I encountered the following problems when using foreach:

Question

\documentclass{standalone}
\usepackage{tikz}
\begin{document}
  \begin{tikzpicture}
    \foreach \x in{0,0.1,...,0.5}{
       \node[] at (20*\x,0) {\x};
       }
  \end{tikzpicture}
\end{document}

However, as follows:

\documentclass{standalone}
\usepackage{tikz}
\begin{document}
  \begin{tikzpicture}
    \foreach \x in{0,0.125,...,0.5}{
       \node[] at (20*\x,0) {\x};
       }
  \end{tikzpicture}
\end{document}

it is normally.

Hello and welcome to Tex.SE. This is due to an rounding error. See here for a full explanation. — Roland, Aug 12 '21 at 01:53
Just use integers in in{1, ..., 10} whenever possible. You can always divide later — Symbol 1, Aug 12 '21 at 03:51
In binary representation numbers of the form n * 2^-m are exact. Whereas n * 10^-m have inherent rounding errors in binary representation. I don't know pfg internal representation, but since 0.125, 0.25 do not give rounding errors I suppose internal representation is binary. — Johannes Linkels, Aug 17 '21 at 23:48

score 9 · Answer 1 · answered Aug 12 '21 at 11:27

Welcome to TeX.SE. As said in comments, this is a precision issue due to how TikZ performs floating point computations (it uses TeX \dimen registers by default). Since the computed numbers are very close to the desired ones, you just need to use \pgfmathprintnumber in order to typeset the numbers rounded to the desired precision:

\documentclass[tikz,border=1mm]{standalone}
\begin{document}
\begin{tikzpicture}
  \foreach \x in {0, 0.1, ..., 0.5} {
    \node at (20*\x, 0) {%
      \pgfmathprintnumber[fixed, precision=1]{\x}%
  };
}
\end{tikzpicture}
\end{document}

Bottom line: computation is one thing, often not perfectly accurate. Number formatting is another thing that \pgfmathprintnumber does well.

Thanks for your answer！ I already understand – 芒果不盲 Aug 12 '21 at 12:15 — 芒果不盲, Aug 12 '21 at 12:15

score 8 · Answer 2 · answered Aug 12 '21 at 15:27

As well as the precision of the displayed number (as answered in the other solutions), there is the issue with the missing node. In the first foreach loop the last node is not displayed because PGF thinks that the next value is 0.50003 which is more than 0.5 so is outside the bounds of the loop.

As Symbol1 said in the comments, it is best to use integers in the loop when using the ... syntax to ensure that this sort of thing doesn't happen. This then means that the coordinate calculation needs changing, but this is easy to incorporate. Slightly more complicated is computing the displayed number. This can be done either in the node text itself using pgfmathparse{\x/10}\pgfmathprintnumber[fixed, precision=1]{\pgfmathresult} or it can be done in the foreach loop itself (which is more useful if you are going to use the number more than once).

\documentclass{standalone}
%\url{https://tex.stackexchange.com/q/610368/86}
\usepackage{tikz}
\begin{document}
  \begin{tikzpicture}
    \foreach \x in{0,0.1,...,0.5}{
       \node[] at (20*\x,0) {\x};
       }
  \end{tikzpicture}
\begin{tikzpicture}
    \foreach[evaluate=\x as \displayx using \x/10] \x in{0,1,...,5}{
       \node[] at (2*\x,0) {\pgfmathprintnumber[fixed, precision=1]{\displayx}};
       }
  \end{tikzpicture}
\end{document}

Good point, you have my upvote! ;-) – frougon Aug 12 '21 at 16:00 — frougon, Aug 12 '21 at 16:00
@frougon And you have mine! – Andrew Stacey Aug 12 '21 at 16:07 — Andrew Stacey, Aug 12 '21 at 16:07

Steven B. Segletes · Answer 3 · 2021-08-12T12:48:50.480

7

Borrowing and adapting the technique from frougon, Precision for semilogyaxis

\documentclass{standalone}
\usepackage{pgfplots, tikz}
\begin{document}
\begin{tikzpicture}
    \foreach \x in{0,0.1,...,0.5}{
       \node[] at (20*\x,0) {\pgfmathprintnumber{\x}};
}
\end{tikzpicture}
\end{document}

edited Aug 12 '21 at 12:48

answered Aug 12 '21 at 09:16

Steven B. Segletes

237,551

1

Thanks for the reference. Note that we only need to round the input numbers here (which are computed by the \foreach statement with imperfect precision before your \pgfset{fixed point arithmetic,...} kicks in); so, the \pgfmathparse{\x} step is useless and might even result in a loss of precision in some situations (depending on what was in \x before). On the other hand, \pgfmathprintnumber parses the input number with \pgfmathfloatparsenumber, which “allows arbitrary precision” (as written in the documentation of \pgfmathprintnumber). :-) – frougon Aug 12 '21 at 11:35
@frougon I have edited, hopefully capturing the essence of your comment. Thanks for your advice. – Steven B. Segletes Aug 12 '21 at 12:18
That's better, but the \pgfset{fixed point arithmetic} is still useless here, AFAICS: no computation whatsoever is performed with this engine! The only thing we need to do here is some rounding before typesetting the number... which is what my answer does. – frougon Aug 12 '21 at 12:22

I encountered the following problems when using foreach:

3 Answers3

Linked