Hi all,

this is my first question here... I hope I am in the right place. So, this is a quite esoteric question so the example is a bit convoluted (and verbose, because I think it's easier to understand like this). 

The idea for the code is to have a command like `\appearingbox<2->{name}{content}`. This will typeset the content more or less "plain" in the first overlay and then boxed in the second one. Moreover, I will have the `name` available to link things with drawings.

I will dump the code here, will explain the problem after it. 

```latex
\documentclass[]{beamer}
\usepackage{tikz}
\usetikzlibrary{fit}
\newcommand\insideappinbox[4]{%
    \begin{tikzpicture}[remember picture,baseline=(#2-inner.base)]
        \node[draw=none] (#2-inner) {#3};
        \only#4{
            \node[overlay, draw=#1, fit=(#2-inner)](#2) {};
        }
    \end{tikzpicture}%
}
\newcommand<>{\appearingbox}[3][red]{\relax
    % we make a tigtht inner box here
    % notice that #4 is the overlay spec with <>
    \ifmmode
        \mathchoice{%
            \appearingboxD#4[#1]{#2}{#3}%
        }%
        {%
            \appearingboxT#4[#1]{#2}{#3}%
        }%
        {%
            \appearingboxS#4[#1]{#2}{#3}%
        }%
        {%
            \appearingboxSS#4[#1]{#2}{#3}%
        }%
    \else
        \appearingboxP#4[#1]{#2}{#3}%
    \fi
}

\newcommand<>{\appearingboxD}[3][red]{\relax
        \insideappinbox{#1}{#2}{$\displaystyle #3$}{#4}%
}
\newcommand<>{\appearingboxT}[3][red]{\relax
        \insideappinbox{#1}{#2}{$\textstyle #3$}{#4}%
}
\newcommand<>{\appearingboxS}[3][red]{\relax
        \insideappinbox{#1}{#2}{$\scriptstyle #3$}{#4}%
}
\newcommand<>{\appearingboxSS}[3][red]{\relax
        \insideappinbox{#1}{#2}{$\scriptscriptstyle #3$}{#4}%
}
\newcommand<>{\appearingboxP}[3][red]{\relax
        \insideappinbox{#1}{#2}{$\displaystyle #3$}{#4}%
}
\begin{document}
    \begin{frame}
        \frametitle{F1: Test with specific ones}

        \appearingboxP{textmode}{textmode}

        \[ \appearingboxD<2->{dstyle}{\sum^1 A_D} \]

        $\appearingboxT<2->[blue]{tstyle}{\sum_0^1 A_T}$

        \only<2->{\tikz[remember picture, overlay] \draw[->] (textmode) -- (dstyle) -- (tstyle);}
    \end{frame}

    \begin{frame}
        \frametitle{F2: Test with mathchoice}

        \appearingbox{textmode}{textmode}

        \[ \appearingbox<2->{dstyle}{\sum^1 A_D} \]

        $\appearingbox<2->[blue]{tstyle}{\sum_0^1 A_T}$

        \only<2->{\tikz[remember picture, overlay] \draw[->] (textmode) -- (dstyle) -- (tstyle);}
    \end{frame}
\end{document}

```

This is the output:

![ksnip_20200902-155136.png](/image?hash=594a4f8519e3f0bd318b7db7975cd45f167002c1954dd063095d6eabf66ec106)

Notice that when I use manually the versions of the `\appearingbox` macro for each style (non-math, displaystyle, etc.) like in the first frame (first row in the image) all is ok. The nodes are usable outside  the command and, after a couple of runs, they work beautifully.

On the other hand, if I try to automatically select the appropriate function by way of `\ifmmode` and `mathchoice` I have a very puzzling (at leat for me) result. The appropriate macro is invoked but, alas, the nodes are in the wrong site. 

I tried with `\expandafter` trickery, with older and newer TeXLives, by dancing around the computer... nothing works. I can survive with the manual method, but now I lost two hours trying to understand what happens and I'd like to understand it. Although probably the problem is that I misused `\mathchoice` (although I have a very similar construct in my package `circledsteps`).

Thanks!