forme/text/

mod.rs

//! # Text Layout
//!
//! Line breaking, text measurement, and glyph positioning.
//!
//! Uses real font metrics from the FontContext for accurate character widths.

pub mod bidi;
pub mod knuth_plass;
pub mod shaping;

use crate::font::FontContext;
use crate::layout::{PAGE_NUMBER_SENTINEL, TOTAL_PAGES_SENTINEL};
use crate::style::{Color, FontStyle, Hyphens, TextDecoration};
use unicode_linebreak::{linebreaks, BreakOpportunity};

/// A line of text after line-breaking.
#[derive(Debug, Clone)]
pub struct BrokenLine {
    /// The characters on this line.
    pub chars: Vec<char>,
    /// The text as a string.
    pub text: String,
    /// X position of each character relative to line start.
    pub char_positions: Vec<f64>,
    /// Total width of the line.
    pub width: f64,
}

/// A styled character for multi-style line breaking.
#[derive(Debug, Clone)]
pub struct StyledChar {
    pub ch: char,
    pub font_family: String,
    pub font_size: f64,
    pub font_weight: u32,
    pub font_style: FontStyle,
    pub color: Color,
    pub href: Option<String>,
    pub text_decoration: TextDecoration,
    pub letter_spacing: f64,
    /// Added to the advance of word-separator characters (the space).
    /// Measured here, alongside letter_spacing, so line breaking sees it:
    /// it used to be applied only at PDF-write time via `Tw`, so a
    /// word-spaced line was broken as though it had no word spacing and
    /// then rendered wider than the box it was measured into.
    pub word_spacing: f64,
}

/// A line of text from multi-style (runs) line breaking.
#[derive(Debug, Clone)]
pub struct RunBrokenLine {
    pub chars: Vec<StyledChar>,
    pub char_positions: Vec<f64>,
    pub width: f64,
}

/// Override widths for page placeholder sentinel characters.
/// Delegates to `FontContext::char_width` which returns the correct
/// width based on the current `sentinel_digit_count`.
#[allow(clippy::too_many_arguments)]
fn fix_sentinel_widths(
    chars: &[char],
    widths: &mut [f64],
    font_context: &FontContext,
    font_family: &str,
    font_weight: u32,
    italic: bool,
    font_size: f64,
    letter_spacing: f64,
) {
    for (i, &ch) in chars.iter().enumerate() {
        if ch == PAGE_NUMBER_SENTINEL || ch == TOTAL_PAGES_SENTINEL {
            widths[i] = font_context.char_width(ch, font_family, font_weight, italic, font_size)
                + letter_spacing;
        }
    }
}

/// The per-character advance a spacing pair contributes.
///
/// CSS applies `word-spacing` to word-separator characters; the space is the
/// only one in this subset. Kept as one function so every measurement site
/// answers the question identically: `word_spacing` used to be absent from
/// this file entirely, applied only at PDF-write time through `Tw`, so lines
/// were broken as though it were zero and then rendered wider than the box
/// they were measured into.
#[inline]
fn extra_advance(ch: char, letter_spacing: f64, word_spacing: f64) -> f64 {
    letter_spacing + if ch == ' ' { word_spacing } else { 0.0 }
}

/// Compute UAX#14 break opportunities indexed by char position.
///
/// Returns a vec of length `text.chars().count()`. Each entry is the break
/// opportunity *before* that character position (i.e. "can we break before
/// char[i]?"). Index 0 is always `None` (no break before the first char).
fn compute_break_opportunities(text: &str) -> Vec<Option<BreakOpportunity>> {
    let char_count = text.chars().count();
    let mut result = vec![None; char_count];

    // linebreaks() yields (byte_offset, opportunity) where byte_offset is the
    // position AFTER the break — i.e. the start of the next segment.
    // We need to convert byte offsets to char indices.
    let byte_to_char: Vec<usize> = {
        let mut map = vec![0usize; text.len() + 1];
        let mut char_idx = 0;
        for (byte_idx, _) in text.char_indices() {
            map[byte_idx] = char_idx;
            char_idx += 1;
        }
        map[text.len()] = char_idx;
        map
    };

    for (byte_offset, opp) in linebreaks(text) {
        let char_idx = byte_to_char[byte_offset];
        if char_idx < char_count {
            result[char_idx] = Some(opp);
        }
        // byte_offset == text.len() means "break at end" — we ignore that
    }

    // Suppress breaks before and after sentinel characters so they stay
    // glued to surrounding text (e.g. "Page \x02" won't break between them).
    let chars: Vec<char> = text.chars().collect();
    for (i, &ch) in chars.iter().enumerate() {
        if ch == PAGE_NUMBER_SENTINEL || ch == TOTAL_PAGES_SENTINEL {
            // No break before the sentinel
            result[i] = None;
            // No break after the sentinel
            if i + 1 < char_count {
                result[i + 1] = None;
            }
        }
    }

    result
}

/// Map a BCP 47 language tag to a `hypher::Lang` for hyphenation.
///
/// Returns `Some(lang)` for supported languages, `None` for unsupported ones
/// (which disables algorithmic hyphenation). Defaults to English when no tag
/// is provided, for backward compatibility.
fn resolve_hypher_lang(lang: Option<&str>) -> Option<hypher::Lang> {
    let tag = match lang {
        Some(t) => t,
        None => return Some(hypher::Lang::English),
    };
    let primary = tag.split('-').next().unwrap_or(tag).to_lowercase();
    match primary.as_str() {
        "af" => Some(hypher::Lang::Afrikaans),
        "sq" => Some(hypher::Lang::Albanian),
        "be" => Some(hypher::Lang::Belarusian),
        "bg" => Some(hypher::Lang::Bulgarian),
        "ca" => Some(hypher::Lang::Catalan),
        "hr" => Some(hypher::Lang::Croatian),
        "cs" => Some(hypher::Lang::Czech),
        "da" => Some(hypher::Lang::Danish),
        "nl" => Some(hypher::Lang::Dutch),
        "en" => Some(hypher::Lang::English),
        "et" => Some(hypher::Lang::Estonian),
        "fi" => Some(hypher::Lang::Finnish),
        "fr" => Some(hypher::Lang::French),
        "ka" => Some(hypher::Lang::Georgian),
        "de" => Some(hypher::Lang::German),
        "el" => Some(hypher::Lang::Greek),
        "hu" => Some(hypher::Lang::Hungarian),
        "is" => Some(hypher::Lang::Icelandic),
        "it" => Some(hypher::Lang::Italian),
        "ku" => Some(hypher::Lang::Kurmanji),
        "la" => Some(hypher::Lang::Latin),
        "lt" => Some(hypher::Lang::Lithuanian),
        "mn" => Some(hypher::Lang::Mongolian),
        "nb" | "nn" | "no" => Some(hypher::Lang::Norwegian),
        "pl" => Some(hypher::Lang::Polish),
        "pt" => Some(hypher::Lang::Portuguese),
        "ru" => Some(hypher::Lang::Russian),
        "sr" => Some(hypher::Lang::Serbian),
        "sk" => Some(hypher::Lang::Slovak),
        "sl" => Some(hypher::Lang::Slovenian),
        "es" => Some(hypher::Lang::Spanish),
        "sv" => Some(hypher::Lang::Swedish),
        "tr" => Some(hypher::Lang::Turkish),
        "tk" => Some(hypher::Lang::Turkmen),
        "uk" => Some(hypher::Lang::Ukrainian),
        _ => None,
    }
}

pub struct TextLayout;

impl Default for TextLayout {
    fn default() -> Self {
        Self::new()
    }
}

impl TextLayout {
    pub fn new() -> Self {
        Self
    }

    /// Break a string into lines that fit within `max_width`.
    ///
    /// Uses a greedy line-breaking algorithm with optional hyphenation.
    /// When `hyphens` is `Auto`, long words that don't fit are split at
    /// syllable boundaries using the Knuth-Liang algorithm. When `Manual`,
    /// only soft hyphens (U+00AD) in the text are used as break points.
    #[allow(clippy::too_many_arguments)]
    pub fn break_into_lines(
        &self,
        font_context: &FontContext,
        text: &str,
        max_width: f64,
        font_size: f64,
        font_family: &str,
        font_weight: u32,
        font_style: FontStyle,
        letter_spacing: f64,
        word_spacing: f64,
        hyphens: Hyphens,
        lang: Option<&str>,
    ) -> Vec<BrokenLine> {
        if text.is_empty() {
            return vec![BrokenLine {
                chars: vec![],
                text: String::new(),
                char_positions: vec![],
                width: 0.0,
            }];
        }

        let char_widths = self.measure_chars(
            font_context,
            text,
            font_size,
            font_family,
            font_weight,
            font_style,
            letter_spacing,
            word_spacing,
        );

        let hyphen_width = font_context.char_width(
            '-',
            font_family,
            font_weight,
            matches!(font_style, FontStyle::Italic | FontStyle::Oblique),
            font_size,
        ) + letter_spacing;

        let mut lines = Vec::new();
        let mut line_start = 0;
        let mut line_width = 0.0;
        let mut last_break_point = None;
        let mut _last_break_width = 0.0;

        let chars: Vec<char> = text.chars().collect();
        let break_opps = compute_break_opportunities(text);

        for (i, &ch) in chars.iter().enumerate() {
            let char_width = char_widths[i];

            // UAX#14 break opportunities: a break *before* char[i] means
            // we can end the previous line at char[i-1].
            // We record break points at i-1 (the char before the break).
            if i > 0 {
                if let Some(opp) = break_opps[i] {
                    match opp {
                        BreakOpportunity::Mandatory => {
                            // Mandatory break: flush the current line
                            let end = if chars[i - 1] == '\n'
                                || chars[i - 1] == '\r'
                                || chars[i - 1] == '\u{2028}'
                                || chars[i - 1] == '\u{2029}'
                            {
                                i - 1
                            } else {
                                i
                            };
                            let line_chars = self.filter_soft_hyphens(&chars[line_start..end]);
                            let line_widths = self.filter_soft_hyphen_widths(
                                &chars[line_start..end],
                                &char_widths[line_start..end],
                            );
                            lines.push(self.make_line(&line_chars, &line_widths));
                            line_start = i;
                            line_width = 0.0;
                            last_break_point = None;
                            // Don't skip — still need to process char[i] width below
                        }
                        BreakOpportunity::Allowed => {
                            // Record the char BEFORE this position as a break point
                            last_break_point = Some(i - 1);
                            _last_break_width = line_width;
                        }
                    }
                }
            }

            // Soft hyphens are additional break points for Manual and Auto modes
            if ch == '\u{00AD}' && hyphens != Hyphens::None {
                last_break_point = Some(i);
                _last_break_width = line_width;
            }

            // Soft hyphens are zero-width when not at a break
            if ch == '\u{00AD}' {
                continue;
            }

            // Skip newline/CR chars (already handled by mandatory break above)
            if ch == '\n' || ch == '\r' || ch == '\u{2028}' || ch == '\u{2029}' {
                continue;
            }

            if line_width + char_width > max_width && line_start < i {
                // Line overflow — break at the last break point if possible
                if let Some(bp) = last_break_point {
                    if bp >= line_start {
                        if chars[bp] == '\u{00AD}' {
                            // Break at soft hyphen: render visible hyphen
                            let mut line_chars = self.filter_soft_hyphens(&chars[line_start..bp]);
                            let mut line_widths = self.filter_soft_hyphen_widths(
                                &chars[line_start..bp],
                                &char_widths[line_start..bp],
                            );
                            line_chars.push('-');
                            line_widths.push(hyphen_width);
                            lines.push(self.make_line(&line_chars, &line_widths));
                        } else {
                            // bp is the last char on this line (UAX#14 break is *after* bp)
                            let break_at = bp + 1;
                            let line_chars = self.filter_soft_hyphens(&chars[line_start..break_at]);
                            let line_widths = self.filter_soft_hyphen_widths(
                                &chars[line_start..break_at],
                                &char_widths[line_start..break_at],
                            );
                            lines.push(self.make_line(&line_chars, &line_widths));
                        }

                        line_start = bp + 1;
                        // Recalculate width excluding soft hyphens
                        line_width = chars[line_start..=i]
                            .iter()
                            .zip(char_widths[line_start..=i].iter())
                            .filter(|(c, _)| **c != '\u{00AD}')
                            .map(|(_, w)| w)
                            .sum();
                        last_break_point = None;
                        continue;
                    }
                }

                // No space/hyphen break point — try algorithmic hyphenation
                if hyphens == Hyphens::Auto {
                    if let Some((hyphen_line_chars, hyphen_line_widths, new_start)) = self
                        .try_hyphenate_word(
                            &chars,
                            &char_widths,
                            line_start,
                            i,
                            line_width,
                            max_width,
                            hyphen_width,
                            lang,
                        )
                    {
                        lines.push(self.make_line(&hyphen_line_chars, &hyphen_line_widths));
                        line_start = new_start;
                        line_width = chars[line_start..=i]
                            .iter()
                            .zip(char_widths[line_start..=i].iter())
                            .filter(|(c, _)| **c != '\u{00AD}')
                            .map(|(_, w)| w)
                            .sum();
                        last_break_point = None;
                        continue;
                    }
                }

                // No good break point — force break at current position
                let line_chars = self.filter_soft_hyphens(&chars[line_start..i]);
                let line_widths = self
                    .filter_soft_hyphen_widths(&chars[line_start..i], &char_widths[line_start..i]);
                lines.push(self.make_line(&line_chars, &line_widths));
                line_start = i;
                line_width = char_width;
                last_break_point = None;
                continue;
            }

            line_width += char_width;
        }

        // Last line
        if line_start < chars.len() {
            let line_chars = self.filter_soft_hyphens(&chars[line_start..]);
            let line_widths =
                self.filter_soft_hyphen_widths(&chars[line_start..], &char_widths[line_start..]);
            lines.push(self.make_line(&line_chars, &line_widths));
        }

        lines
    }

    /// Create a BrokenLine from characters and their widths.
    fn make_line(&self, chars: &[char], widths: &[f64]) -> BrokenLine {
        let mut positions = Vec::with_capacity(chars.len());
        let mut x = 0.0;
        for &w in widths {
            positions.push(x);
            x += w;
        }

        // Trim trailing spaces from width calculation
        let mut effective_width = x;
        let mut i = chars.len();
        while i > 0 && chars[i - 1] == ' ' {
            i -= 1;
            effective_width -= widths[i];
        }

        BrokenLine {
            text: chars.iter().collect(),
            chars: chars.to_vec(),
            char_positions: positions,
            width: effective_width,
        }
    }

    /// Filter out soft hyphens from a char slice.
    fn filter_soft_hyphens(&self, chars: &[char]) -> Vec<char> {
        chars.iter().copied().filter(|c| *c != '\u{00AD}').collect()
    }

    /// Filter out widths corresponding to soft hyphens.
    fn filter_soft_hyphen_widths(&self, chars: &[char], widths: &[f64]) -> Vec<f64> {
        chars
            .iter()
            .zip(widths.iter())
            .filter(|(c, _)| **c != '\u{00AD}')
            .map(|(_, w)| *w)
            .collect()
    }

    /// Try to hyphenate the current word at a syllable boundary that fits.
    ///
    /// Looks backward from the overflow point to find word boundaries, then
    /// uses `hypher` to find syllable breaks within the word. Returns the
    /// rightmost break that fits (with hyphen char appended).
    ///
    /// Returns `Some((line_chars, line_widths, new_line_start))` on success.
    #[allow(clippy::too_many_arguments)]
    fn try_hyphenate_word(
        &self,
        chars: &[char],
        char_widths: &[f64],
        line_start: usize,
        overflow_at: usize,
        _line_width: f64,
        max_width: f64,
        hyphen_width: f64,
        lang: Option<&str>,
    ) -> Option<(Vec<char>, Vec<f64>, usize)> {
        // Find the start of the current word (scan backward from overflow)
        let mut word_start = overflow_at;
        while word_start > line_start && !chars[word_start - 1].is_whitespace() {
            word_start -= 1;
        }

        // Collect the word chars (up to and including overflow_at - 1)
        let word_end = overflow_at; // exclusive — the char at overflow_at triggered overflow
        if word_end <= word_start {
            return None;
        }

        let word: String = chars[word_start..word_end].iter().collect();
        let hypher_lang = resolve_hypher_lang(lang)?;
        let syllables = hypher::hyphenate(&word, hypher_lang);

        let syllables: Vec<&str> = syllables.collect();
        if syllables.len() < 2 {
            return None;
        }

        // Width of content before the word on this line
        let prefix_width: f64 = chars[line_start..word_start]
            .iter()
            .zip(char_widths[line_start..word_start].iter())
            .filter(|(c, _)| **c != '\u{00AD}')
            .map(|(_, w)| w)
            .sum();

        // Find the rightmost syllable boundary that fits
        let mut best_break: Option<usize> = None; // index into chars[] to break AFTER
        let mut syllable_offset = word_start;
        for (si, syllable) in syllables.iter().enumerate() {
            if si == syllables.len() - 1 {
                break; // don't break after the last syllable
            }
            syllable_offset += syllable.chars().count();

            // Width of word chars from word_start..syllable_offset
            let word_part_width: f64 = chars[word_start..syllable_offset]
                .iter()
                .zip(char_widths[word_start..syllable_offset].iter())
                .filter(|(c, _)| **c != '\u{00AD}')
                .map(|(_, w)| w)
                .sum();

            if prefix_width + word_part_width + hyphen_width <= max_width {
                best_break = Some(syllable_offset);
            }
        }

        let break_at = best_break?;

        let mut line_chars = self.filter_soft_hyphens(&chars[line_start..break_at]);
        let mut line_widths = self.filter_soft_hyphen_widths(
            &chars[line_start..break_at],
            &char_widths[line_start..break_at],
        );
        line_chars.push('-');
        line_widths.push(hyphen_width);

        Some((line_chars, line_widths, break_at))
    }

    /// Measure individual character widths using real font metrics.
    ///
    /// For custom fonts with available font data, uses OpenType shaping via
    /// rustybuzz to produce accurate widths that account for kerning and
    /// ligatures. For standard fonts, uses per-char width lookup.
    #[allow(clippy::too_many_arguments)]
    fn measure_chars(
        &self,
        font_context: &FontContext,
        text: &str,
        font_size: f64,
        font_family: &str,
        font_weight: u32,
        font_style: FontStyle,
        letter_spacing: f64,
        word_spacing: f64,
    ) -> Vec<f64> {
        let italic = matches!(font_style, FontStyle::Italic | FontStyle::Oblique);
        let chars: Vec<char> = text.chars().collect();

        // Detect if BiDi shaping is needed (Arabic, Hebrew, etc.)
        let has_bidi = !bidi::is_pure_ltr(text, crate::style::Direction::Auto);
        let bidi_runs = if has_bidi {
            bidi::analyze_bidi(text, crate::style::Direction::Auto)
        } else {
            vec![]
        };

        // Fast path: single font family (no comma)
        if !font_family.contains(',') {
            if let Some(font_data) = font_context.font_data(font_family, font_weight, italic) {
                let units_per_em = font_context.units_per_em(font_family, font_weight, italic);

                if has_bidi {
                    // Shape each BiDi run with correct direction to match glyph builder
                    let mut widths = vec![0.0_f64; chars.len()];
                    for bidi_run in &bidi_runs {
                        let run_text: String = chars[bidi_run.char_start..bidi_run.char_end]
                            .iter()
                            .collect();
                        if let Some(shaped) = shaping::shape_text_with_direction(
                            &run_text,
                            font_data,
                            bidi_run.is_rtl,
                        ) {
                            let num_chars = bidi_run.char_end - bidi_run.char_start;
                            let cluster_w = shaping::cluster_widths(
                                &shaped,
                                num_chars,
                                units_per_em,
                                font_size,
                                letter_spacing,
                            );
                            for (j, w) in cluster_w.into_iter().enumerate() {
                                widths[bidi_run.char_start + j] = w;
                            }
                        } else {
                            for i in bidi_run.char_start..bidi_run.char_end {
                                widths[i] =
                                    font_context.char_width(
                                        chars[i],
                                        font_family,
                                        font_weight,
                                        italic,
                                        font_size,
                                    ) + extra_advance(chars[i], letter_spacing, word_spacing);
                            }
                        }
                    }
                    fix_sentinel_widths(
                        &chars,
                        &mut widths,
                        font_context,
                        font_family,
                        font_weight,
                        italic,
                        font_size,
                        letter_spacing,
                    );
                    return widths;
                }

                if let Some(shaped) = shaping::shape_text(text, font_data) {
                    let num_chars = chars.len();
                    let mut widths = shaping::cluster_widths(
                        &shaped,
                        num_chars,
                        units_per_em,
                        font_size,
                        letter_spacing,
                    );
                    fix_sentinel_widths(
                        &chars,
                        &mut widths,
                        font_context,
                        font_family,
                        font_weight,
                        italic,
                        font_size,
                        letter_spacing,
                    );
                    return widths;
                }
            }

            return text
                .chars()
                .map(|ch| {
                    font_context.char_width(ch, font_family, font_weight, italic, font_size)
                        + extra_advance(ch, letter_spacing, word_spacing)
                })
                .collect();
        }

        // Per-char fallback path: use context-independent per-character widths.
        // We intentionally avoid full-text shaping here because the glyph builder
        // shapes individual lines (different shaping context), which can produce
        // different total widths. Per-char hmtx widths are stable regardless of
        // context, so the line breaker never underestimates and lines never overflow.
        // The glyph builder still uses shaping for visual quality (kerning, ligatures).
        chars
            .iter()
            .map(|&ch| {
                font_context.char_width(ch, font_family, font_weight, italic, font_size)
                    + extra_advance(ch, letter_spacing, word_spacing)
            })
            .collect()
    }

    /// Shape text and return shaped glyphs for a custom font.
    /// Returns `None` for standard fonts or if shaping fails.
    #[allow(clippy::too_many_arguments)]
    pub fn shape_text(
        &self,
        font_context: &FontContext,
        text: &str,
        font_family: &str,
        font_weight: u32,
        font_style: FontStyle,
    ) -> Option<Vec<shaping::ShapedGlyph>> {
        let italic = matches!(font_style, FontStyle::Italic | FontStyle::Oblique);
        let font_data = font_context.font_data(font_family, font_weight, italic)?;
        shaping::shape_text(text, font_data)
    }

    /// Measure widths for styled chars, using shaping for contiguous runs
    /// that share the same custom font. Falls back to per-char measurement
    /// for standard fonts or when shaping fails.
    fn measure_styled_chars(&self, font_context: &FontContext, chars: &[StyledChar]) -> Vec<f64> {
        if chars.is_empty() {
            return vec![];
        }

        let mut widths = vec![0.0_f64; chars.len()];
        let mut i = 0;

        while i < chars.len() {
            let sc = &chars[i];
            let italic = matches!(sc.font_style, FontStyle::Italic | FontStyle::Oblique);

            // Check if this char's font is a custom font with shaping data
            if let Some(font_data) = font_context.font_data(&sc.font_family, sc.font_weight, italic)
            {
                // Find the end of the contiguous run with the same font
                let run_start = i;
                let mut run_end = i + 1;
                while run_end < chars.len() {
                    let next = &chars[run_end];
                    let next_italic =
                        matches!(next.font_style, FontStyle::Italic | FontStyle::Oblique);
                    if next.font_family == sc.font_family
                        && next.font_weight == sc.font_weight
                        && next_italic == italic
                        && (next.font_size - sc.font_size).abs() < 0.001
                    {
                        run_end += 1;
                    } else {
                        break;
                    }
                }

                // Shape this run
                let run_text: String = chars[run_start..run_end].iter().map(|c| c.ch).collect();
                if let Some(shaped) = shaping::shape_text(&run_text, font_data) {
                    let num_chars = run_end - run_start;
                    let units_per_em =
                        font_context.units_per_em(&sc.font_family, sc.font_weight, italic);
                    let cluster_w = shaping::cluster_widths(
                        &shaped,
                        num_chars,
                        units_per_em,
                        sc.font_size,
                        sc.letter_spacing,
                    );
                    for (j, w) in cluster_w.into_iter().enumerate() {
                        widths[run_start + j] = w;
                    }
                    // Fix sentinel widths that shaping may have given wrong values
                    for j in run_start..run_end {
                        let ch = chars[j].ch;
                        if ch == PAGE_NUMBER_SENTINEL || ch == TOTAL_PAGES_SENTINEL {
                            widths[j] = font_context.char_width(
                                ch,
                                &chars[j].font_family,
                                chars[j].font_weight,
                                italic,
                                chars[j].font_size,
                            ) + extra_advance(
                                chars[j].ch,
                                chars[j].letter_spacing,
                                chars[j].word_spacing,
                            );
                        }
                    }
                    i = run_end;
                    continue;
                }
            }

            // Fallback: per-char measurement
            widths[i] = font_context.char_width(
                sc.ch,
                &sc.font_family,
                sc.font_weight,
                italic,
                sc.font_size,
            ) + extra_advance(sc.ch, sc.letter_spacing, sc.word_spacing);
            i += 1;
        }

        widths
    }

    /// Break multi-style text (runs) into lines that fit within `max_width`.
    pub fn break_runs_into_lines(
        &self,
        font_context: &FontContext,
        chars: &[StyledChar],
        max_width: f64,
        hyphens: Hyphens,
        lang: Option<&str>,
    ) -> Vec<RunBrokenLine> {
        if chars.is_empty() {
            return vec![RunBrokenLine {
                chars: vec![],
                char_positions: vec![],
                width: 0.0,
            }];
        }

        // Measure each character width using shaping for custom fonts
        let char_widths = self.measure_styled_chars(font_context, chars);

        let mut lines = Vec::new();
        let mut line_start = 0;
        let mut line_width = 0.0;
        let mut last_break_point: Option<usize> = None;

        // Build plain text for UAX#14 break analysis
        let plain_text: String = chars.iter().map(|sc| sc.ch).collect();
        let break_opps = compute_break_opportunities(&plain_text);

        for (i, sc) in chars.iter().enumerate() {
            let char_width = char_widths[i];

            // UAX#14 break opportunities
            if i > 0 {
                if let Some(opp) = break_opps[i] {
                    match opp {
                        BreakOpportunity::Mandatory => {
                            let end = if chars[i - 1].ch == '\n'
                                || chars[i - 1].ch == '\r'
                                || chars[i - 1].ch == '\u{2028}'
                                || chars[i - 1].ch == '\u{2029}'
                            {
                                i - 1
                            } else {
                                i
                            };
                            let filtered = self.filter_soft_hyphens_runs(&chars[line_start..end]);
                            let filtered_widths = self.filter_soft_hyphen_widths_runs(
                                &chars[line_start..end],
                                &char_widths[line_start..end],
                            );
                            lines.push(self.make_run_line(&filtered, &filtered_widths));
                            line_start = i;
                            line_width = 0.0;
                            last_break_point = None;
                        }
                        BreakOpportunity::Allowed => {
                            last_break_point = Some(i - 1);
                        }
                    }
                }
            }

            // Soft hyphens are additional break points
            if sc.ch == '\u{00AD}' && hyphens != Hyphens::None {
                last_break_point = Some(i);
            }

            // Soft hyphens are zero-width when not at a break
            if sc.ch == '\u{00AD}' {
                continue;
            }

            // Skip newline/CR chars (already handled by mandatory break above)
            if sc.ch == '\n' || sc.ch == '\r' || sc.ch == '\u{2028}' || sc.ch == '\u{2029}' {
                continue;
            }

            if line_width + char_width > max_width && line_start < i {
                if let Some(bp) = last_break_point {
                    if bp >= line_start {
                        if chars[bp].ch == '\u{00AD}' {
                            // Break at soft hyphen: render visible hyphen
                            let mut filtered =
                                self.filter_soft_hyphens_runs(&chars[line_start..bp]);
                            let mut filtered_widths = self.filter_soft_hyphen_widths_runs(
                                &chars[line_start..bp],
                                &char_widths[line_start..bp],
                            );
                            // Add a visible hyphen with the style of the char before the soft hyphen
                            let hyphen_style = if bp > 0 {
                                chars[bp - 1].clone()
                            } else {
                                chars[bp].clone()
                            };
                            let italic = matches!(
                                hyphen_style.font_style,
                                FontStyle::Italic | FontStyle::Oblique
                            );
                            let hw = font_context.char_width(
                                '-',
                                &hyphen_style.font_family,
                                hyphen_style.font_weight,
                                italic,
                                hyphen_style.font_size,
                            ) + hyphen_style.letter_spacing;
                            let mut hyphen_sc = hyphen_style;
                            hyphen_sc.ch = '-';
                            filtered.push(hyphen_sc);
                            filtered_widths.push(hw);
                            lines.push(self.make_run_line(&filtered, &filtered_widths));
                        } else {
                            // bp is the last char on this line (UAX#14 break after bp)
                            let break_at = bp + 1;
                            let filtered =
                                self.filter_soft_hyphens_runs(&chars[line_start..break_at]);
                            let filtered_widths = self.filter_soft_hyphen_widths_runs(
                                &chars[line_start..break_at],
                                &char_widths[line_start..break_at],
                            );
                            lines.push(self.make_run_line(&filtered, &filtered_widths));
                        }

                        line_start = bp + 1;
                        line_width = chars[line_start..=i]
                            .iter()
                            .zip(char_widths[line_start..=i].iter())
                            .filter(|(sc, _)| sc.ch != '\u{00AD}')
                            .map(|(_, w)| w)
                            .sum();
                        last_break_point = None;
                        continue;
                    }
                }

                // Try algorithmic hyphenation
                if hyphens == Hyphens::Auto {
                    let plain_chars: Vec<char> = chars.iter().map(|sc| sc.ch).collect();
                    let italic = if !chars.is_empty() {
                        matches!(
                            chars[line_start].font_style,
                            FontStyle::Italic | FontStyle::Oblique
                        )
                    } else {
                        false
                    };
                    let hyphen_width = if !chars.is_empty() {
                        font_context.char_width(
                            '-',
                            &chars[line_start].font_family,
                            chars[line_start].font_weight,
                            italic,
                            chars[line_start].font_size,
                        ) + extra_advance(
                            chars[line_start].ch,
                            chars[line_start].letter_spacing,
                            chars[line_start].word_spacing,
                        )
                    } else {
                        0.0
                    };

                    if let Some((_, _, new_start)) = self.try_hyphenate_word(
                        &plain_chars,
                        &char_widths,
                        line_start,
                        i,
                        line_width,
                        max_width,
                        hyphen_width,
                        lang,
                    ) {
                        // Build the run line with hyphen
                        let mut filtered =
                            self.filter_soft_hyphens_runs(&chars[line_start..new_start]);
                        let mut filtered_widths = self.filter_soft_hyphen_widths_runs(
                            &chars[line_start..new_start],
                            &char_widths[line_start..new_start],
                        );
                        let hyphen_style_ref = if new_start > 0 {
                            &chars[new_start - 1]
                        } else {
                            &chars[0]
                        };
                        let mut hyphen_sc = hyphen_style_ref.clone();
                        hyphen_sc.ch = '-';
                        filtered.push(hyphen_sc);
                        filtered_widths.push(hyphen_width);
                        lines.push(self.make_run_line(&filtered, &filtered_widths));

                        line_start = new_start;
                        line_width = chars[line_start..=i]
                            .iter()
                            .zip(char_widths[line_start..=i].iter())
                            .filter(|(sc, _)| sc.ch != '\u{00AD}')
                            .map(|(_, w)| w)
                            .sum();
                        last_break_point = None;
                        continue;
                    }
                }

                let filtered = self.filter_soft_hyphens_runs(&chars[line_start..i]);
                let filtered_widths = self.filter_soft_hyphen_widths_runs(
                    &chars[line_start..i],
                    &char_widths[line_start..i],
                );
                lines.push(self.make_run_line(&filtered, &filtered_widths));
                line_start = i;
                line_width = char_width;
                last_break_point = None;
                continue;
            }

            line_width += char_width;
        }

        if line_start < chars.len() {
            let filtered = self.filter_soft_hyphens_runs(&chars[line_start..]);
            let filtered_widths = self
                .filter_soft_hyphen_widths_runs(&chars[line_start..], &char_widths[line_start..]);
            lines.push(self.make_run_line(&filtered, &filtered_widths));
        }

        lines
    }

    /// Filter out soft hyphens from styled char slices.
    fn filter_soft_hyphens_runs(&self, chars: &[StyledChar]) -> Vec<StyledChar> {
        chars
            .iter()
            .filter(|sc| sc.ch != '\u{00AD}')
            .cloned()
            .collect()
    }

    /// Filter out widths corresponding to soft hyphens in styled char slices.
    fn filter_soft_hyphen_widths_runs(&self, chars: &[StyledChar], widths: &[f64]) -> Vec<f64> {
        chars
            .iter()
            .zip(widths.iter())
            .filter(|(sc, _)| sc.ch != '\u{00AD}')
            .map(|(_, w)| *w)
            .collect()
    }

    fn make_run_line(&self, chars: &[StyledChar], widths: &[f64]) -> RunBrokenLine {
        let mut positions = Vec::with_capacity(chars.len());
        let mut x = 0.0;
        for &w in widths {
            positions.push(x);
            x += w;
        }

        // Trim trailing spaces from width calculation
        let mut effective_width = x;
        let mut i = chars.len();
        while i > 0 && chars[i - 1].ch == ' ' {
            i -= 1;
            effective_width -= widths[i];
        }

        RunBrokenLine {
            chars: chars.to_vec(),
            char_positions: positions,
            width: effective_width,
        }
    }

    /// Measure the widest single word in a string (min-content width).
    ///
    /// When `hyphens` is `Auto`, returns the widest *syllable* width instead
    /// of the widest word, since hyphenation allows breaking within words.
    #[allow(clippy::too_many_arguments)]
    pub fn measure_widest_word(
        &self,
        font_context: &FontContext,
        text: &str,
        font_size: f64,
        font_family: &str,
        font_weight: u32,
        font_style: FontStyle,
        letter_spacing: f64,
        word_spacing: f64,
        hyphens: Hyphens,
        lang: Option<&str>,
    ) -> f64 {
        if hyphens == Hyphens::Auto {
            if let Some(hypher_lang) = resolve_hypher_lang(lang) {
                // With auto hyphenation, min-content is the widest syllable
                return text
                    .split_whitespace()
                    .flat_map(|word| {
                        let syllables = hypher::hyphenate(word, hypher_lang);
                        syllables
                            .into_iter()
                            .map(|s| {
                                self.measure_width(
                                    font_context,
                                    s,
                                    font_size,
                                    font_family,
                                    font_weight,
                                    font_style,
                                    letter_spacing,
                                    word_spacing,
                                )
                            })
                            .collect::<Vec<_>>()
                    })
                    .fold(0.0f64, f64::max);
            }
            // Unsupported language — fall through to word-level measurement
        }
        text.split_whitespace()
            .map(|word| {
                self.measure_width(
                    font_context,
                    word,
                    font_size,
                    font_family,
                    font_weight,
                    font_style,
                    letter_spacing,
                    word_spacing,
                )
            })
            .fold(0.0f64, f64::max)
    }

    /// Measure the width of a string on a single line.
    #[allow(clippy::too_many_arguments)]
    pub fn measure_width(
        &self,
        font_context: &FontContext,
        text: &str,
        font_size: f64,
        font_family: &str,
        font_weight: u32,
        font_style: FontStyle,
        letter_spacing: f64,
        word_spacing: f64,
    ) -> f64 {
        self.measure_chars(
            font_context,
            text,
            font_size,
            font_family,
            font_weight,
            font_style,
            letter_spacing,
            word_spacing,
        )
        .iter()
        .sum()
    }

    /// Break text into lines using the Knuth-Plass optimal algorithm.
    ///
    /// Falls back to greedy breaking if KP finds no feasible solution.
    #[allow(clippy::too_many_arguments)]
    pub fn break_into_lines_optimal(
        &self,
        font_context: &FontContext,
        text: &str,
        max_width: f64,
        font_size: f64,
        font_family: &str,
        font_weight: u32,
        font_style: FontStyle,
        letter_spacing: f64,
        word_spacing: f64,
        hyphens: Hyphens,
        lang: Option<&str>,
        justify: bool,
    ) -> Vec<BrokenLine> {
        if text.is_empty() {
            return vec![BrokenLine {
                chars: vec![],
                text: String::new(),
                char_positions: vec![],
                width: 0.0,
            }];
        }

        let char_widths = self.measure_chars(
            font_context,
            text,
            font_size,
            font_family,
            font_weight,
            font_style,
            letter_spacing,
            word_spacing,
        );

        let hyphen_width = font_context.char_width(
            '-',
            font_family,
            font_weight,
            matches!(font_style, FontStyle::Italic | FontStyle::Oblique),
            font_size,
        ) + letter_spacing;

        let chars: Vec<char> = text.chars().collect();
        let break_opps = compute_break_opportunities(text);

        // Check for mandatory breaks — if present, handle each segment separately
        let mut segments = Vec::new();
        let mut seg_start = 0;
        for (i, opp) in break_opps.iter().enumerate() {
            if let Some(BreakOpportunity::Mandatory) = opp {
                // End of previous segment is just before this char
                // But the mandatory break could be at \n, so the end is i-1 or earlier
                let end = if i > 0
                    && (chars[i - 1] == '\n'
                        || chars[i - 1] == '\r'
                        || chars[i - 1] == '\u{2028}'
                        || chars[i - 1] == '\u{2029}')
                {
                    i - 1
                } else {
                    i
                };
                segments.push(seg_start..end);
                seg_start = i;
            }
        }
        segments.push(seg_start..chars.len());

        if segments.len() > 1 {
            // Multiple mandatory-break segments: run KP on each
            let mut all_lines = Vec::new();
            for seg in &segments {
                if seg.is_empty() {
                    all_lines.push(BrokenLine {
                        chars: vec![],
                        text: String::new(),
                        char_positions: vec![],
                        width: 0.0,
                    });
                    continue;
                }
                let seg_chars: Vec<char> = chars[seg.clone()]
                    .iter()
                    .copied()
                    .filter(|c| *c != '\n' && *c != '\r' && *c != '\u{2028}' && *c != '\u{2029}')
                    .collect();
                if seg_chars.is_empty() {
                    continue;
                }
                let seg_text: String = seg_chars.iter().collect();
                let seg_lines = self.break_into_lines_optimal(
                    font_context,
                    &seg_text,
                    max_width,
                    font_size,
                    font_family,
                    font_weight,
                    font_style,
                    letter_spacing,
                    word_spacing,
                    hyphens,
                    lang,
                    justify,
                );
                all_lines.extend(seg_lines);
            }
            return all_lines;
        }

        // Single segment — run KP
        let items = knuth_plass::build_items(
            &chars,
            &char_widths,
            hyphen_width,
            hyphens,
            &break_opps,
            lang,
        );
        let config = knuth_plass::Config {
            line_width: max_width,
            ..Default::default()
        };

        if let Some(solutions) = knuth_plass::find_breaks(&items, &config) {
            knuth_plass::reconstruct_lines(
                &solutions,
                &items,
                &chars,
                &char_widths,
                max_width,
                justify,
            )
        } else {
            // Fallback to greedy
            self.break_into_lines(
                font_context,
                text,
                max_width,
                font_size,
                font_family,
                font_weight,
                font_style,
                letter_spacing,
                word_spacing,
                hyphens,
                lang,
            )
        }
    }

    /// Break multi-style text into lines using the Knuth-Plass optimal algorithm.
    ///
    /// Falls back to greedy breaking if KP finds no feasible solution.
    pub fn break_runs_into_lines_optimal(
        &self,
        font_context: &FontContext,
        chars: &[StyledChar],
        max_width: f64,
        hyphens: Hyphens,
        lang: Option<&str>,
        justify: bool,
    ) -> Vec<RunBrokenLine> {
        if chars.is_empty() {
            return vec![RunBrokenLine {
                chars: vec![],
                char_positions: vec![],
                width: 0.0,
            }];
        }

        let char_widths = self.measure_styled_chars(font_context, chars);

        // Use the first char's style for hyphen width
        let hyphen_width = if !chars.is_empty() {
            let sc = &chars[0];
            let italic = matches!(sc.font_style, FontStyle::Italic | FontStyle::Oblique);
            font_context.char_width('-', &sc.font_family, sc.font_weight, italic, sc.font_size)
                + sc.letter_spacing
        } else {
            0.0
        };

        let plain_text: String = chars.iter().map(|sc| sc.ch).collect();
        let break_opps = compute_break_opportunities(&plain_text);

        // Handle mandatory breaks by splitting into segments
        let plain_chars: Vec<char> = chars.iter().map(|sc| sc.ch).collect();
        let has_mandatory = break_opps
            .iter()
            .any(|o| matches!(o, Some(BreakOpportunity::Mandatory)));

        if has_mandatory {
            let mut all_lines = Vec::new();
            let mut seg_start = 0;

            for (i, opp) in break_opps.iter().enumerate() {
                if let Some(BreakOpportunity::Mandatory) = opp {
                    let end = if i > 0
                        && (plain_chars[i - 1] == '\n'
                            || plain_chars[i - 1] == '\r'
                            || plain_chars[i - 1] == '\u{2028}'
                            || plain_chars[i - 1] == '\u{2029}')
                    {
                        i - 1
                    } else {
                        i
                    };
                    let seg_chars: Vec<StyledChar> = chars[seg_start..end]
                        .iter()
                        .filter(|sc| {
                            sc.ch != '\n'
                                && sc.ch != '\r'
                                && sc.ch != '\u{2028}'
                                && sc.ch != '\u{2029}'
                        })
                        .cloned()
                        .collect();
                    let seg_lines = self.break_runs_into_lines_optimal(
                        font_context,
                        &seg_chars,
                        max_width,
                        hyphens,
                        lang,
                        justify,
                    );
                    all_lines.extend(seg_lines);
                    seg_start = i;
                }
            }
            // Last segment
            let seg_chars: Vec<StyledChar> = chars[seg_start..]
                .iter()
                .filter(|sc| {
                    sc.ch != '\n' && sc.ch != '\r' && sc.ch != '\u{2028}' && sc.ch != '\u{2029}'
                })
                .cloned()
                .collect();
            if !seg_chars.is_empty() {
                let seg_lines = self.break_runs_into_lines_optimal(
                    font_context,
                    &seg_chars,
                    max_width,
                    hyphens,
                    lang,
                    justify,
                );
                all_lines.extend(seg_lines);
            }
            return all_lines;
        }

        let items = knuth_plass::build_items_styled(
            chars,
            &char_widths,
            hyphen_width,
            hyphens,
            &break_opps,
            lang,
        );
        let config = knuth_plass::Config {
            line_width: max_width,
            ..Default::default()
        };

        if let Some(solutions) = knuth_plass::find_breaks(&items, &config) {
            knuth_plass::reconstruct_run_lines(
                &solutions,
                &items,
                chars,
                &char_widths,
                max_width,
                justify,
            )
        } else {
            // Fallback to greedy
            self.break_runs_into_lines(font_context, chars, max_width, hyphens, lang)
        }
    }

    /// Truncate lines to a single line with ellipsis appended if it exceeds max_width.
    #[allow(clippy::too_many_arguments)]
    pub fn truncate_with_ellipsis(
        &self,
        font_context: &FontContext,
        mut lines: Vec<BrokenLine>,
        max_width: f64,
        font_size: f64,
        font_family: &str,
        font_weight: u32,
        font_style: FontStyle,
        letter_spacing: f64,
        word_spacing: f64,
    ) -> Vec<BrokenLine> {
        if lines.is_empty() {
            return lines;
        }

        // Take only the first line's content; if there were multiple lines,
        // reconstruct all chars from all lines into one
        let mut all_chars: Vec<char> = Vec::new();
        for line in &lines {
            all_chars.extend(&line.chars);
        }
        lines.truncate(1);

        let ellipsis = '\u{2026}'; // …
        let italic = matches!(font_style, FontStyle::Italic | FontStyle::Oblique);
        let ellipsis_width =
            font_context.char_width(ellipsis, font_family, font_weight, italic, font_size)
                + letter_spacing;

        // Measure full first line — if it fits, return as-is
        let char_widths = self.measure_chars(
            font_context,
            &all_chars.iter().collect::<String>(),
            font_size,
            font_family,
            font_weight,
            font_style,
            letter_spacing,
            word_spacing,
        );

        let total_width: f64 = char_widths.iter().sum();
        if total_width <= max_width {
            // Fits — rebuild line with all chars
            let mut x = 0.0;
            let positions: Vec<f64> = char_widths
                .iter()
                .map(|w| {
                    let pos = x;
                    x += w;
                    pos
                })
                .collect();
            lines[0] = BrokenLine {
                chars: all_chars.clone(),
                text: all_chars.iter().collect(),
                char_positions: positions,
                width: total_width,
            };
            return lines;
        }

        // Truncate: remove chars from end until line + ellipsis fits
        let target_width = max_width - ellipsis_width;
        let mut width = 0.0;
        let mut keep = 0;
        for (i, &cw) in char_widths.iter().enumerate() {
            if width + cw > target_width {
                break;
            }
            width += cw;
            keep = i + 1;
        }

        // Trim trailing whitespace
        while keep > 0 && all_chars[keep - 1].is_whitespace() {
            keep -= 1;
        }

        let mut truncated_chars: Vec<char> = all_chars[..keep].to_vec();
        truncated_chars.push(ellipsis);

        let mut x = 0.0;
        let mut positions: Vec<f64> = char_widths[..keep]
            .iter()
            .map(|w| {
                let pos = x;
                x += w;
                pos
            })
            .collect();
        positions.push(x);
        let final_width = x + ellipsis_width;

        lines[0] = BrokenLine {
            text: truncated_chars.iter().collect(),
            chars: truncated_chars,
            char_positions: positions,
            width: final_width,
        };
        lines
    }

    /// Truncate lines to a single line by clipping (no indicator appended).
    #[allow(clippy::too_many_arguments)]
    pub fn truncate_clip(
        &self,
        font_context: &FontContext,
        mut lines: Vec<BrokenLine>,
        max_width: f64,
        font_size: f64,
        font_family: &str,
        font_weight: u32,
        font_style: FontStyle,
        letter_spacing: f64,
        word_spacing: f64,
    ) -> Vec<BrokenLine> {
        if lines.is_empty() {
            return lines;
        }

        let mut all_chars: Vec<char> = Vec::new();
        for line in &lines {
            all_chars.extend(&line.chars);
        }
        lines.truncate(1);

        let char_widths = self.measure_chars(
            font_context,
            &all_chars.iter().collect::<String>(),
            font_size,
            font_family,
            font_weight,
            font_style,
            letter_spacing,
            word_spacing,
        );

        let total_width: f64 = char_widths.iter().sum();
        if total_width <= max_width {
            let mut x = 0.0;
            let positions: Vec<f64> = char_widths
                .iter()
                .map(|w| {
                    let pos = x;
                    x += w;
                    pos
                })
                .collect();
            lines[0] = BrokenLine {
                chars: all_chars.clone(),
                text: all_chars.iter().collect(),
                char_positions: positions,
                width: total_width,
            };
            return lines;
        }

        let mut width = 0.0;
        let mut keep = 0;
        for (i, &cw) in char_widths.iter().enumerate() {
            if width + cw > max_width {
                break;
            }
            width += cw;
            keep = i + 1;
        }

        let truncated_chars: Vec<char> = all_chars[..keep].to_vec();
        let mut x = 0.0;
        let positions: Vec<f64> = char_widths[..keep]
            .iter()
            .map(|w| {
                let pos = x;
                x += w;
                pos
            })
            .collect();

        lines[0] = BrokenLine {
            text: truncated_chars.iter().collect(),
            chars: truncated_chars,
            char_positions: positions,
            width,
        };
        lines
    }

    /// Truncate multi-style run lines to a single line with ellipsis.
    pub fn truncate_runs_with_ellipsis(
        &self,
        font_context: &FontContext,
        mut lines: Vec<RunBrokenLine>,
        max_width: f64,
    ) -> Vec<RunBrokenLine> {
        if lines.is_empty() {
            return lines;
        }

        // Gather all chars from all lines
        let mut all_chars: Vec<StyledChar> = Vec::new();
        for line in &lines {
            all_chars.extend(line.chars.iter().cloned());
        }
        lines.truncate(1);

        let char_widths = self.measure_styled_chars(font_context, &all_chars);
        let total_width: f64 = char_widths.iter().sum();

        if total_width <= max_width {
            let mut x = 0.0;
            let positions: Vec<f64> = char_widths
                .iter()
                .map(|w| {
                    let pos = x;
                    x += w;
                    pos
                })
                .collect();
            lines[0] = RunBrokenLine {
                chars: all_chars,
                char_positions: positions,
                width: total_width,
            };
            return lines;
        }

        // Measure ellipsis using the last char's style
        let last_style = all_chars.last().unwrap();
        let italic = matches!(
            last_style.font_style,
            FontStyle::Italic | FontStyle::Oblique
        );
        let ellipsis_width = font_context.char_width(
            '\u{2026}',
            &last_style.font_family,
            last_style.font_weight,
            italic,
            last_style.font_size,
        ) + last_style.letter_spacing;

        let target_width = max_width - ellipsis_width;
        let mut width = 0.0;
        let mut keep = 0;
        for (i, &cw) in char_widths.iter().enumerate() {
            if width + cw > target_width {
                break;
            }
            width += cw;
            keep = i + 1;
        }

        while keep > 0 && all_chars[keep - 1].ch.is_whitespace() {
            keep -= 1;
        }

        let mut truncated: Vec<StyledChar> = all_chars[..keep].to_vec();
        // Add ellipsis char with the style of the last kept char (or last_style)
        let ellipsis_style = if keep > 0 {
            all_chars[keep - 1].clone()
        } else {
            last_style.clone()
        };
        truncated.push(StyledChar {
            ch: '\u{2026}',
            ..ellipsis_style
        });

        let mut x = 0.0;
        let mut positions: Vec<f64> = char_widths[..keep]
            .iter()
            .map(|w| {
                let pos = x;
                x += w;
                pos
            })
            .collect();
        positions.push(x);

        lines[0] = RunBrokenLine {
            chars: truncated,
            char_positions: positions,
            width: x + ellipsis_width,
        };
        lines
    }

    /// Truncate multi-style run lines to a single line by clipping.
    pub fn truncate_runs_clip(
        &self,
        font_context: &FontContext,
        mut lines: Vec<RunBrokenLine>,
        max_width: f64,
    ) -> Vec<RunBrokenLine> {
        if lines.is_empty() {
            return lines;
        }

        let mut all_chars: Vec<StyledChar> = Vec::new();
        for line in &lines {
            all_chars.extend(line.chars.iter().cloned());
        }
        lines.truncate(1);

        let char_widths = self.measure_styled_chars(font_context, &all_chars);
        let total_width: f64 = char_widths.iter().sum();

        if total_width <= max_width {
            let mut x = 0.0;
            let positions: Vec<f64> = char_widths
                .iter()
                .map(|w| {
                    let pos = x;
                    x += w;
                    pos
                })
                .collect();
            lines[0] = RunBrokenLine {
                chars: all_chars,
                char_positions: positions,
                width: total_width,
            };
            return lines;
        }

        let mut width = 0.0;
        let mut keep = 0;
        for (i, &cw) in char_widths.iter().enumerate() {
            if width + cw > max_width {
                break;
            }
            width += cw;
            keep = i + 1;
        }

        let truncated: Vec<StyledChar> = all_chars[..keep].to_vec();
        let mut x = 0.0;
        let positions: Vec<f64> = char_widths[..keep]
            .iter()
            .map(|w| {
                let pos = x;
                x += w;
                pos
            })
            .collect();

        lines[0] = RunBrokenLine {
            chars: truncated,
            char_positions: positions,
            width,
        };
        lines
    }
}

#[cfg(test)]
mod tests {
    use super::*;

    fn ctx() -> FontContext {
        FontContext::new()
    }

    #[test]
    fn test_single_line() {
        let tl = TextLayout::new();
        let fc = ctx();
        let lines = tl.break_into_lines(
            &fc,
            "Hello",
            200.0,
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
            Hyphens::Manual,
            None,
        );
        assert_eq!(lines.len(), 1);
        assert_eq!(lines[0].text, "Hello");
    }

    #[test]
    fn test_line_break_at_space() {
        let tl = TextLayout::new();
        let fc = ctx();
        let lines = tl.break_into_lines(
            &fc,
            "Hello World",
            40.0,
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
            Hyphens::Manual,
            None,
        );
        assert!(lines.len() >= 2);
    }

    #[test]
    fn test_explicit_newline() {
        let tl = TextLayout::new();
        let fc = ctx();
        let lines = tl.break_into_lines(
            &fc,
            "Hello\nWorld",
            200.0,
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
            Hyphens::Manual,
            None,
        );
        assert_eq!(lines.len(), 2);
        assert_eq!(lines[0].text, "Hello");
        assert_eq!(lines[1].text, "World");
    }

    #[test]
    fn test_empty_string() {
        let tl = TextLayout::new();
        let fc = ctx();
        let lines = tl.break_into_lines(
            &fc,
            "",
            200.0,
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
            Hyphens::Manual,
            None,
        );
        assert_eq!(lines.len(), 1);
        assert_eq!(lines[0].width, 0.0);
    }

    #[test]
    fn test_bold_text_wider() {
        let tl = TextLayout::new();
        let fc = ctx();
        let regular = tl.measure_width(
            &fc,
            "ABCDEFG",
            32.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
        );
        let bold = tl.measure_width(
            &fc,
            "ABCDEFG",
            32.0,
            "Helvetica",
            700,
            FontStyle::Normal,
            0.0,
            0.0,
        );
        assert!(
            bold > regular,
            "Bold text should be wider: bold={bold}, regular={regular}"
        );
    }

    #[test]
    fn test_hyphenation_auto_breaks_long_word() {
        let tl = TextLayout::new();
        let fc = ctx();
        // "extraordinary" is long enough to need hyphenation in a narrow column
        let lines = tl.break_into_lines(
            &fc,
            "extraordinary",
            50.0, // very narrow
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
            Hyphens::Auto,
            None,
        );
        // Should break into multiple lines with hyphens
        assert!(
            lines.len() >= 2,
            "Auto hyphenation should break 'extraordinary' into multiple lines, got {}",
            lines.len()
        );
        // First line should end with a hyphen
        assert!(
            lines[0].text.ends_with('-'),
            "First line should end with hyphen, got: '{}'",
            lines[0].text
        );
    }

    #[test]
    fn test_hyphenation_none_forces_break() {
        let tl = TextLayout::new();
        let fc = ctx();
        let lines = tl.break_into_lines(
            &fc,
            "extraordinary",
            50.0,
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
            Hyphens::None,
            None,
        );
        // Should still break (force break), but NO hyphens inserted
        assert!(lines.len() >= 2);
        // No line should end with '-' from hyphenation
        assert!(
            !lines[0].text.ends_with('-'),
            "hyphens:none should not insert hyphens, got: '{}'",
            lines[0].text
        );
    }

    #[test]
    fn test_hyphenation_manual_uses_soft_hyphens() {
        let tl = TextLayout::new();
        let fc = ctx();
        // "extra\u{00AD}ordinary" — soft hyphen between "extra" and "ordinary"
        let lines = tl.break_into_lines(
            &fc,
            "extra\u{00AD}ordinary",
            40.0, // narrow enough to trigger break
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
            Hyphens::Manual,
            None,
        );
        assert!(
            lines.len() >= 2,
            "Should break at soft hyphen, got {} lines",
            lines.len()
        );
        // First line should end with visible hyphen
        assert!(
            lines[0].text.ends_with('-'),
            "Should render visible hyphen at soft-hyphen break, got: '{}'",
            lines[0].text
        );
        // The soft hyphen itself should not appear in output
        for line in &lines {
            assert!(
                !line.text.contains('\u{00AD}'),
                "Soft hyphens should be filtered from output"
            );
        }
    }

    #[test]
    fn test_hyphenation_prefers_space_over_hyphen() {
        let tl = TextLayout::new();
        let fc = ctx();
        // "Hello extraordinary" — should break at space first
        let lines = tl.break_into_lines(
            &fc,
            "Hello extraordinary",
            60.0,
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
            Hyphens::Auto,
            None,
        );
        assert!(lines.len() >= 2);
        // First line should break at the space, not hyphenate "Hello"
        assert!(
            lines[0].text.starts_with("Hello"),
            "Should break at space first, got: '{}'",
            lines[0].text
        );
    }

    #[test]
    fn test_min_content_width_with_hyphenation() {
        let tl = TextLayout::new();
        let fc = ctx();
        let auto_width = tl.measure_widest_word(
            &fc,
            "extraordinary",
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
            Hyphens::Auto,
            None,
        );
        let manual_width = tl.measure_widest_word(
            &fc,
            "extraordinary",
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
            Hyphens::Manual,
            None,
        );
        assert!(
            auto_width < manual_width,
            "Auto hyphenation min-content ({auto_width}) should be less than manual ({manual_width})"
        );
    }

    #[test]
    fn test_cjk_break_opportunities() {
        // UAX#14 should identify break opportunities between CJK chars
        let opps = compute_break_opportunities("\u{4F60}\u{597D}\u{4E16}\u{754C}"); // 你好世界
                                                                                    // Between CJK ideographs, UAX#14 should allow breaks
        let allowed_count = opps
            .iter()
            .filter(|o| matches!(o, Some(BreakOpportunity::Allowed)))
            .count();
        assert!(
            allowed_count >= 2,
            "Should have at least 2 break opportunities between 4 CJK chars, got {}",
            allowed_count
        );
    }

    #[test]
    fn test_hyphenation_german() {
        let tl = TextLayout::new();
        let fc = ctx();
        // German compound word — should hyphenate with lang "de"
        let lines = tl.break_into_lines(
            &fc,
            "Donaudampfschifffahrt",
            60.0,
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
            Hyphens::Auto,
            Some("de"),
        );
        assert!(
            lines.len() >= 2,
            "German word should hyphenate with lang='de', got {} lines",
            lines.len()
        );
        assert!(
            lines[0].text.ends_with('-'),
            "First line should end with hyphen, got: '{}'",
            lines[0].text
        );
    }

    #[test]
    fn test_hyphenation_unsupported_lang() {
        // Unknown lang disables algorithmic hyphenation
        let lang = resolve_hypher_lang(Some("xx-unknown"));
        assert!(lang.is_none(), "Unsupported language should return None");
    }

    #[test]
    fn test_resolve_hypher_lang_mapping() {
        assert!(matches!(
            resolve_hypher_lang(None),
            Some(hypher::Lang::English)
        ));
        assert!(matches!(
            resolve_hypher_lang(Some("en")),
            Some(hypher::Lang::English)
        ));
        assert!(matches!(
            resolve_hypher_lang(Some("en-US")),
            Some(hypher::Lang::English)
        ));
        assert!(matches!(
            resolve_hypher_lang(Some("de")),
            Some(hypher::Lang::German)
        ));
        assert!(matches!(
            resolve_hypher_lang(Some("fr")),
            Some(hypher::Lang::French)
        ));
        assert!(matches!(
            resolve_hypher_lang(Some("es")),
            Some(hypher::Lang::Spanish)
        ));
        assert!(matches!(
            resolve_hypher_lang(Some("nb")),
            Some(hypher::Lang::Norwegian)
        ));
        assert!(matches!(
            resolve_hypher_lang(Some("nn")),
            Some(hypher::Lang::Norwegian)
        ));
        assert!(resolve_hypher_lang(Some("zz")).is_none());
    }

    #[test]
    fn test_knuth_plass_fallback_to_greedy() {
        let tl = TextLayout::new();
        let fc = ctx();
        // Very narrow width — KP may fail, should fall back to greedy
        let lines = tl.break_into_lines_optimal(
            &fc,
            "Hello World",
            1.0, // impossibly narrow
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
            Hyphens::Manual,
            None,
            false,
        );
        assert!(
            !lines.is_empty(),
            "Should still produce lines via greedy fallback"
        );
    }

    #[test]
    fn test_min_content_width_without_hyphenation() {
        let tl = TextLayout::new();
        let fc = ctx();
        let manual_width = tl.measure_widest_word(
            &fc,
            "extraordinary",
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
            Hyphens::Manual,
            None,
        );
        let full_width = tl.measure_width(
            &fc,
            "extraordinary",
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
        );
        assert!(
            (manual_width - full_width).abs() < 0.01,
            "Manual min-content ({manual_width}) should equal full word width ({full_width})"
        );
    }

    #[test]
    fn test_truncate_ellipsis_narrow() {
        let tl = TextLayout::new();
        let fc = ctx();
        let lines = tl.break_into_lines(
            &fc,
            "Hello World this is a long text",
            200.0,
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
            Hyphens::Manual,
            None,
        );
        // Multiple lines exist
        assert!(!lines.is_empty());

        let truncated = tl.truncate_with_ellipsis(
            &fc,
            lines,
            60.0, // Very narrow
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
        );
        assert_eq!(truncated.len(), 1, "Should be single line");
        assert!(
            truncated[0].text.ends_with('\u{2026}'),
            "Should end with ellipsis: {:?}",
            truncated[0].text
        );
        assert!(
            truncated[0].width <= 60.0 + 0.1,
            "Should fit within max_width"
        );
    }

    #[test]
    fn test_truncate_ellipsis_fits() {
        let tl = TextLayout::new();
        let fc = ctx();
        let lines = tl.break_into_lines(
            &fc,
            "Hi",
            200.0,
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
            Hyphens::Manual,
            None,
        );
        let truncated = tl.truncate_with_ellipsis(
            &fc,
            lines,
            200.0,
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
        );
        assert_eq!(truncated.len(), 1);
        assert_eq!(
            truncated[0].text, "Hi",
            "Short text should not get ellipsis"
        );
    }

    #[test]
    fn test_truncate_clip() {
        let tl = TextLayout::new();
        let fc = ctx();
        let lines = tl.break_into_lines(
            &fc,
            "Hello World this is a long text",
            200.0,
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
            Hyphens::Manual,
            None,
        );
        let truncated = tl.truncate_clip(
            &fc,
            lines,
            60.0,
            12.0,
            "Helvetica",
            400,
            FontStyle::Normal,
            0.0,
            0.0,
        );
        assert_eq!(truncated.len(), 1, "Should be single line");
        assert!(
            !truncated[0].text.contains('\u{2026}'),
            "Clip should not have ellipsis"
        );
        assert!(
            truncated[0].width <= 60.0 + 0.1,
            "Should fit within max_width"
        );
    }

    #[test]
    fn test_greedy_vs_optimal_produce_different_breaks() {
        // This test verifies that greedy and optimal algorithms actually diverge.
        // We use the proof paragraph at various widths/font sizes.
        let tl = TextLayout::new();
        let fc = ctx();
        let text = "The extraordinary effectiveness of mathematics in the natural sciences is something bordering on the mysterious. There is no rational explanation for it. It is not at all natural that laws of nature exist, much less that man is able to discover them. The miracle of the appropriateness of the language of mathematics for the formulation of the laws of physics is a wonderful gift which we neither understand nor deserve.";

        let mut found_divergence = false;
        let mut divergence_info = String::new();

        // Test at 10pt font (matching the proof template: 200pt col - 16pt padding = 184pt)
        for width in [
            100.0, 120.0, 140.0, 150.0, 160.0, 170.0, 180.0, 184.0, 200.0,
        ] {
            let greedy = tl.break_into_lines(
                &fc,
                text,
                width,
                10.0,
                "Helvetica",
                400,
                FontStyle::Normal,
                0.0,
                0.0,
                Hyphens::Auto,
                Some("en"),
            );
            let optimal = tl.break_into_lines_optimal(
                &fc,
                text,
                width,
                10.0,
                "Helvetica",
                400,
                FontStyle::Normal,
                0.0,
                0.0,
                Hyphens::Auto,
                Some("en"),
                false,
            );

            let greedy_texts: Vec<&str> = greedy.iter().map(|l| l.text.as_str()).collect();
            let optimal_texts: Vec<&str> = optimal.iter().map(|l| l.text.as_str()).collect();

            if greedy_texts != optimal_texts {
                found_divergence = true;
                divergence_info = format!(
                    "font_size=10, width={}: greedy={} lines, optimal={} lines\nGreedy:\n{}\nOptimal:\n{}",
                    width, greedy.len(), optimal.len(),
                    greedy_texts.iter().enumerate().map(|(i, t)| format!("  {}: {:?}", i, t)).collect::<Vec<_>>().join("\n"),
                    optimal_texts.iter().enumerate().map(|(i, t)| format!("  {}: {:?}", i, t)).collect::<Vec<_>>().join("\n"),
                );
                break;
            }
        }

        assert!(
            found_divergence,
            "Greedy and optimal should produce different line breaks at some width with 10pt font."
        );
        eprintln!("Found divergence: {}", divergence_info);
    }

    #[test]
    fn test_line_widths_do_not_exceed_available_width() {
        let tl = TextLayout::new();
        let fc = ctx();
        let french = "Le chiffre d'affaires consolide a atteint douze virgule un millions de dollars, soit une augmentation de vingt-trois pour cent par rapport a l'exercice precedent. L'expansion dans trois nouveaux marches a contribue a une croissance trimestrielle de trente et un pour cent des nouvelles acquisitions de clients.";
        let german = "Das vierte Quartal verzeichnete ein starkes Umsatzwachstum in allen Regionen. Die Kundenbindungsrate blieb mit vierundneunzig Prozent auf einem hervorragenden Niveau, was die kontinuierlichen Investitionen in Produktqualitat und Kundenbetreuung widerspiegelt.";

        let max_width = 229.0;
        let font_size = 8.0;

        for (label, text, lang) in [("French", french, "fr"), ("German", german, "de")] {
            // Test both greedy and optimal
            let greedy = tl.break_into_lines(
                &fc,
                text,
                max_width,
                font_size,
                "Helvetica",
                400,
                FontStyle::Normal,
                0.0,
                0.0,
                Hyphens::Auto,
                Some(lang),
            );
            let optimal = tl.break_into_lines_optimal(
                &fc,
                text,
                max_width,
                font_size,
                "Helvetica",
                400,
                FontStyle::Normal,
                0.0,
                0.0,
                Hyphens::Auto,
                Some(lang),
                true,
            );

            for (algo, lines) in [("greedy", &greedy), ("optimal", &optimal)] {
                for (i, line) in lines.iter().enumerate() {
                    // Check line.width
                    assert!(
                        line.width <= max_width + 0.01,
                        "{} {} line {} width exceeds: {:.4} > {:.4} (text: {:?})",
                        label,
                        algo,
                        i,
                        line.width,
                        max_width,
                        line.text,
                    );

                    // Compute rendered_width same way as layout code
                    if !line.chars.is_empty() {
                        let last_idx = line.chars.len() - 1;
                        let last_pos = line.char_positions.get(last_idx).copied().unwrap_or(0.0);
                        let last_char = line.chars[last_idx];
                        let last_advance =
                            fc.char_width(last_char, "Helvetica", 400, false, font_size);
                        let rendered_width = (last_pos + last_advance).max(line.width * 0.5);
                        eprintln!(
                            "{} {} line {}: width={:.4}, rendered={:.4}, max={:.4}, last_char={:?}, text={:?}",
                            label, algo, i, line.width, rendered_width, max_width, last_char, line.text,
                        );
                    }
                }
            }
        }
    }
}