yscale=-1 causes incorrect drawing of bipoles

[asinghvi17]

MWE:

\documentclass[tikz, export=png]{standalone}

\usepackage{circuitikzgit}

\begin{document}

\begin{tikzpicture}
  \draw (0, 0) to [R] (0, 2);
  \draw (1, 2) to [R] (1, 0);
\end{tikzpicture}

\begin{tikzpicture}[yscale = -1]
  \draw (0, 0) to [R] (0, 2);
  \draw (1, 2) to [R] (1, 0);
\end{tikzpicture}

\end{document}

The first tikzpicture has no yscale=-1, and the resistors are both drawn normally. However, in the second, there is a vertical line through both resistors; I suspect this is because the resistor are being drawn without the yscale=-1 option, even when it is set globally.

Is there any change to circuitikz that could fix this, or should I just work with no y-flip, on the negative y-axis?

[Rmano]

Hmmm... the problem is that circuititk is putting the "left" anchor to the first node, and the "right" anchor to the second one; when yscale=-1 is active, the coordinates are swapped, but the nodes are not rotated, so the point above is connected to the point below and the other way around. You can check this because adding tranform shape to the options all is ok, but now the labels are mirrored, too.

This behavior was the same in at least 0.8.3, so it's not a regression, at least.

It's a bug, but I am not sure on how to tackle it.

[asinghvi17]

Thanks, this isn't high priority at all so don't worry if it takes some time. I converted my breadboard layout to use negative y which is actually more intuitive for displacements.

[Rmano]

This is really complex, and it goes all way back at least to 0.6. I am trying to understand exactly what's happening, but touching the main code in pgfcircpath is more difficult than expected, and it has a bad effect on labels, too. If you like to dive in, I have a test branch on my clone: https://github.com/Rmano/circuitikz/tree/bug-211

To see real fireworks, try to play with:

\documentclass[]{article}

\usepackage{circuitikz-0.4}

\begin{document}

\begin{tikzpicture}
  \draw (0, 0) to [D, l=D1] (0, 2);
  \draw (0, 0) to [D, l=D2] (2, 0);
  \draw (0, 0) to [D, l=D3] (3, 3);
  \path (0,0) node[draw]{A};
  \path (0,2) node[draw]{B};
  \path (2,0) node[draw]{C};
\end{tikzpicture}

\bigskip

\begin{tikzpicture}[
        % rotate=45,
        xscale = 2,
        % scale = 2,
        % transform shape, every node/.style={transform shape}
    ]
  \draw (0, 0) to [D, l=D1] (0, 2);
  \draw (0, 0) to [D, l=D2] (2, 0);
  \draw (0, 0) to [D, l=D3] (3, 3);
  \path (0,0) node[draw]{A};
  \path (0,2) node[draw]{B};
  \path (2,0) node[draw]{C};
\end{tikzpicture}

\end{document}

So for now I conclude that circuitikz is not working with global negative scale and with rotation parameters...

I will leave this here, opened, in the hope that I can find the time to rewrite all the code

[Rmano]

Also, xscale=1 and yscale=2 is funny (was there in 0.4, checked).

The problem is the \pgf@circ@direction calculation, and all the anchor shuffling; changes will affect almost everything...

I will merge a documentation update for now, see PR #217

[asinghvi17]

Also, @Rmano, you can probably make branches on the main repo as a collaborator, right? Helps with discovery, and you don't need to resync your fork's master every time you want to make a PR :)

[Rmano]

I prefer (probably because my git skills are quite basic) to branch on my fork to test things... and let the main repository clean. I fear that it could become like some other projects I'm in and that have grown to a forest, really...

[Rmano]

A lot of time ago there was a develop branch --- to split maintenance and new features. Maybe I can try to do it again, with topic branches... will study a bit more git and I'll see.

[Rmano]

Asked a question on tex.stackexchange.com https://tex.stackexchange.com/questions/501929/tikz-angle-calculation-and-rotation-with-transformed-coordinates

[Rmano]

Well, with the help of @marmot of tex.stackexchange.com, I prepared a patch that seems to fix this issue. It's a quite scary thing, changing a basic part of the package, but... the problem is that it now breaks everything if transform shape is active. Will try to experiment a bit more.

[Rmano]

Ok, I hacked something out in #257. I'd really appreciate testing on this, because it's scary. It should not change anything when not needed, but you know...

With this change, the original snippet renders as this:

\documentclass[]{article}
\usepackage{circuitikz}
\begin{document}
\begin{tikzpicture}
  \draw (0, 0) to [R=R] (0, 2);
  \draw (1, 2) to [R=R] (1, 0);
\end{tikzpicture}

\begin{tikzpicture}[yscale = -1]
  \draw (0, 0) to [R=R] (0, 2);
  \draw (1, 2) to [R=R] (1, 0);
\end{tikzpicture}
\end{document}

renders like:

which seems correct. Moreover, the longish example:

\documentclass[]{article}

\usepackage{circuitikz}

\begin{document}

\begin{tikzpicture}
  \draw (0, 0) to [D, l=D1, a=A1] (0, 3);
  \draw (0, 0) to [D, l=D2, a=A2] (3, 0);
  \draw (0, 0) to [D, l=D3, a=A3] (3, 3);
  \path (0,0) node[draw]{A};
  \path (0,3) node[draw]{B};
  \path (3,0) node[draw]{C};
\end{tikzpicture}

\bigskip
\typeout{Now the scaling one:}
\begin{tikzpicture}[
        % rotate=30,
        xscale = 2,
        % yscale = -1,
        % scale = 2,
        % transform shape,
        % every node/.style={transform shape}
    ]
  \draw (0, 0) to [D, l=D1, a=A1, ] (0, 3);
  \draw (0, 0) to [D, l=D2, a=A2, ] (3, 0);
  \draw (0, 0) to [D, l=D3, a=A3, ] (3, 3);
  \path (0,0) node[draw]{A};
  \path (0,3) node[draw]{B};
  \path (3,0) node[draw]{C};
\end{tikzpicture}
\typeout{Now the rotating one:}
\begin{tikzpicture}[
        rotate=30,
        % xscale = 2,
        % scale = 2,
        % transform shape, every node/.style={transform shape}
    ]
  \draw (0, 0) to [D, l=D1, a=A1, ] (0, 3);
  \draw (0, 0) to [D, l=D2, a=A2, ] (3, 0);
  \draw (0, 0) to [D, l=D3, a=A3, ] (3, 3);
  \path (0,0) node[draw]{A};
  \path (0,3) node[draw]{B};
  \path (3,0) node[draw]{C};
\end{tikzpicture}


\begin{circuitikz}[american, rotate=-20, 
        transform shape,
        ]
    \draw (0,0) -- ++(1,0) to[R=R] ++(2,0)
    to [rmeterwa, t=A, i=$i$] ++(0,-2) node[ground]{};
    \draw (1,0) to[rmeterwa, t=V, v=$v$] ++(0,-2) node[ground]{};
\end{circuitikz}

\end{document}

produces