Inline: Escaped characters stripped vs preserved at word boundaries

[nlopes]

Problem

acdc and asciidoctor handle backslash escapes differently at word boundaries.

Reproduction

Text with \^caret at word boundary.
Also \*asterisk at boundary.
Mid-word es\*cape test.

acdc output:

<p>Text with ^caret at word boundary.</p>
<p>Also *asterisk at boundary.</p>
<p>Mid-word es*cape test.</p>

asciidoctor output:

<p>Text with \^caret at word boundary.</p>
<p>Also \*asterisk at boundary.</p>
<p>Mid-word es\*cape test.</p>

Analysis

• acdc: Always strips the backslash, outputs the literal character
• asciidoctor: Keeps the backslash when it's at a word boundary (not actually escaping anything)

The difference is subtle: when \* isn't actually preventing formatting (because *text without closing * wouldn't be bold anyway), asciidoctor preserves the backslash literally.

Expected behavior

Match asciidoctor: only strip backslash when it's actually preventing a formatting interpretation.

Implementation notes

This requires changes to inline processing in acdc-parser/src/grammar/inline_processing.rs or the escape handling in converters/core/src/substitutions.rs.

The logic needs to check: "Would this character sequence have been interpreted as formatting if the backslash wasn't there?" Only strip the backslash if yes.

Workaround

Use passthrough: pass:[*] or pass:[\*]

Priority

Low - edge case, passthrough workaround available.

[nlopes]

Not sure about the right approach here. I don't understand what keeping the backslash in the output adds or alternatively prevents.