\exp_last_unbraced:No fits this case, but it won't obviously increase the readability.

It's a pity that there's no expl3 function to handle this two-step expansion case.

\exp_args:Nf should do that. It basically equals \expandafter#1\expandafter{\romannumeral-`0#2}

We used to have them (d-type), but there are very few places we found we needed them - as we've required e-TeX for a long time, we've been able to rely on most functions including \unexpanded to protect their return values. Also, now we have \expanded, we can use that ... sorry it doesn't work quite so well for you.

What I have in my mind is sth more general (and/but less efficient, of course), like \exp_by_steps:nn {<num of steps>} {<tokens>}

\documentclass{article}
\usepackage{pgfmath}

\ExplSyntaxOn\makeatletter
% Expand #2 by exactly #1 steps and leaves the result in input stream. 
% The result is returned within \unexpanded.
\cs_new:Npn \exp_by_steps:nn #1#2
  {
    \int_compare:nNnTF {#1} > 0
      {
        \exp_by_steps:eo { \int_eval:n {#1-1} } {#2}
      }
      {
        \exp_not:n {#2}
      }
  }
\cs_generate_variant:Nn \exp_by_steps:nn {eo}

\exp_args:Ne \tl_show:n { \exp_by_steps:nn {0} {\undefined} }
\exp_args:Ne \tl_show:n { \exp_by_steps:nn {1} {\tl_if_eq:ccTF} }
\exp_args:Ne \tl_show:n { \exp_by_steps:nn {2} {\pgfutil@trimspaces{ \x}} }
\makeatother\ExplSyntaxOff

\begin{document}
\end{document}

% > \undefined .
% > \exp_args:Ncc \tl_if_eq:NNTF .
% > \x .

This should be inlined into pgfutil-common.tex directly. Loading files in TeX is slow. @josephwright promised to develop some kind of luadocstrip for l3build to merge files during build. Once that arrives we can split it again.

Made inlined in commit 190342a

Since pgfutil-<format>.def is loaded after pgfutil-common.tex, all wrapped primitives [1] are undefined in pgfutil-common.tex. We have to pay attention to not actually expand any of them in defining pgfexpl macros. Not sure if this is achievable in the long term.

[1] Currently there're \pgfutil@protected and \pgfutil@unexpanded. \pgfutil@expanded should be added (see also #1028).

I think with a little bit of work we could make _ a letter. There are very few uses of _ outside of \write and comments.

: is a bit more difficult, because it's used in \pgfutil@ifnextchar (and some other stuff that needs to skip spaces).

doable. Do you expect it happens in current PR?

I'd say yes, because we should address this immediately as pgfexpl is introduced. However, if you can't do it in a timely manner, we can go ahead without it.

What do you think about

% Naming conventions:
%   pgf_cs = "__pgfexpl_" + expl3_cs.replace(":", "@")
% For example,
%   expl3_cs: \exp_args:Ne,          \tl_put_right:Nn
%   pgf_cs:   \__pgfexpl_exp_args@Ne, \__pgfexpl_tl_put_right@Nn

: is a bit more difficult, because it's used in \pgfutil@ifnextchar (and some other stuff that needs to skip spaces).

Since the ultimate goal is to depend on expl3-code.tex, now I'm kind of leaning more towards making both _ and : letters at once. A constant token list \c__pgfexpl_semicolon_tl and some careful expansion works will overcome the : issue.