Define coordinate with anchor point

Question

I have macros that use anchor points in pgfplots to access certain coordinate. For instance, in a graph I can invoke a macro such as:

\newcommand*{\DoSomethingWithXAxis}{%
   \draw [ultra thick, blue] (current axis.left of origin) -- (current axis.right of origin);
}

I would like to be able to reuse such macros in a basic tikzpicture environment by manually defining the coordinates (current axis.left of origin) and (current axis.right of origin). The problem stems from the fact that those are anchor points, so was wondering what is the simplest approach to set those.

Notes:

• I realize that \DoSomethingWithXAxis presented here is not very interesting and that there are easier ways to accomplish the same goal. But, this is just a MWE, with emphasis on the M part of the MWE.

The MWE below has \DoSomethingWithXAxis commented out in the first tikzpicture and yields the following. The desired result is for the tikzpicture axis to also be in blue via \DoSomethingWithXAxis}.

References:

• How to access \xmin, \xmax, \ymin, \ymax from within PGFplots axis environment

Code:

\documentclass{article}
\usepackage{pgfplots}

\newcommand*{\DoSomethingWithXAxis}{%
    \draw [ultra thick, blue] (current axis.left of origin) -- (current axis.right of origin);
}

%% Only specifying one side to ensure that this works with both  
%% manually specified end point and an auto determined one.
\newcommand*{\XMax}{8}

\begin{document}
\begin{tikzpicture}
    \draw [-latex] (1,0) -- (\XMax,0) node [above, pos=0.5] {tikzpicture};

    %% Don't think these coordinates are being set correctly?
    \coordinate (current axis.left of origin)  at (1,0);
    \coordinate (current axis.right of origin) at (\XMax,0);
    %\DoSomethingWithXAxis;%%% <---- How do this ????
\end{tikzpicture}
\par\medskip
\begin{tikzpicture}
    \begin{axis}[axis lines=middle, samples=50, xmax=\XMax]
        \addplot [mark=none, red, thick] {x*x} node [pos=0.75, right] {pgfplots};
        \DoSomethingWithXAxis;
    \end{axis}
\end{tikzpicture}
\end{document}

Answer 1

(Posting answer as requested.)

A small improvement of Peter's method could be to use \pgfplotsifinaxis only to define new coordinates, and do all the work with those coordinates. So one would have something like

\newcommand*{\DoSomethingWithXAxis}{%
    \pgfplotsifinaxis{% https://tex.stackexchange.com/a/17436/4301
        \coordinate (current axis left of origin) at (current axis.left of origin);
        \coordinate (current axis right of origin) at (current axis.right of origin);
    }{}%
    \draw [ultra thick, blue] (current axis left of origin) -- (current axis right of origin);
}

The benefit of this is that you don't have to duplicate all the code in \DoSomethingWithXAxis.

Answer 2

One way to address this problem is to use \pgfplotsifinaxis and have a different code when outside of the axis environment:

\newcommand*{\DoSomethingWithXAxis}{%
    \pgfplotsifinaxis{% https://tex.stackexchange.com/a/17436/4301
        \draw [ultra thick, blue] (current axis.left of origin) -- (current axis.right of origin);
    }{%
        \draw [ultra thick, blue] (current axis left of origin) -- (current axis right of origin);
    }%
}

where I define a current axis left of origin (note the missing .) when I am not in an axis environment.

Notes:

• I really dislike this solution as it requires two mode of each macro that uses the coordinates of the ends of the axis.

References:

• Determine if within a PGF axis environment

Code:

\documentclass{article}
\usepackage{pgfplots}

\newcommand*{\DoSomethingWithXAxis}{%
    \pgfplotsifinaxis{% https://tex.stackexchange.com/a/17436/4301
        \draw [ultra thick, blue] (current axis.left of origin) -- (current axis.right of origin);
    }{%
        \draw [ultra thick, blue] (current axis left of origin) -- (current axis right of origin);
    }%
}

%% Only specifying one side to ensure that this works with both  
%% manually specified end point and an auto determined one.
\newcommand*{\XMax}{8}

\begin{document}
\begin{tikzpicture}
    \draw [-latex] (1,0) -- (\XMax,0) node [above, pos=0.5] {tikzpicture};

    \coordinate (current axis left of origin)  at (1,0);
    \coordinate (current axis right of origin) at (\XMax,0);
    \DoSomethingWithXAxis;%%% <---- Works! BUT I needed to define different coordinate above
\end{tikzpicture}
\par\medskip
\begin{tikzpicture}
    \begin{axis}[axis lines=middle, samples=50, xmax=\XMax]
        \addplot [mark=none, red, thick] {x*x} node [pos=0.75, right] {pgfplots};
        \DoSomethingWithXAxis;
    \end{axis}
\end{tikzpicture}
\end{document}

Answer 3

Here is a simple shape and a command to do what you want. \dualcoordinate is limited, it just looks like TikZ code and the node is always named current axis. Also the center anchor is not in the center.

\documentclass{article}
\usepackage{pgfplots}

\newcommand*{\DoSomethingWithXAxis}{%
    \draw [ultra thick, blue] (current axis.left of origin) -- (current axis.right of origin);
}

\makeatletter
% options for the new shape
\pgfset{
  second point x/.initial=0,
  second point y/.initial=0,
  second point/.style 2 args={/pgf/second point x=#1,/pgf/second point y=#2},
}

% to store the origin of the node
\newdimen\pgf@base@x
\newdimen\pgf@base@y

% new shape
\pgfdeclareshape{dual coordinate}
{
  \savedanchor\firstpoint{%
    % save the origin of the node, needed for \secondpoint
    % otherwise, it would be relative to the first
    \global\pgf@base@x=\pgf@x
    \global\pgf@base@y=\pgf@y
    % copied from \pgfdeclareshape{coordinate}
    % x
    \pgf@x=.5\wd\pgfnodeparttextbox%
    % y
    \pgf@y=.5\ht\pgfnodeparttextbox%
    \advance\pgf@y by -.5\dp\pgfnodeparttextbox%
  }
  \savedanchor\secondpoint{%
    \pgfpointdiff{\pgfpoint{\pgf@base@x}{\pgf@base@y}}%
        {\pgfpointxy{\pgfkeysvalueof{/pgf/second point x}}{\pgfkeysvalueof{/pgf/second point y}}}%
  }

  \anchor{left of origin}{\secondpoint}
  \anchor{right of origin}{\firstpoint}
  \anchor{center}{\firstpoint}% necessary to prevent odd errors
  \anchorborder{\firstpoint}

  \nodeparts{}% no text
}

% split to allow space between first and second coordinate
\def\dualcoordinate(#1,#2){\dual@coordinate{#1}{#2}}
\def\dual@coordinate#1#2(#3,#4){%
    \node[dual coordinate, second point={#3}{#4}] (current axis)  at (#1,#2) {}%
    % no semicolon here, so it's necessary in the tikz picture
}
\makeatother

%% Only specifying one side to ensure that this works with both  
%% manually specified end point and an auto determined one.
\newcommand*{\XMax}{8}

\begin{document}
\begin{tikzpicture}
    \draw [-latex] (1,0) -- (\XMax,0) node [above, pos=0.5] {tikzpicture};

    \dualcoordinate (1,0) (\XMax,0);
    \DoSomethingWithXAxis;
\end{tikzpicture}
\par\medskip
\begin{tikzpicture}
    \begin{axis}[axis lines=middle, samples=50, xmax=\XMax]
        \addplot [mark=none, red, thick] {x*x} node [pos=0.75, right] {pgfplots};
        \DoSomethingWithXAxis;
    \end{axis}
\end{tikzpicture}
\end{document}