understand why you're using \matrix for all the Nn nodes though, seems unnecessary. I'll edit in a bit, showing one possible other method for making this diagram. — Torbjørn T., May 02 '18 at 10:24

Joining corners of a node and matrix

1 Answers1

Addendum