Creating new counter formats on-the-fly with enumitem's \AddEnumerateCounter*

Question

Aesthetic caveat: I am actually using small-caps with hanging figures and not lining. However, in creating a minimal example, I have succumbed to the defaults and created a typographical monstrosity.

---

I am trying to create an environment with something like the following syntax:

\begin{<environment name>}{<single letter>}[<options>}
  \item <stuff>
\end{<environment name>}

I want the list to be enumerated and I want each item to be labelled with \textsc{<single letter>}\arabic{<number of item>}.

This is straightforward with enumitem. For example, I can do this:

\NewDocumentEnvironment{letters}{ O {} m }{%
  \begin{enumerate}[label=\textsc{#2}\arabic*., ref=\textsc{#2}\arabic*, #1]
}{%
  \end{enumerate}%
}

and then

\begin{letters}{a}
  \item first
  \item second
\end{letters}

will produce

However, when I reference an item, I would like the format of the reference to be \textsc{<letter>}\arabic{<value>} (without a dot). Again, this is fine and the ref in enumitem above takes care of this nicely.

But I would also like to be able to capitalise the letter when the reference occurs at the start of a sentence, for example. Both fancyref and cleveref offer this kind of facility, but the use of small-caps for lower case makes this problematic.

So I started to investigate enumitem and started messing around with the \AddEnumerateCounter(*). As far as I can tell, this command does nothing of the sort. Instead, it seems to add a counter format so it is for adding an analogue to \arabic rather than analogue to c@page, for example. That said, I'm not at all sure I understand it in any way.

I did manage to reproduce my existing result with the following code

\makeatletter
\def\csimplelettered#1{\expandafter\@csimplelettered\csname c@#1\endcsname}
\def\@csimplelettered#1{\@arabic#1}
\makeatother
\AddEnumerateCounter{\csimplelettered}{\@csimplelettered}{2}
\NewDocumentEnvironment{simplelettered}{ O {} m }{%
  \begin{enumerate}[label=\textsc{#2}\csimplelettered*., ref=\textsc{#2}\csimplelettered*, #1]
}{%
  \end{enumerate}%
}

which lets me write

\begin{simplelettered}{b}
  \item first
  \item second
\end{simplelettered}

to produce

So then I tried to create a new counter format on-the-fly within the environment definition. Here's one (surely laughable) attempt

\makeatletter
\NewDocumentEnvironment{lettered}{ O {} m }{%
  \expandafter\def\csname clettered#2\endcsname ##1{\expandafter\expandafter\expandafter\csname @clettered#2\endcsname\csname c@##1\endcsname}%
  \expandafter\def\csname @clettered#2\endcsname ##1{\@arabic##1}%
  \AddEnumerateCounter*{\expandafter\csname clettered#2\endcsname}{\expandafter\csname @clettered#2\endcsname}{2}%
  \begin{enumerate}[label=\textsc{#2}\arabic*., ref=\textsc{#2}\arabic*, #1]
}{%
  \end{enumerate}%
}
\makeatother

Clearly my default strategy of throwing \expandafters randomly in arbitrary places is not proving especially fruitful in this particular case.

And now I'm more interested in this problem than I am in the original one (which I can solve in another way). How can I make the lettered environment work so that it creates a new counter format and uses it in the enumeration?

Complete NMWE:

\documentclass{article}
\usepackage{xparse,enumitem}
\NewDocumentEnvironment{letters}{ O {} m }{%
  \begin{enumerate}[label=\textsc{#2}\arabic*., ref=\textsc{#2}\arabic*, #1]
}{%
  \end{enumerate}%
}
\makeatletter
\def\csimplelettered#1{\expandafter\@csimplelettered\csname c@#1\endcsname}
\def\@csimplelettered#1{\@arabic#1}
\makeatother
\AddEnumerateCounter{\csimplelettered}{\@csimplelettered}{2}
\NewDocumentEnvironment{simplelettered}{ O {} m }{%
  \begin{enumerate}[label=\textsc{#2}\csimplelettered*., ref=\textsc{#2}\csimplelettered*, #1]
}{%
  \end{enumerate}%
}
\makeatletter
\NewDocumentEnvironment{lettered}{ O {} m }{%
  \expandafter\def\csname clettered#2\endcsname ##1{\expandafter\expandafter\expandafter\csname @clettered#2\endcsname\csname c@##1\endcsname}%
  \expandafter\def\csname @clettered#2\endcsname ##1{\@arabic##1}%
  \AddEnumerateCounter*{\expandafter\csname clettered#2\endcsname}{\expandafter\csname @clettered#2\endcsname}{2}%
  \begin{enumerate}[label=\textsc{#2}\arabic*., ref=\textsc{#2}\arabic*, #1]      
  % want to use something like:
  %\begin{enumerate}[label=\textsc{\csname clettered#2\endcsname*}., ref=\csname clettered#2\endcsname*, #1]
}{%
  \end{enumerate}%
}
\makeatother
\begin{document}
\begin{letters}{a}
  \item first
  \item second
\end{letters}
\begin{simplelettered}{b}
  \item first
  \item second
\end{simplelettered}
\begin{lettered}{c}
  \item first
  \item second
\end{lettered}
\end{document}

Answer 1

I don't understand the question, so don't expect this to answer it. But in any case, if you substitute this

\expandafter\def\csname clettered#2\endcsname ##1{\expandafter\expandafter\expandafter\csname @clettered#2\endcsname\csname c@##1\endcsname}%
\expandafter\def\csname @clettered#2\endcsname ##1{\@arabic##1}%
\AddEnumerateCounter*{\expandafter\csname clettered#2\endcsname}{\expandafter\csname @clettered#2\endcsname}{2}%

with

\def\clettered##1{\expandafter\@clettered\csname c@##1\endcsname}%
\def\@clettered{\@arabic}%
\AddEnumerateCounter*{\clettered}{\@clettered}{2}%

it does compile. Apart from the fact that you don't need to define any variants with \csname..\endcsname since you are inside an environment (if this has nothing to do with the question, ignore it).

---

In any case, this is your solution with correct use of \expandafters

\expandafter\def\csname clettered#2\endcsname##1{\csname @clettered#2\expandafter\endcsname\csname c@##1\endcsname}%
\expandafter\def\csname @clettered#2\endcsname##1{\@arabic{##1}}%
\expandafter\AddEnumerateCounter\expandafter*\csname clettered#2\expandafter\endcsname\csname @clettered#2\endcsname{2}%

To explain this, you only need one \expandafter before the \endcsname to create a token from the next \csname so

\csname a\expandafter\endcsname\csname b\endcsname

would first create \b and then \a, leaving \a\b.

Answer 2

I'm not sure I've understood the question, but I think the best approach is to save the label unformatted:

\documentclass{article}

\usepackage{enumitem}

\makeatletter
\newcommand\mylabel[2]{\mylabelfont{#1}\@arabic{#2}}
\makeatother
\let\mylabelfont\textsc % Default
\newcommand\ucref[1]{%
  {\let\mylabelfont\MakeUppercase\ref{#1}}}

\begin{document}

\begin{enumerate}[label=\textsc{a}\arabic*,ref=\protect\mylabel{a}{\arabic*}]
\item \label{1st} First
\item \label{2nd} Second
\end{enumerate}

First is \ref{1st} and second is \ref{2nd}. But compare with \ucref{1st} and \ucref{2nd}.

\end{document}

Here \mylabel just saves the letter and the item number. The format is decided when the label is actually used. This is a general way to delay how the label is formatted when it depends on the context.