Reuse of coordinate names in Tikz

Question

Many times I reuse help coordinate names in tikz pictures. Normally, they are overwritten. Now I have found an instance where they are maintained.

\documentclass{article}
\usepackage{tikz}
\begin{document}
Reuse of coordinate names overwrites them:
\begin{tikzpicture}
  \foreach \x in {1,...,5}
    \node (A) at (\x,0){.};
  \draw[red] (A) circle(2pt);
\end{tikzpicture}
When placed in a macro also:
\def\overwritecoords{%
  \foreach \x in {1,...,5}
    \node (A) at (\x,0){.};}
\begin{tikzpicture}
  \overwritecoords
  \draw[green] (A) circle(2pt);
\end{tikzpicture}
But when the \verb|\foreach| uses a macro, not:
\def\placecoord{\node (A) at (\x,0){.};}
\def\keepcoords{%
  \foreach \x in {1,...,5}
    \placecoord}
\begin{tikzpicture}
  \keepcoords
  \draw[blue] (A) circle(2pt);
\end{tikzpicture}
Although, they can only be used once:
\begin{tikzpicture}
  \keepcoords
  \draw[blue] (A) circle(2pt);
  \draw[->] (A)--++(0,-1); 
\end{tikzpicture}
\begin{tikzpicture}
  \keepcoords
  \draw[->] (A)--++(0,-1); 
  \draw[blue] (A) circle(2pt);
\end{tikzpicture}
If the same name is reused again, the previous ones are lost, as usual:
\begin{tikzpicture}
  \keepcoords
  \node (A) at (6,0){.};
  \draw[orange] (A) circle(2pt);
\end{tikzpicture}
\end{document}

I could not find a description of this behaviour but, since the manual is so extensive, I might have overlooked it. I do not understand why and how should tikz maintain a list of coordinates with the same name, when normally they are overwritten. I am a bit worried that at some point, in such a set-up, tikz might pick the wrong help coordinate, or more than one, for drawing my picture. Could that happen? Is there a way to avoid the accumulation of coordinates with the same name?

On the other hand, if I use this behaviour for some purpose, would that be stable in new updates of the package?

Answer 1

The PGF manual explains how \foreach figures out the body of its loop in Section 88 "Repeating Things: The Foreach Statement":

Syntax for the commands. Let us move on to a more complicated setting. The first complication occurs when the are not some text in curly braces. If the \foreach statement does not encounter an opening brace, it will instead scan everything up to the next semicolon and use this as .

(bold-emphasis mine)

A key part of this is that the scan does not expand as it goes. So the semi-colon inside \placecoord is not seen and the \foreach takes everything up to the semi-colon at the end of the \draw command. Thus the \draw is dragged into the \foreach loop.

To block this behaviour (assuming that this is what you want to do) then you need to ensure that TikZ/PGF encounters the semi-colon at the right place without expansion.

One option is to put it in after \keepcoords:

\begin{tikzpicture}
  \keepcoords;
  \draw[blue] (A) circle(2pt);
\end{tikzpicture}

Another is to put it inside \keepcoords:

\def\keepcoords{%
  \foreach \x in {1,...,5}
    \placecoord;}

Yet another is to use braces to delimit the body of the loop.

\def\keepcoords{%
  \foreach \x in {1,...,5}
  {\placecoord}}

As for your question about relying on this behaviour, I would say "It's documented, so it's likely to stay.". But from how you write about it then I'm not sure it's behaving how you think it is behaving - in particular, TikZ/PGF is not maintaining a list of nodes defined with the same label, each overwrites the previous one, but rather the \draw is being used as each one is defined. So if this is what you want, then you can rely on it (as much as anything).