Safe character to use as a separator in string

Question

For some reasons, I need to choose one or two character to delimit placeholders in LaTeX3 strings, like:

I like ``NAME`OF`FRUITS``

in order to allow a safer replace (hence the doubling of the first and last symbol to get rid of ambiguity) + easy reading.

But finding a good character is not that easy:

• Many characters have special meanings for LaTeX and babel that redefine many characters, like ^, _, -, ;, ~, #, @… I’d prefer to avoid using them to avoid espace nightmares. And I don’t know all packages, and I’m sure that some popular packages can redefine other chars.
• I’m thinking that non-ascii characters might be an issue since they might be interpreted differently on different computers. In particular, I don’t know if ° could be a valid choice.

Is there a list of character that are basically as safe as letters to use?

EDIT

Here is a more concrete use case I want to consider. I basically want to be sure that no matter what is around \robExtGetPlaceholder{__VEGETABLE__}, the _ should not be turned into another symbol, even after loading a popular package.

\documentclass{article}
\ExplSyntaxOn
\seq_clear_new:N \l_robExt_placeholders_seq
% Make sure that the placeholder is in the list \l_robExt_placeholders_seq.
% This should automatically be called by other tools
\NewDocumentCommand{\robExtAddPlaceholderToList}{m}{
  \seq_put_left:Nn \l_robExt_placeholders_seq { #1 }
}
\NewDocumentCommand{\robExtPlaceholderFromContent}{mm}{
  \str_gset:cn { l_robExt_placeholder_#1_str } {#2}
  \message{aaaaaaaaaaaaa#1}
  \robExtAddPlaceholderToList{#1}
}
\NewDocumentCommand{\robExtDebugPlaceholder}{sm}{
  \message{Placeholder ~ #2 ~ contains: ~ \use:c{l_robExt_placeholder_#2_str}}
  \IfBooleanTF{#1}{\cs_show:c { l_robExt_placeholder_#2_str }}{}
}
\NewDocumentCommand{\robExtGetPlaceholder}{m}{
  \use:c{l_robExt_placeholder_#1_str}
}
\NewDocumentCommand{\robExtDebugPlaceholdersContents}{s}{
  \message{List ~ of ~ placeholders:}
  \seq_map_inline:Nn \l_robExt_placeholders_seq {\robExtDebugPlaceholder{##1}}
  \IfBooleanTF{#1}{\cs_show:N \l_robExt_placeholders_seq}{}
}
%% I also have other commands, for instance to replace placeholders etc...
\ExplSyntaxOff
\begin{document}
\robExtPlaceholderFromContent{FRUIT}{Orange}
\robExtPlaceholderFromContent{SENTENCE}{I like FRUIT and VEGETABLE}
$\robExtPlaceholderFromContent{VEGETABLE}{Salad}$
\robExtDebugPlaceholdersContents*
Does it mean that whatever think is put around the get placeholder here, it will still be interpreted correctly? (no escape, no weird replacement of the character with another character…)
$1 + \robExtGetPlaceholder{FRUIT} + \robExtGetPlaceholder{VEGETABLE}$
\end{document}

Answer 1

In full genenerality the answer is

There is no safe character.

A slightly more helpful answer would be

Only you can say which are the safe (sequences of) characters. You can use any sequence guaranteed not to appear as data.

The bullet points on your question don't seem that relevant, babel, input encodings, etc (mostly) only matter if you are typesetting, but here you are just delimiting substrings.

You may think of ° as a degree sign character, but to pdftex that is the two character tokens C2 B0 and it's just safe to use as a delimiter if the two bytes C2 B0 will never appear in the delimited string. If you double it then the delimiter is the four tokens C2 B0 C2 B0 and the answer is the same, this is safe to use as a delimiter so long as the four bytes C2 B0 C2 B0 do not appear in the data to be delimited.

The same is true of higher characters to pdftex is not a duck but the four bytes F0 9F A6 86 and you can use ... as the delimiter so long as those four bytes do not appear in ...

You can also use multiple bytes of printable ascii, eg [[....]] delimits ... safely so long as you ensure ... never includes [[ or ]].

The details here are for classic 8bit tex such as latex or pdflatex, but the main point of the answer would be the same with xelatex or lualatex.

Answer 2

Edit

Turns out I was wrong on the assumption that some characters accessible by the Alt key were ASCII. While this workaround "works" it may introduce a plethora of unseen problems. Thanks to David Carlise for the clarification.

Original asnwer

Out of curiosity, I tested how far classic ASCII characters can be pushed to be used as functions or potentially escape characters as what you need, and turns out they run! At least in this example you can see something that may be close to your needs:

\documentclass{article}
\newcount\newcase
\newcase=0
\def♪{%
  \ifnum\newcase=0
    \newcase=1
    [
  \else
    \newcase=0
    ]
  \fi
}
\begin{document}
Musical math: ♪1+1=2♪
\end{document}

At least, this works on LaTeX while running under UTF-8. There is a caveat though: only one “especial character” can be defined at the same time, for instance, doing something like this:

\def♪{A thing}
\def♫{Another thing}

Will cause an “Use of � doesn't match its definition” error. I'm not entirely sure of the reason, but at least this seems to work for one “odd” definition. And in that one “odd” definition you can go wild with characters like ☺☻♪♫←↓→↑§ and more!

Hope that helps!