Get a transparent, nestable, wrappable background

Question

I like the way soul package enables coloring text background because it:

• adds no padding around \hl{}ed text (unlike \colorbox unless we set \fboxsep=0cm in a surrounding group)
• enables line wrapping (unlike \colorbox which is unbreakable)

However, it does not support:

• transparent background coloring
• nesting highlighted texts (which does makes sense if the background is transparent)

In other words, I am looking for a magical command which would allow me define a custom \mhl highlighting command so that:

\documentclass{standalone}

\newcommand{\mhl}[1]{\whichCommandIsThat{black}{.2}{#1}}

\begin{document}

\begin{minipage}{3.4cm}
here is a \mhl{weird \mhl{nested, wrappable} weird} text
\end{minipage}

\end{document}

would produce:

Any idea?

Answer 1

Edit : see the updated full solution at the end.

I do have a partial answer by using the tcolorbox package, where nesting and transparancy are possible:

\documentclass{article}
\usepackage{tcolorbox}
\tcbuselibrary{skins,breakable,xparse}
\newcommand{\mhl}[1]{%
   \tcbox[tcbox raise base, 
  left=0mm,right=0mm,top=0mm,bottom=0mm,boxsep=0.5pt,arc=0mm,
  boxrule=0pt,opacityfill=0.3,enhanced jigsaw,colback=gray!85!white,
   before=\relax,after=\relax]{#1}
}
\begin{document}
I say \mhl{Hello World} !
This box has a no \mhl{defined \mhl{height}} as you can see. 
The total \mhl{colored box is \mhl{shrunk to the 
\mhl{dimensions} of the \mhl{upper} part}. There} should be  no title. 
\end{document}

resulting in: Allas it is not breakable across lines. May be would it possible to emulate breakable capability by using xstring to handle text word by word...

EDIT: A better solution uses the rather new listofitems package allowing to split the content of \mhl into words or inner \mhl, an to process them one by one. For this purpose the former macro \mhl calling \tcbox is recast into an internal macro \mhl@int. The line breaking works in the first level of \mhl, but not in the nested one.

New code:

\documentclass{article}
\usepackage{tcolorbox}
\usepackage{trimspaces}
\usepackage{listofitems}
\tcbuselibrary{skins,breakable}
%
\makeatletter
\newcommand{\trim}[1]{\trim@spaces@noexp{#1}}
\newlength{\mhl@spacelen}
\settowidth{\mhl@spacelen}{\ }
\def\mhl@space{\kern\mhl@spacelen}
\def\mhl@backspace{\kern-1.1\mhl@spacelen}
%
\newcommand{\mhl@int}[1]{%
\mhl@backspace%
\tcbox[tcbox raise base,
left=0mm,right=0mm,top=0mm,bottom=0mm,boxsep=0pt,arc=0mm,boxrule=0pt,
opacityfill=0.3,enhanced jigsaw,colback=gray!85!white,
before=\relax,after=\relax]{\trim{#1}}
}
%
\newcommand{\mhl}[1]{%
\setsepchar{ }
\ignoreemptyitems
\readlist\phrase{#1}
\foreachitem\mot\in\phrase{\mhl@int{\strut\mot\mhl@space}}
}
%
\makeatother
%
\begin{document}
I say \mhl{Hello World}! This box has a no \mhl{defined \mhl{height}} 
as you can see. The total colored \mhl{box is shrunk to 
the\mhl{dimensions} of the upper part. \mhl{There} should be 
no lower part and no title.}
\end{document}

New result:

Further improvement would likely use recursive calls to mhl with a tracking of the level to get rid of some extra spaces and pushes the breakability into the nested highlighted sentences.

Edit: full solution

After a while, I finaly obtained a full solution of this question. It still rely on tcolorbox for coloring the background and on parsing the text with listofitems. As nested tcolorbox are not breakable, the procedure loop on the sentence, highlighting it word by word, keeping a track of the level, and handling space as carefully as possible.

Here is the MWE (sorry for the French words, phrase=sentence, mot=word):

\documentclass{standalone}
\usepackage{listofitems}
\usepackage{xstring}
\usepackage{tcolorbox}
\tcbuselibrary{skins,breakable,xparse}
%
\makeatletter
\newlength{\mhl@spacelen}
\settowidth{\mhl@spacelen}{\space}
\def\mhl@space{\hspace*{\mhl@spacelen}}
\def\mhl@sep{\hspace{\z@}}
\newcounter{nesthl}
\newif\ifmhlstart \mhlstartfalse
\newif\ifmhlend \mhlendfalse
%

\newcommand{\mhl@int}[2]{%
\ifcase#1\def\col{black!10!white}\or\def\col{black!10!white}\or\def\col{black!50!white}\or\def\col{black!85!white}\fi%
\tcbox[tcbox raise base, spartan, left=0mm,right=0mm,top=0mm,bottom=0mm,boxsep=0pt,arc=0mm,
boxrule=0pt,before=\relax,after=\relax, opacityfill=0.35,colback=\col
]{\strut #2}}

\newcommand{\mhl}[1]{%
\setcounter{nesthl}{1}%
\def\beforeskip{}%
\setsepchar{ }\ignoreemptyitems\readlist*\phrase{#1}%
\foreachitem\mot\in\phrase{%
    \IfBeginWith{\mot}{+}{\mhlstarttrue}{\mhlstartfalse}%
    \IfEndWith{\mot}{+}{\mhlendtrue}{\mhlendfalse}%
    \ifmhlstart \StrBehind{\mot}{+}[\nmot]\else\edef\nmot{\mot}\fi%
    \ifmhlend\StrBefore{\nmot}{+}[\nnmot]\else\edef\nnmot{\nmot}\fi%
    \ifmhlstart\addtocounter{nesthl}{1}\fi%
    \ifmhlend \ifnum\thenesthl>1 \def\beforeskip{\mhl@space}\fi
    \mhl@int{\thenesthl}{\nnmot}\mhl@sep\addtocounter{nesthl}{-1}%
    \else\mhl@int{\thenesthl}{\beforeskip\nnmot\mhl@space}\mhl@sep\def\beforeskip{}\fi%
    }
}
\makeatother

\begin{document}
\framebox{
\begin{minipage}{3.7cm}
here is a  \mhl{weird nested, +wrappable, +dummy blob+ and nice+} text
\end{minipage}
}
\end{document}

and the result:

The key idea in this solution consists in not using nested \mhl command that would prevent wrapping, but to use a simple wiki-like markup inside the \mhl command, with + (it works also with *, and likely punctuation character if they are not active). Words starting with a + (without space) initialize a new level, and those finishing with a + close this level. The number of + must be even, and their relative position does matter.

The maximal nesting depth is not really limited, but to use a larger one than the 3 levels used here, one would have to modify the \mhl@int macro to enlarge the number of shades available (first line) and likely the opacityfill parameter of the boxes.