In tikz is there a way to set a node style to center the center of each node within a given scope

Question

In the following MWE, I know that the commented line works. What I want to know is why the every node/.style approach is not working.

\documentclass{article}
\usepackage{tikz}

\begin{document}

\begin{tikzpicture}[x=2in,y=2in]

  \node (A) at (1,0)   {};
  \node (B) at (0,1)   {};
  \node (C) at (-1,0)  {};
  \node (D) at (1,1)   {}; 

%      \draw  (A.center) -- (B.center) -- (D.center) -- (C.center) -- cycle;
  \draw[every node/.style={anchor=center}] (A) -- (B) -- (D) -- (C) -- cycle;
\end{tikzpicture}

\end{document}

Is there some way I can do this without having to write (<node_name>.center) for each node I wish to connect with a line?

UPDATE

I can write something like the following:

  \def\concatenatedpoints{}
  \foreach \x in {A,B,C,D}
      {\xdef\concatenatedpoints{\concatenatedpoints (\x.center) -- } }
  \draw \concatenatedpoints     cycle;

which has the desired effect. And, of course, I could write a macro to accomplish this for whenever I want to be able to connect nodes this way.
But it seems that they's got to be a more tikz-flavored approach to efficiently connect the nodes as though they were coordinates.

FURTHER UPDATE

I've tried something very similar to a suggestion made by @percusse in the comments and @jldiaz in his answer. What I had tried was

\draw \foreach \x in {A,B,D,C}{(\x.center) -- } cycle; 

I'd be very curious for an explanation of why my approach fails but either of theirs works.

Answer 1

For the \foreach part, a little clarification.

When \foreach is used within the path, it doesn't simply concatenates strings but parses each spin of the loop and concatenates the paths(!)

So what you have is incomplete path snippets that requires a final coordinate, in other words you tell TikZ

\draw (A.center) --
\draw (A.center) -- (B.center) --
\draw (A.center) -- (B.center) -- (C.center) --
\draw (A.center) -- (B.center) -- (C.center) -- (D.center) --
\draw (A.center) -- (B.center) -- (C.center) -- (D.center) -- cycle;

which is only correct in the last spin but it never reaches to that point due to the error. Instead if you use

\draw (A.center) \foreach \x in {B,D,C}{-- (\x.center)}-- cycle;

Then you have

\draw (A.center)
\draw (A.center) -- (B.center)
\draw (A.center) -- (B.center) -- (C.center)
\draw (A.center) -- (B.center) -- (C.center) -- (D.center)
\draw (A.center) -- (B.center) -- (C.center) -- (D.center) -- cycle;

and each spin is a valid statement. OK, mr. know it all how do you explain this?

\draw \foreach \x in {A,B,D,C}{-- (\x.center)}-- cycle;

No error and no path! Well not really. This is due to the missing internal \pgfpathmoveto{} such that the pen is put on the canvas. For example, a nice riddle is

\pgfpathmoveto{}
\draw  -- (A.center)-- (B.center)--(C.center)--(D.center); 

:) Where does that (a) to (d) line come from?

Answer 2

A silly idea with an ugly syntax, but perhaps you do like it.

\documentclass{article}
\usepackage{tikz}

\def\drawPathThrough#1#2{
  \def\firstnode{#1}
  \draw \foreach \n in {#1,#2} {\ifx\n\firstnode\else -- \fi (\n.center) } --cycle;
}

\begin{document}
\begin{tikzpicture}[x=2in,y=2in]
  \node (A) at (1,0)   {A};
  \node (B) at (0,1)   {B};
  \node (C) at (-1,0)  {C};
  \node (D) at (1,1)   {D};
  \drawPathThrough{A}{B,D,C};
\end{tikzpicture}
\end{document}

Answer 3

This answer provides three solutions:

1. The first solution uses your exact syntax (i.e. node name without a specified anchor or the indicating .). Unfortunately, this needs a fix of an existing macro that is used to parse the implicit version of the node cs.

   This macro checks for an . in the specified coordinate: If one is found it is used to split up node name and referenced anchor. If no .<anchor> is given it automatically adds .center which it only does in case the tikz@shapeborder if is ignored. This part is the cause why nodes connect so intelligently when you do not specify an anchor.

   If we intercept this, we can make your syntax possible.

2. Solution two simply allows to define a generic anchor locally.

   The macro \pgfdeclaregenericanchor{<name>}{<code>} creates a macro \pgf@anchor@generic@<name> that excepts one argument which we can refer to in <code> as ##1. This is used to simply refer to another anchor.

   If you leave out the name of that generic anchor, you can use (A.) to refer to short-cut.

3. Seeing the other solutions to avoid the anchor problem I’d define a few styles that build a path via the insert path style and the .list handler to connect coordinates/nodes:

   • open polygon={<first coordinate>, <list of other coordinates>} and
   • closed polygon with the same syntax that includes an additional -- cycle.

   The use of insert path instead of a rather fixed macro makes it possible to use that path as part of another path (the possibilites aren’t that great for a closed polygon but still).

Solution 1

The patching builds a \tikz@parse@node macro defined as (comments from the original source):

\def\tikz@parse@node#1(#2){%
  \pgfutil@in@.{#2}% Ok, flag this
  \ifpgfutil@in@
    \tikz@calc@anchor#2\tikz@stop%
  \else%
    \pgfkeysgetvalue{/tikz/use anchor}\tikz@temp
    \ifx\tikz@temp\pgfutil@empty
      \tikz@calc@anchor#2.center\tikz@stop% to be on the save side, in
                                  % case iftikz@shapeborder is ignored...
      \expandafter\ifx\csname pgf@sh@ns@#2\endcsname\tikz@coordinate@text%
      \else
        \tikz@shapebordertrue%
        \def\tikz@shapeborder@name{#2}%
      \fi%
    \else
      \expandafter\tikz@calc@anchor#2\tikz@temp\tikz@stop
    \fi
  \fi%
  \edef\tikz@marshal{\noexpand#1{\noexpand\pgfqpoint{\the\pgf@x}{\the\pgf@y}}}%
  \tikz@marshal}

The \pgfkeysgetvalue and the \ifx as well as its \else part is new.

What happens here? First of all, our code is only used if none . is in #2 that means, if an anchor is specified it is used. But if no . is inside #2 it will check whether the value of /tikz/use anchor is empty, if it is, the usual behavior is applied: The .center anchor is used and the boolean tikz@shapeborder is switched on. This Boolean along with the name \tikz@shapeborder@name is later used in many path operators to actually find the start of a line. If you usually say

\draw (<node1>) -- (<some other coordinate>);

The line actually starts at the angular anchor in <some other coordinate>’s direction. Those path operators check \iftikz@shapeborder if a node has been specified without an explicit anchor and an additional move-to is included.

In our case we disable this intelligence (as long as /tikz/use anchor is not empty) and simply add the anchor anyway without setting tikz@shapeborder.

Code 1

\documentclass[tikz]{standalone}
\tikzset{use anchor/.initial=}
\usepackage{etoolbox}
\makeatletter
\patchcmd\tikz@parse@node{\else\tikz@calc@anchor}{%
  \else\pgfkeysgetvalue{/tikz/use anchor}\tikz@temp
  \ifx\tikz@temp\pgfutil@empty\tikz@calc@anchor}{}{}
\patchcmd\tikz@parse@node{\fi\fi}{\fi\else\tikz@calc@anchor#2.\tikz@temp\tikz@stop\fi}{}{}
\makeatother
\begin{document}
\begin{tikzpicture}[x=2in,y=2in, nodes={draw, help lines}]
  \node (A) at (1,0)   {A}; \node (B) at (0,1)   {B};
  \node (C) at (-1,0)  {C}; \node (D) at (1,1)   {D}; 
  \draw[use anchor=center]           (A) -- (B) -- (D) -- (C) -- cycle;
  \draw[use anchor=south west, blue] (A) -- (B) -- (D) -- (C) -- cycle;
\end{tikzpicture}
\end{document}

Solution 2

Not much is to be added to the initial notes of this solution.

This may be the most flexible solution as you actually can (globally) define a few short-cuts like

\tikzset{
  anchor shortcut/.list={
    {nw}{north west},
    {n}{north},
    {ne}{north east}}}

Code 2

\documentclass[tikz]{standalone}
\makeatletter
\tikzset{
  use anchor/.style={anchor shortcut={}{#1}},
  anchor shortcut/.code 2 args=\pgfdeclaregenericanchor{#1}{\pgf@sh@reanchor{##1}{#2}}}
\makeatother
\begin{document}
\begin{tikzpicture}[x=2in, y=2in, nodes={draw, help lines}]
  \node (A) at (1,0)   {A}; \node (B) at (0,1)   {B};
  \node (C) at (-1,0)  {C}; \node (D) at (1,1)   {D}; 
  \draw[use anchor=center]                    (A.)   -- (B.)   -- (D.)   -- (C.)   -- cycle;
  \draw[anchor shortcut={sw}{south west},blue](A.sw) -- (B.sw) -- (D.sw) -- (C.sw) -- cycle;
\end{tikzpicture}
\end{document}

Solution 3

I also can’t say much additional to this solution.

It may be noted that due to the fact how the argument to open polygon (and thus closed polygon) is parsed one cannot write

open polygon={A , …}

but only

open polygon={A, …}

Otherwise, spaces do no harm as all other coordinates are parsed by the \foreach parser.

Please note that the . has to be given in the argument of open polygon anchor. (This can be changed with another text and an \edef without a problem though.)

---

If you only draw tetragons, you could actually just define

\tikzset{
  tetragon/.style args={#1:#2,#3,#4,#5}{
    insert path={(#2.#1) -- (#3.#1) -- (#4.#1) -- (#5.#1) -- cycle}}
}

which you can use as

\draw[red, thick, tetragon={east:A,B,C,D}];

and be done with it.

Code 3

\documentclass[tikz]{standalone}
\makeatletter
\tikzset{
  open polygon anchor/.initial=,
  open polygon/.code args={#1,#2}{%
    \edef\tikz@temp{\pgfkeysvalueof{/tikz/open polygon anchor}}%
    \tikzset{@open polygon/.expanded={\tikz@temp}{#1}{#2}}},
  @open polygon/.style n args={3}{%
    insert path={(#2#1)},
    @@open polygon/.style={insert path={-- (##1#1)}},
    @@open polygon/.list={#3}},
  closed polygon/.style={open polygon={#1},insert path={-- cycle}}}
\makeatother
\begin{document}
\begin{tikzpicture}[x=2in,y=2in, nodes={draw, help lines}]
  \node (A) at (1,0)   {A};  \node (B) at (0,1)   {B};
  \node (C) at (-1,0)  {C};  \node (D) at (1,1)   {D}; 
  \draw[open polygon anchor=.center,     closed polygon={A,B,D,C}]      ;
  \draw[open polygon anchor=.south west, closed polygon={A,B,D,C}, blue];
\end{tikzpicture}
\end{document}

Outputs