xfa_layout_engine/

text.rs

//! # Text Measurement and Wrapping
//!
//! Provides text wrapping for the XFA layout engine.
//! Key functions:
//! - `wrap_text()` — wraps text to fit within a maximum width
//! - `measure_text()` — measures text dimensions without wrapping
//!
//! ## Font Metrics
//!
//! Text measurement uses `FontMetrics.resolved_widths` when available
//! (populated from PDF /Widths arrays). Falls back to glyph advances
//! from the font file via ttf_parser.
//!
//! ## Coordinate System
//!
//! All measurements are in PDF points (1pt = 1/72 inch).
//!
//! XFA Spec 3.3 §8.1 — Text Placement in Growable Containers (p277-279):
//! - Growable width: text records interpreted as lines, width = longest line.
//! - Growable height: container increases height to accommodate text.
//! - Text split between lines only (§8.7 p291), NOT within a line.
//! - Orphan/widow controls may restrict split points.
//!   Not implemented yet because the layout engine does not currently track
//!   widow/orphan state across container and page splits.
//! - Text within rotated containers cannot be split.
//!   Not checked yet because rotation metadata is not propagated into this
//!   measurement layer.

use crate::types::{Size, TextAlign};

/// Font properties for text measurement.
#[derive(Debug, Clone)]
pub struct FontMetrics {
    /// Font size in points.
    pub size: f64,
    /// Line height as a multiplier of font size (typically 1.2).
    pub line_height: f64,
    /// Average character width as a fraction of font size (fallback).
    pub avg_char_width: f64,
    /// Horizontal text alignment (from XFA `<para hAlign>`).
    pub text_align: TextAlign,
    /// Font family for per-character width lookup.
    pub typeface: FontFamily,
    /// Resolved glyph widths from the actual PDF font (units per `resolved_upem`).
    pub resolved_widths: Option<Vec<u16>>,
    /// Units-per-em of the resolved font (typically 1000 or 2048).
    pub resolved_upem: Option<u16>,
    /// Typographic ascender of the resolved font (font units).
    pub resolved_ascender: Option<i16>,
    /// Typographic descender of the resolved font (font units, typically negative).
    pub resolved_descender: Option<i16>,
}

/// Font family classification for width table selection.
#[derive(Debug, Clone, Copy, PartialEq, Eq, Default)]
pub enum FontFamily {
    /// Times New Roman, Times, Georgia, etc.
    Serif,
    /// Arial, Helvetica, Verdana, Myriad Pro, etc.
    #[default]
    SansSerif,
    /// Courier New, Courier, Consolas, etc.
    Monospace,
}

impl FontFamily {
    /// XFA Spec 3.3 §17 (p716) — Map genericFamily attribute to FontFamily.
    pub fn from_generic_family(gf: &str) -> Self {
        match gf {
            "serif" | "decorative" | "cursive" => FontFamily::Serif,
            "monospaced" => FontFamily::Monospace,
            // sansSerif, fantasy, and unknown values default to SansSerif
            _ => FontFamily::SansSerif,
        }
    }

    /// Classify a typeface name into a font family.
    pub fn from_typeface(name: &str) -> Self {
        let lower = name.to_ascii_lowercase();
        if lower.contains("courier") || lower.contains("consolas") || lower.contains("mono") {
            FontFamily::Monospace
        } else if lower.contains("times")
            || lower.contains("georgia")
            || lower.contains("garamond")
            || lower.contains("palatino")
            || lower.contains("cambria")
            || lower.contains("book antiqua")
            || lower.contains("century")
            || lower.contains("serif")
        {
            FontFamily::Serif
        } else {
            // Arial, Helvetica, Verdana, Myriad Pro, Calibri, Tahoma, etc.
            FontFamily::SansSerif
        }
    }
}

impl Default for FontMetrics {
    fn default() -> Self {
        Self {
            // Match the engine-wide default body text size used by existing XFA forms.
            size: 10.0,
            // Adobe/XFA uses a 1.2x fallback line-height when font metrics are absent.
            line_height: 1.2,
            // Coarse fallback used only when neither resolved widths nor AFM tables apply.
            avg_char_width: 0.50,
            text_align: TextAlign::Left,
            typeface: FontFamily::SansSerif,
            resolved_widths: None,
            resolved_upem: None,
            resolved_ascender: None,
            resolved_descender: None,
        }
    }
}

impl FontMetrics {
    pub fn new(size: f64) -> Self {
        Self {
            size,
            ..Default::default()
        }
    }

    /// Uses resolved ascender/descender when available, else `size * 1.2`.
    ///
    /// XFA Spec 3.3 §28.1 (p1228) — Adobe Non-conformance: Font metrics.
    /// Adobe's AXTE text engine ignores font-supplied line gap and uses 20% of
    /// font height. Our approach matches: (asc−desc)/upem gives the em-square
    /// height (no line gap added), and the fallback is size × 1.2 (20% extra).
    pub fn line_height_pt(&self) -> f64 {
        if let (Some(asc), Some(desc), Some(upem)) = (
            self.resolved_ascender,
            self.resolved_descender,
            self.resolved_upem,
        ) {
            if upem > 0 {
                return ((asc as f64) - (desc as f64)) / (upem as f64) * self.size;
            }
        }
        self.size * self.line_height
    }

    /// Measure the width of `text` in points.
    ///
    /// When `resolved_widths` is present, widths are interpreted as per-1000
    /// text-space units indexed directly by character code in the 0..255 range.
    /// Upstream code must therefore apply any PDF `/FirstChar` offset as padding
    /// before storing widths here, so that code 65 (`'A'`) is read from
    /// `resolved_widths[65]`.
    ///
    /// Characters outside the available resolved-width table fall back to the
    /// width of ASCII space. This keeps wrapping stable for partially populated
    /// tables and avoids treating unknown characters as zero-width.
    ///
    /// When `resolved_widths` is absent, measurement falls back to AFM tables for
    /// the generic serif/sans/monospace families. Non-ASCII characters in that
    /// path use the width of `'n'` as a conservative average for one visible glyph.
    ///
    /// The function iterates over Unicode scalar values rather than UTF-8 bytes,
    /// so multibyte characters contribute a single glyph width.
    pub fn measure_width(&self, text: &str) -> f64 {
        if let (Some(ref widths), Some(_upem)) = (&self.resolved_widths, self.resolved_upem) {
            // Use space as the fallback width because it is usually present in
            // PDF width tables and is safer for wrapping than a zero-width default.
            let space_w = widths.get(b' ' as usize).copied().unwrap_or(0) as f64;
            let mut w = 0.0;
            for ch in text.chars() {
                let code = ch as u32;
                let cw = if (code as usize) < widths.len() {
                    widths[code as usize] as f64
                } else {
                    space_w
                };
                w += cw / 1000.0 * self.size;
            }
            return w;
        }
        let table = match self.typeface {
            FontFamily::Serif => &TIMES_WIDTHS,
            FontFamily::SansSerif => &HELVETICA_WIDTHS,
            FontFamily::Monospace => &COURIER_WIDTHS,
        };
        // The AFM fallback uses 'n' as the representative width for unsupported
        // characters because it is a common mid-width Latin glyph.
        let default_w = table[b'n' as usize] as f64;
        let mut width = 0.0;
        for ch in text.chars() {
            let code = ch as u32;
            let char_width = if code < 128 {
                table[code as usize] as f64
            } else {
                default_w
            };
            width += char_width / 1000.0 * self.size;
        }
        width
    }
}

// ---------------------------------------------------------------------------
// Per-character width tables (Adobe Font Metrics, units per 1000 em)
// ---------------------------------------------------------------------------

/// Times-Roman character widths (ASCII 0-127).
/// Source: Adobe AFM for Times-Roman.
#[rustfmt::skip]
static TIMES_WIDTHS: [u16; 128] = [
    // 0x00-0x0F: control characters → use space width (250)
    250,250,250,250,250,250,250,250,250,250,250,250,250,250,250,250,
    // 0x10-0x1F: control characters
    250,250,250,250,250,250,250,250,250,250,250,250,250,250,250,250,
    // 0x20-0x2F: SP ! " # $ % & ' ( ) * + , - . /
    250,333,408,500,500,833,778,180,333,333,500,564,250,333,250,278,
    // 0x30-0x3F: 0 1 2 3 4 5 6 7 8 9 : ; < = > ?
    500,500,500,500,500,500,500,500,500,500,278,278,564,564,564,444,
    // 0x40-0x4F: @ A B C D E F G H I J K L M N O
    921,722,667,667,722,611,556,722,722,333,389,722,611,889,722,722,
    // 0x50-0x5F: P Q R S T U V W X Y Z [ \ ] ^ _
    556,722,667,556,611,722,722,944,722,722,611,333,278,333,469,500,
    // 0x60-0x6F: ` a b c d e f g h i j k l m n o
    333,444,500,444,500,444,333,500,500,278,278,500,278,778,500,500,
    // 0x70-0x7F: p q r s t u v w x y z { | } ~ DEL
    500,500,333,389,278,500,500,722,500,500,444,480,200,480,541,250,
];

/// Helvetica character widths (ASCII 0-127).
/// Source: Adobe AFM for Helvetica.
#[rustfmt::skip]
static HELVETICA_WIDTHS: [u16; 128] = [
    // 0x00-0x0F: control characters → use space width (278)
    278,278,278,278,278,278,278,278,278,278,278,278,278,278,278,278,
    // 0x10-0x1F: control characters
    278,278,278,278,278,278,278,278,278,278,278,278,278,278,278,278,
    // 0x20-0x2F: SP ! " # $ % & ' ( ) * + , - . /
    278,278,355,556,556,889,667,191,333,333,389,584,278,333,278,278,
    // 0x30-0x3F: 0 1 2 3 4 5 6 7 8 9 : ; < = > ?
    556,556,556,556,556,556,556,556,556,556,278,278,584,584,584,556,
    // 0x40-0x4F: @ A B C D E F G H I J K L M N O
    1015,667,667,722,722,611,556,778,722,278,500,667,556,833,722,778,
    // 0x50-0x5F: P Q R S T U V W X Y Z [ \ ] ^ _
    667,778,722,667,611,722,667,944,667,667,611,278,278,278,469,556,
    // 0x60-0x6F: ` a b c d e f g h i j k l m n o
    333,556,556,500,556,556,278,556,556,222,222,500,222,833,556,556,
    // 0x70-0x7F: p q r s t u v w x y z { | } ~ DEL
    556,556,333,500,278,556,500,722,500,500,500,334,260,334,584,278,
];

/// Courier character widths (ASCII 0-127).
/// All characters are 600 units wide (monospace).
#[rustfmt::skip]
static COURIER_WIDTHS: [u16; 128] = [
    600,600,600,600,600,600,600,600,600,600,600,600,600,600,600,600,
    600,600,600,600,600,600,600,600,600,600,600,600,600,600,600,600,
    600,600,600,600,600,600,600,600,600,600,600,600,600,600,600,600,
    600,600,600,600,600,600,600,600,600,600,600,600,600,600,600,600,
    600,600,600,600,600,600,600,600,600,600,600,600,600,600,600,600,
    600,600,600,600,600,600,600,600,600,600,600,600,600,600,600,600,
    600,600,600,600,600,600,600,600,600,600,600,600,600,600,600,600,
    600,600,600,600,600,600,600,600,600,600,600,600,600,600,600,600,
];

/// Text wrapping and measurement result.
#[derive(Debug, Clone)]
pub struct TextLayout {
    /// The wrapped lines of text.
    pub lines: Vec<String>,
    /// Per-line flag: `true` when the line is the first line of a paragraph.
    pub first_line_of_para: Vec<bool>,
    /// Total size of the text block.
    pub size: Size,
}

/// Wrap text to fit within a given width, and compute the resulting size.
///
/// Parameters:
/// - `text`: the source text. Explicit `\n` characters always start a new paragraph.
/// - `max_width`: maximum available width in points for non-indented lines.
/// - `font`: font metrics used to measure words and spaces.
/// - `text_indent`: first-line indent in points, subtracted from the first line of
///   each paragraph only.
/// - `line_height_override`: optional baseline-to-baseline distance in points.
///   When `None`, `font.line_height_pt()` is used.
///
/// Returns:
/// - `TextLayout.lines`: wrapped lines in visual order.
/// - `TextLayout.first_line_of_para`: flags indicating whether each output line is
///   the first line of a paragraph.
/// - `TextLayout.size`: width of the longest emitted line and total block height.
///
/// Edge cases:
/// - Empty input returns no lines and a zero-size block.
/// - Empty paragraphs caused by consecutive `\n` are preserved as blank lines.
/// - Whitespace inside a paragraph is normalized by `split_whitespace()`, so runs
///   of spaces do not survive wrapping.
/// - Words are never split internally; a word wider than `max_width` is placed on
///   its own line and may overflow horizontally.
/// - If `text_indent >= max_width`, the first line's available width clamps to `0`
///   so the paragraph still wraps deterministically.
pub fn wrap_text(
    text: &str,
    max_width: f64,
    font: &FontMetrics,
    text_indent: f64,
    line_height_override: Option<f64>,
) -> TextLayout {
    if text.is_empty() {
        return TextLayout {
            lines: vec![],
            first_line_of_para: vec![],
            size: Size {
                width: 0.0,
                height: 0.0,
            },
        };
    }

    let mut lines = Vec::new();
    let mut first_line_of_para = Vec::new();
    let mut max_line_width = 0.0_f64;

    for paragraph in text.split('\n') {
        if paragraph.is_empty() {
            lines.push(String::new());
            first_line_of_para.push(true);
            continue;
        }

        let words: Vec<&str> = paragraph.split_whitespace().collect();
        if words.is_empty() {
            lines.push(String::new());
            first_line_of_para.push(true);
            continue;
        }

        let mut is_first_line = true;
        let mut current_line = String::new();
        let mut current_width = 0.0;

        for word in words {
            let word_width = font.measure_width(word);
            let space_width = if current_line.is_empty() {
                0.0
            } else {
                font.measure_width(" ")
            };

            let effective_max = if is_first_line {
                (max_width - text_indent).max(0.0)
            } else {
                max_width
            };

            if current_width + space_width + word_width > effective_max && !current_line.is_empty()
            {
                // Wrap only at whitespace boundaries. The engine deliberately does
                // not hyphenate or split within a word because XFA pagination
                // rules operate on whole rendered lines.
                max_line_width = max_line_width.max(current_width);
                lines.push(current_line);
                first_line_of_para.push(is_first_line);
                is_first_line = false;
                current_line = word.to_string();
                current_width = word_width;
            } else {
                if !current_line.is_empty() {
                    current_line.push(' ');
                    current_width += space_width;
                }
                current_line.push_str(word);
                current_width += word_width;
            }
        }

        if !current_line.is_empty() {
            max_line_width = max_line_width.max(current_width);
            lines.push(current_line);
            first_line_of_para.push(is_first_line);
        }
    }

    let lh = line_height_override.unwrap_or_else(|| font.line_height_pt());
    let height = lines.len() as f64 * lh;

    TextLayout {
        lines,
        first_line_of_para,
        size: Size {
            width: max_line_width,
            height,
        },
    }
}

/// Compute the bounding box of text without wrapping.
///
/// Each `\n` starts a new measured line, but lines are never reflowed to fit a width.
/// Width is the maximum measured line width; height is `line_count * line_height_pt()`.
pub fn measure_text(text: &str, font: &FontMetrics) -> Size {
    if text.is_empty() {
        return Size {
            width: 0.0,
            height: 0.0,
        };
    }

    let lines: Vec<&str> = text.split('\n').collect();
    let max_width = lines
        .iter()
        .map(|l| font.measure_width(l))
        .fold(0.0, f64::max);
    let height = lines.len() as f64 * font.line_height_pt();

    Size {
        width: max_width,
        height,
    }
}

/// Compute valid split positions (y-offsets) for multiline text.
///
/// XFA Spec 3.3 §8.7 (p291): text may be split between lines only.
/// Returns the y-position of each line boundary (after line 1, after line 2, …).
/// A text with N lines has N−1 split points.
pub fn text_split_points(line_count: usize, line_height: f64) -> Vec<f64> {
    if line_count <= 1 {
        return Vec::new();
    }
    (1..line_count).map(|i| i as f64 * line_height).collect()
}

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

    #[test]
    fn font_metrics_defaults() {
        let f = FontMetrics::default();
        assert_eq!(f.size, 10.0);
        assert_eq!(f.line_height_pt(), 12.0); // 10 * 1.2
    }

    #[test]
    fn measure_width_times() {
        let f = FontMetrics {
            size: 10.0,
            typeface: FontFamily::Serif,
            ..Default::default()
        };
        // "Hello" in Times: H=722 e=444 l=278 l=278 o=500 = 2222/1000*10 = 22.22
        let w = f.measure_width("Hello");
        assert!((w - 22.22).abs() < 0.01, "Times Hello={w}, expected ~22.22");
    }

    #[test]
    fn measure_width_helvetica() {
        let f = FontMetrics {
            size: 10.0,
            typeface: FontFamily::SansSerif,
            ..Default::default()
        };
        // "Hello" in Helvetica: H=722 e=556 l=222 l=222 o=556 = 2278/1000*10 = 22.78
        let w = f.measure_width("Hello");
        assert!((w - 22.78).abs() < 0.01, "Helv Hello={w}, expected ~22.78");
    }

    #[test]
    fn measure_width_courier() {
        let f = FontMetrics {
            size: 10.0,
            typeface: FontFamily::Monospace,
            ..Default::default()
        };
        // "Hello" in Courier: 5 * 600/1000*10 = 30.0
        let w = f.measure_width("Hello");
        assert!((w - 30.0).abs() < 0.01, "Courier Hello={w}, expected 30.0");
    }

    #[test]
    fn times_narrower_than_old_avg() {
        // The old avg_char_width=0.5 would give: 5*10*0.5=25.0 for "Hello"
        // Times should be narrower (~22.22)
        let f = FontMetrics {
            size: 10.0,
            typeface: FontFamily::Serif,
            ..Default::default()
        };
        let w = f.measure_width("Hello");
        assert!(w < 25.0, "Times should be narrower than old 0.5 avg");
    }

    #[test]
    fn font_family_classification() {
        assert_eq!(
            FontFamily::from_typeface("Times New Roman"),
            FontFamily::Serif
        );
        assert_eq!(FontFamily::from_typeface("Arial"), FontFamily::SansSerif);
        assert_eq!(
            FontFamily::from_typeface("Courier New"),
            FontFamily::Monospace
        );
        assert_eq!(
            FontFamily::from_typeface("Myriad Pro"),
            FontFamily::SansSerif
        );
        assert_eq!(FontFamily::from_typeface("Verdana"), FontFamily::SansSerif);
        assert_eq!(FontFamily::from_typeface("Georgia"), FontFamily::Serif);
    }

    #[test]
    fn measure_text_single_line() {
        let f = FontMetrics::default();
        let s = measure_text("Hello", &f);
        assert!(s.width > 0.0);
        assert_eq!(s.height, 12.0);
    }

    #[test]
    fn measure_text_multiline() {
        let f = FontMetrics::default();
        let s = measure_text("Line 1\nLine 2\nLine 3", &f);
        assert_eq!(s.height, 36.0); // 3 lines * 12pt
    }

    #[test]
    fn wrap_text_no_wrap_needed() {
        let f = FontMetrics::default();
        let result = wrap_text("Short", 200.0, &f, 0.0, None);
        assert_eq!(result.lines.len(), 1);
        assert_eq!(result.lines[0], "Short");
    }

    #[test]
    fn wrap_text_preserves_newlines() {
        let f = FontMetrics::default();
        let result = wrap_text("Line 1\nLine 2", 200.0, &f, 0.0, None);
        assert_eq!(result.lines.len(), 2);
        assert_eq!(result.lines[0], "Line 1");
        assert_eq!(result.lines[1], "Line 2");
    }

    #[test]
    fn wrap_text_empty_string() {
        let f = FontMetrics::default();
        let result = wrap_text("", 100.0, &f, 0.0, None);
        assert_eq!(result.lines.len(), 0);
        assert_eq!(result.size.height, 0.0);
    }

    #[test]
    fn resolved_widths_measure() {
        let mut widths = vec![0u16; 256];
        widths[b'H' as usize] = 700;
        widths[b'i' as usize] = 300;
        let f = FontMetrics {
            size: 10.0,
            resolved_widths: Some(widths),
            resolved_upem: Some(1000),
            ..Default::default()
        };
        let w = f.measure_width("Hi");
        assert!((w - 10.0).abs() < 0.01, "resolved Hi={w}, expected 10.0");
    }

    #[test]
    fn resolved_line_height() {
        let f = FontMetrics {
            size: 12.0,
            resolved_ascender: Some(800),
            resolved_descender: Some(-200),
            resolved_upem: Some(1000),
            ..Default::default()
        };
        let lh = f.line_height_pt();
        assert!((lh - 12.0).abs() < 0.01, "resolved lh={lh}, expected 12.0");
    }

    #[test]
    fn wrap_text_times_given_name() {
        // Regression: "Given Name (First Name)" should fit in ~140pt at 9pt Times
        let f = FontMetrics {
            size: 9.0,
            typeface: FontFamily::Serif,
            ..Default::default()
        };
        let result = wrap_text("Given Name (First Name)", 140.0, &f, 0.0, None);
        assert_eq!(
            result.lines.len(),
            1,
            "Should fit on 1 line but got: {:?}",
            result.lines
        );
    }

    #[test]
    fn resolved_widths_upem_2048() {
        // Widths from pdf_glyph_widths() are per-1000 units regardless of upem.
        // With upem=2048, measure_width must still divide by 1000, not 2048.
        let mut widths = vec![0u16; 256];
        widths[b'A' as usize] = 600; // per-1000 unit width
        let f = FontMetrics {
            size: 10.0,
            resolved_widths: Some(widths),
            resolved_upem: Some(2048),
            ..Default::default()
        };
        let w = f.measure_width("A");
        // Correct: 600/1000*10 = 6.0
        // Old bug: 600/2048*10 = 2.93 (too narrow)
        assert!(
            (w - 6.0).abs() < 0.01,
            "upem=2048: A width={w}, expected 6.0"
        );
    }

    #[test]
    fn multibyte_utf8_single_char_width() {
        // 'é' (U+00E9) is 2 bytes in UTF-8 but must be measured as ONE character.
        let mut widths = vec![0u16; 256];
        widths[0xE9] = 500; // width of é
        let f = FontMetrics {
            size: 10.0,
            resolved_widths: Some(widths),
            resolved_upem: Some(1000),
            ..Default::default()
        };
        let w = f.measure_width("\u{00E9}");
        // Correct: 500/1000*10 = 5.0 (one char)
        // Old bug: two bytes 0xC3+0xA9 → widths[0xC3]+widths[0xA9] = double lookup
        assert!(
            (w - 5.0).abs() < 0.01,
            "é width={w}, expected 5.0 (one glyph, not two bytes)"
        );
    }

    #[test]
    fn afm_fallback_multibyte_char() {
        // Non-ASCII char in AFM fallback should count as one 'n'-width, not two.
        let f = FontMetrics {
            size: 10.0,
            typeface: FontFamily::SansSerif,
            ..Default::default()
        };
        let w_n = f.measure_width("n");
        let w_e_accent = f.measure_width("\u{00E9}");
        // AFM fallback maps non-ASCII to 'n' width: one char = one 'n' width.
        assert!(
            (w_e_accent - w_n).abs() < 0.01,
            "AFM fallback: é={w_e_accent} should equal n={w_n}"
        );
    }

    #[test]
    fn text_split_points_multi() {
        let pts = text_split_points(4, 12.0);
        assert_eq!(pts.len(), 3);
        assert!((pts[0] - 12.0).abs() < 0.01);
        assert!((pts[1] - 24.0).abs() < 0.01);
        assert!((pts[2] - 36.0).abs() < 0.01);
    }

    #[test]
    fn text_split_points_single_line() {
        let pts = text_split_points(1, 12.0);
        assert!(pts.is_empty());
    }
}