MWE: framebox and RTL issue

[mahmadzaid]

\begin[papersize=b6,class=plain]{document}
\use[module=packages.frametricks]
\use[module=packages.framebox]
\use[module=packages.rules]
\use[module=packages.color]
\nofolios
\neverindent
\define[command=sbox]{\framebox[bordercolor=#008000]{\color[color=#8b0000]{\process}}}
\define[command=hlght]{\color[color=#8b0000]{\process}}
\define[command=cmnt]{\breakframevertical\font[size=0.8em]{\color[color=#666666]{\process}}\smallskip}
\define[command=rtlframe]{\breakframevertical\thisframedirection[direction=RTL]}
\define[command=ltrframe]{\breakframevertical}
\font[family="Amiri",size=12pt]


\cmnt{LTR language:}

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua.

\bigskip

\cmnt{LTR}
\ltrframe
Lorem ipsum dolor sit amet, \sbox{consectetur adipiscing} elit, sed do eiusmod tempor \sbox{incididunt} ut labore et dolore magna aliqua.

\bigskip

\cmnt{RTL}
\rtlframe
Lorem ipsum dolor sit amet, \sbox{consectetur adipiscing} elit, sed do eiusmod tempor \sbox{incididunt} ut labore et dolore magna aliqua.

\medskip\fullrule\medskip

\cmnt{RTL language:}
\rtlframe
عندما يريد العالم أن يتكلّم، فهو يتحدّث بلغة \hlght{يونيكود}.

عندما يريد العالم أن يتكلّم، فهو يتحدّث بلغة \hlght{unicode}.

\bigskip

\cmnt{LTR}
\ltrframe
عندما يريد العالم أن يتكلّم، فهو يتحدّث بلغة \hlght{unicode}.

عندما يريد العالم أن يتكلّم، فهو يتحدّث بلغة \sbox{unicode}.

عندما يريد العالم أن يتكلّم، فهو يتحدّث بلغة \sbox{يونيكود}.

\bigskip

\cmnt{RTL}
\rtlframe
عندما يريد العالم أن يتكلّم، فهو يتحدّث بلغة \hlght{unicode}.

عندما يريد العالم أن يتكلّم، فهو يتحدّث بلغة \sbox{unicode}.

عندما يريد العالم أن يتكلّم، فهو يتحدّث بلغة \sbox{يونيكود}.

\end{document}

Notice how the highlighted arabic word is rendered with the letters in reverse order.

[Omikhleia]

Doesn't the same issue occur with a mere \hbox or a (from the rules package, deprecated but still there) \boxaround ?

[Omikhleia]

As far as I can tell, it does occur with a mere hbox and stock SILE 0.14.9

I am interested in a double-check (I can't read arabic, so I may miss some point), but I'm afraid it's a bug with SILE then. (And whenever fixed, we might have to check this package again, it has some "compatibility shim" still in place for earlier versions of SILE that might need to be revisited too).

But for the time being, it seems to me it's a SILE bug, that should be reported there.

[mahmadzaid]

I'm getting the same results with \hbox, so you're right, it's a SILE bug.

[Omikhleia]

Couldn't find an existing SILE issue that would correspond perfectly, so perhaps can your report it there? You'd be better at simplifying a MWE than I ^^
Note: I'd suggest removing the colors, as there are some issues with them in bidi mode, it seems.

[mahmadzaid]

reported to SILE here sile-typesetter/sile#1796

[Omikhleia]

For references, see:

• Core SILE: PR (Draft, at of now) WIP: Initial change to move bidi to core, get bidi working in hboxes sile-typesetter/sile#2278
• Resilient documents: issue Refactor the bidi logic in resilient resilient.sile#150 and related PR Move bidi logic into the sile·nt typesetter resilient.sile#152 - Planned for v3.2.0 soon.

Here is how the document from the original document, with the resilient.book class instead of plain would look:

(Well, resilient will forbid the use of the balanced-frames packages and some of the "frametricks", so I re-enabled them for the test)

@mahmadzaid If your are still around after that long a time (I would understand otherwise), I'll be interested in your feedback whether this does meet expectations or not. (Something still looks wrong to me.)

[mimbrown]

@Omikhleia I think that's working according to the bidi algorithm properly (from what I can see). The one that looks weird is this:

To our eyes that looks wrong, since the content of the hbox (يونيكود) is also RTL, so it seems like it should come to the left of the other content. But the algorithm treats the hbox as a black box, so practically it inherits the paragraph's bidi level. What would be nice would be the ability to set the direction of the box: \sbox[direction = RTL]{يونيكود}. That would give the user the ability to define what they want to happen.

[Omikhleia]

What would be nice would be the ability to set the direction of the box

Hmm. That would be fragile in case of cascades of reboxing (e.g. rotating, scaling, framing, parboxing etc.). An explicit marking would also be very inconvenient when not using (SIL but a higher-lever / more abstract / lightweight input format (e.g. Markdown) where boxes can be hidden in some constructs. Frankly I don't know. I think that what we have for resilient v3.2.0 is better anyway (less obscure monkey-patching of typesetters by packages, more documented workflows, better separation of concerns etc.) and I'll go for it if other tests are passing. Maybe we cannot expect fixing issues open for 10+ years (sile-typesetter/sile#76)... but at least we can try to simplify the ugly code...