How can I parse a # from a macro argument?

Question

I want to write a macro such that \ghpr{myrepo#1234} expands to \href{https://github.com/my-username/myrepo/pull/1234}{myrepo#1234}. (I know how to do it for \ghpr{myrepo}{1234}, but I'd prefer to keep the # separator if possible.)

My attempt at doing this is:

\documentclass{standalone}

\usepackage{xstring}
\usepackage{hyperref}
\newcommand*\ghpr[1]{%
  \StrSubstitute{#1}{##}{/pull/}[\tmpurl]%
  \href{https://github.com/my-username/\tmpurl}{#1}%
}

\begin{document}
\ghpr{myrepo#1234}
\end{document}

but this produces Illegal parameter number in definition of \xs_arg_ii when I try to use it.

I assume the problem here is that I'm not escaping the # correctly; how should I be writing this command?

Answer 1

I can understand your wish for # as the argument delimiter, but note that this is a hard to parse for character in TeX due to its nature as the parameter token.

Anyway, the following proposes 3 solutions to make this work.

• \ghprDetok would break any macros or special characters (read: non-ASCII in pdfTeX) in your argument, but is the fastest of them
• \ghprEtl is the second fastest (but stable)
• \ghprRegex is the slowest (but stable)

In none of the three do I check whether an # was actually part of the argument (which would result in a low level TeX error in the \ghprDetok and \ghprEtl and lead to an empty numeric part in \ghprRegex). If you need detection here I can add it for all three variants.

The \ghprEtl variant reuses the internal of the \ghprDetok, which wouldn't be necessary but it'd need an auxiliary so why not reuse the other one.

\documentclass{article}

\usepackage{hyperref}
\usepackage{etl}

\newcommand*\ghprfastest[2]
  {\href{https://github.com/my-username/#1/pull/#2}{#1\##2}}%

\newcommand*\ghprDetok[1]
  {\expandafter\ghprDetokAUX\detokenize{#1}\stop}
\expanded{\unexpanded{\def\ghprDetokAUX#1}\string#\string#}#2\stop
  {\ghprfastest{#1}{#2}}

\ExplSyntaxOn
\cs_new_protected_nopar:Npn \ghprEtl #1
  {%
    \exp_last_unbraced:Ne \ghprDetokAUX
      { \etl_token_replace_once:nNe {#1} ## { \c_hash_str\c_hash_str } }
    \stop
  }
\cs_generate_variant:Nn \etl_token_replace_once:nNn { nNe }

\cs_new_protected_nopar:Npn \ghprRegex #1
  {
    \regex_split:nnN { \# } {#1} \l_tmpa_seq
    \exp_args:Nee \ghprfastest
      { \seq_item:Nn \l_tmpa_seq 1 }
      { \seq_item:Nn \l_tmpa_seq 2 }
  }
\ExplSyntaxOff

\begin{document}
\ghprfastest{myrepo}{1234}

\ghprDetok{myrepo#1234}

\ghprEtl{myrepo#1234}

\ghprRegex{myrepo#1234}
\end{document}

Answer 2

You can change catcode of # locally:

\def\ghpr{\bgroup \catcode`\#=12 \ghprA}
\def\ghprA #1{\egroup \href{https://github.com/my-username/myrepo/pull/1234}{#1}}

\ghpr{myrepo#1234} % expands to \href{https://github.com/my-username/myrepo/pull/1234}{myrepo#1234}