Name tikz pic as if it was a node

Question

I'm confused about tikz pics. I can freely define them without issue, and even add coordinates inside that will be available by name inheritance like (-A) as for example in \pic (S) at (0,0) {myPic}; \draw (S-A) -- ++(0,0);. This is all documented around p. 255 of the pgf manual.

However I don' understand why it is not possible to give a name to a pic like the one above if no coordinate is defined inside it. By default it would be great if it could behave like a normal node.

MWE:

\documentclass{minimal}
\usepackage{tikz}

\tikzset{
    myBlock/.pic = {
        \node[draw,minimum width=1cm,minimum height=0.5cm] at (0,0) {};
    },
    myBlockNamed/.pic = {
        \node[draw,minimum height=1cm,minimum width=0.5cm] (-m) at (0,0) {};
    },
}

\begin{document}
\begin{tikzpicture}
    \pic () at (0,0) {myBock}; % can't give it a name
    \pic (S) at (3,0) {myBlockNamed}; % can give it a name
    \draw (S-A) -- ++(2,0);
\end{tikzpicture}
\end{document}

The underlying point is I'm working on a control blocs library (so that we can stop the chat below with @Alenanno). For example here is the definition of a math operator

\tikzset{
    mySum background/.style = {
        circle, minimum size=1.5em, fill=white,
    },
    mySum edge/.style = {
        draw=black, circle, minimum size=1.5em, line width=.5pt,
    },
    pics/bOp/.style = {
    code = {
      \node [mySum background] {};
      \foreach \t [count=\i] in {#1}{
        \pgfmathsetmacro{\angle}{\i*90}
        \node[anchor=center, font=\tiny] at (\angle:0.45em) {$\t$};
      }
      \node [mySum edge] {};
    },
  },
}

Answer 1

There is a trick that can be used in some circumstances which exploits the fact that a pic is put inside a scope which can be named with the local bounding box key. Note that this will not work in trees or graphs.

\documentclass[tikz,border=5]{standalone}
\tikzset{pics/.cd,
  pic a/.style={code={
      \node [fill=red!20, shape=circle] {A};
  }}
}

\begin{document}
\begin{tikzpicture}
\pic [local bounding box=A1] at (0,0) {pic a};
\pic [local bounding box=A2] at (3,0) {pic a};
\foreach \i in {0,15,...,345}{
  \draw [red, ->] (\i:1) -- (A1);
  \draw [blue, ->] (3,0)++(\i:1) -- (A2);
}
\end{tikzpicture}
\end{document}

It is also possible to use the name given to the pic by defining a "wrapper" key which explicitly adds a scope to the code:

\documentclass[tikz,border=5]{standalone}
\usetikzlibrary{fit}%
\makeatletter
\tikzset{pics/named scope code/.style={code={\tikz@fig@mustbenamed%
  \begin{scope}[local bounding box/.expanded=\tikz@fig@name]#1\end{scope}%
}}}
\makeatother

\tikzset{pics/.cd,
  pic a/.style={named scope code={
      \node [fill=red!20, shape=circle] {A};
  }}
}

\begin{document}
\begin{tikzpicture}
\pic (A1) at (0,0) {pic a};
\pic (A2) at (3,0) {pic a};
\foreach \i in {0,15,...,345}{
  \draw [red, ->] (\i:1) -- (A1);
  \draw [blue, ->] (3,0)++(\i:1) -- (A2);
}
\end{tikzpicture}
\end{document}

The result is the same as before.

Answer 2

From the Pgf Manual (version 3.0.0), page 252:

1. Unlike nodes, pics cannot be referenced later on. You can reference nodes that are inside a pic, but not “the pic itself.” In particular, you cannot draw lines between pics the way you can draw them between nodes. In general, whenever it makes sense that some drawing could conceivably be connected to other node-like-things, then a node is better than a pic.

2. If pics are used to emulate the full power of a node (which is possible, in principle), they will be slower to construct and take up more memory than a node achieving the same effect.

So, if you're doing something a node does well, and there is no particular reason for using pic, you're better off using a node. There is a workaround in the answer to the question "How to give a name to \pic", but it's exactly that, a workaround.

However, if you want to be able to position nodes around another node, you can check this question I asked some time ago (How to position nodes around another node in TiKZ?) or you use a Tikz foreach with a regular node.

Updated answer

Now that I know what you need, here's a solution with Tikz. The newcommand \mySumn (you can rename it) takes two arguments like \mySumn{#1}{#2}, where #1 is the node's name, and #2 is the position.

I tried to add an optional argument using above of but didn't succeed. However you can do the same with this code but in a different way. Consider the following:

\mySumn{p1}{0,0};
\mySumn{p2}{$(p1.north east)+(1em,1em)$};

It will place a node p1 at the coordinates (0,0) and then it will place another node p2 at p1's north east +1em on the x and the y axis (diagonally above on the right).

Of course you can still draw edges from one node to the other.

Output

Code

\documentclass[margin=10pt]{standalone}
\usepackage{tikz}
\usepackage{xparse}

\usetikzlibrary{calc,arrows.meta,positioning}

\tikzset{
    mySumbk/.style = {
        draw, circle, minimum size=1.5em, fill=white,
    },
}

\newcommand{\mySumn}[2]{
    \node[mySumbk,fill=white] (#1) at (#2) {};
        \foreach \angle/\maop in {
        90/+,
        180/-
        }{
   \node[anchor=center, font=\tiny] at ($(#1)+(\angle:.45em)$) {$\maop$};
}}

\begin{document}
\begin{tikzpicture}[>=latex]

\mySumn{p1}{0,0};
\mySumn{p2}{$(p1.north east)+(1em,1em)$};

\draw (p1) -- (3,0);
\draw (p1) edge[out=90,in=180] (p2);
\end{tikzpicture}
\end{document}

Answer 3

Old question, but I was not happy with the local bounding box solution since I couldn't find a way to have a better control over it (distance to the inner content, for instance).

My solution was to use a nested picture in the pic, together with the remember picture option. The names are passed as arguments in a cascading fashion. With this, you can not only reference the pic as a node (because in fact it is a node) but also reference its contents.

Another point is that with this solution the content of a pic is always centralized, which was also that I was looking for when working on this code.

Be aware:

• you have to compile the document twice because of remember picture.
• nested nodes can be tricky because of inheritance.

Here is the code:

\documentclass[tikz]{standalone}

\usetikzlibrary{positioning}

\begin{document}
\begin{tikzpicture}[remember picture]
  \tikzset{
    mine/.pic={
      \node[draw](#1){
        \begin{tikzpicture}[remember picture]
          \node[draw](#1-1){#1-1};          
          \node[draw, right=of #1-1](#1-2){#1-2};
        \end{tikzpicture}
      };
    },
  }

  \pic{mine=a};
  \pic[right=of a]{mine=b};

  \draw[thick, blue, <->](a-1)to[out=90, in=90](b-1);
  \draw[thick, blue, <->](a-2)to[out=-90, in=-90](b-2);
  \draw[thick, red](a)to(b);
\end{tikzpicture}
\end{document}

Answer 4

It appears that if I define a command like for my operator:

\tikzset{
    mySum background/.style = {
        circle, minimum size=1.5em, fill=white,
    },
    mySum edge/.style = {
        draw=black, circle, minimum size=1.5em, line width=.5pt,
    },
}
\newcommand{\bOp}[1]{
  \tikz[outer sep=0pt]{
    \node [mySum background] {};
    \foreach \t [count=\i] in {#1}{
      \pgfmathsetmacro{\angle}{\i*90}
      \node[anchor=center, font=\tiny] at (\angle:0.45em) {$\t$};
    }
    \node [mySum edge] {};
  }
}

It is then possible to create a node that contains a fake pic, as what is above is exactly what one would use to define a pic.

\node (1) at (0,0) {\bOp{+,-,,}};
\draw (4,0) -- (1);