use unicode_bidi::BidiInfo;
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub struct BidiRun {
pub start: usize,
pub end: usize,
pub rtl: bool,
}
pub fn visual_runs(text: &str) -> Vec<BidiRun> {
if text.is_empty() {
return vec![];
}
let info = BidiInfo::new(text, None);
let mut out = Vec::new();
for para in &info.paragraphs {
let line = para.range.clone();
let (levels, runs) = info.visual_runs(para, line);
for range in runs {
if range.start >= range.end {
continue;
}
let rtl = levels.get(range.start).map(|l| l.is_rtl()).unwrap_or(false);
out.push(BidiRun {
start: range.start,
end: range.end,
rtl,
});
}
}
out
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn latin_is_one_ltr_run() {
let runs = visual_runs("Hello");
assert_eq!(
runs,
vec![BidiRun {
start: 0,
end: 5,
rtl: false
}]
);
}
#[test]
fn hebrew_word_is_rtl() {
let text = "שלום";
let runs = visual_runs(text);
assert_eq!(runs.len(), 1);
assert!(runs[0].rtl);
assert_eq!(&text[runs[0].start..runs[0].end], text);
}
#[test]
fn mixed_english_hebrew_splits_before_shaping() {
let text = "Hello שלום";
let runs = visual_runs(text);
assert!(
runs.len() >= 2,
"expected separate LTR and RTL runs, got {runs:?}"
);
let latin = runs.iter().find(|r| !r.rtl).unwrap();
let hebrew = runs.iter().find(|r| r.rtl).unwrap();
assert!(text[latin.start..latin.end].contains("Hello"));
assert!(text[hebrew.start..hebrew.end].contains("שלום"));
assert!(!runs[0].rtl);
}
}