opentype.rs

//! OpenType のファイルからテーブルを取り出す
//!
//! 読むのは数式を組むのに要るものだけである。head、hhea、maxp、hmtx、cmap、name の 6 つと、
//! 字形の輪郭を持つ glyf か CFF、それに MATH テーブル。字形の輪郭と MATH は別のところで読む。

use crate::error::Error;
use std::collections::HashMap;


/// head から取り出すもの
#[derive(Debug, Clone, Copy, PartialEq)]
pub struct Head {
    /// 1em あたりの座標の単位
    pub units_per_em: f64,
    /// loca が 4 バイト刻みかどうか
    pub long_loca: bool,
}


fn short(at: usize) -> Error {
    Error::new("フォントのファイルが途中で終わっています", at)
}


pub fn read_u8(data: &[u8], at: usize) -> Result<u8, Error> {
    data.get(at).copied().ok_or_else(|| short(at))
}


pub fn read_u16(data: &[u8], at: usize) -> Result<u16, Error> {
    let slice = data.get(at..at + 2).ok_or_else(|| short(at))?;

    Ok(u16::from_be_bytes([slice[0], slice[1]]))
}


pub fn read_i16(data: &[u8], at: usize) -> Result<i16, Error> {
    Ok(read_u16(data, at)? as i16)
}


pub fn read_u32(data: &[u8], at: usize) -> Result<u32, Error> {
    let slice = data.get(at..at + 4).ok_or_else(|| short(at))?;

    Ok(u32::from_be_bytes([slice[0], slice[1], slice[2], slice[3]]))
}


/// テーブルの目次の位置。フォントの集合（ttcf）なら先頭の 1 つを使う
fn directory(data: &[u8]) -> Result<usize, Error> {
    if data.len() >= 4 && &data[0..4] == b"ttcf" {
        return Ok(read_u32(data, 12)? as usize);
    }

    Ok(0)
}


/// テーブルの位置と長さを引く
pub fn find(data: &[u8], tag: &str) -> Result<Option<(usize, usize)>, Error> {
    let base = directory(data)?;
    let count = read_u16(data, base + 4)? as usize;

    for index in 0..count {
        let at = base + 12 + index * 16;
        let name = data.get(at..at + 4).ok_or_else(|| short(at))?;

        if name == tag.as_bytes() {
            let start = read_u32(data, at + 8)? as usize;
            let length = read_u32(data, at + 12)? as usize;

            return Ok(Some((start, length)));
        }
    }

    Ok(None)
}


/// 要るテーブルを引く。なければ何が足りないかを言う
pub fn require(data: &[u8], tag: &str) -> Result<(usize, usize), Error> {
    match find(data, tag)? {
        Some(found) => Ok(found),
        None => Err(Error::new(format!("{tag} テーブルがありません"), 0)),
    }
}


pub fn head(data: &[u8], at: usize) -> Result<Head, Error> {
    let units_per_em = read_u16(data, at + 18)? as f64;

    if units_per_em <= 0.0 {
        return Err(Error::new("unitsPerEm が 0 です", at + 18));
    }

    Ok(Head {units_per_em, long_loca: read_i16(data, at + 50)? != 0})
}


/// 字形の数
pub fn glyph_count(data: &[u8], maxp_at: usize) -> Result<usize, Error> {
    Ok(read_u16(data, maxp_at + 4)? as usize)
}


/// 字形ごとの字送り。フォントの単位のまま返す
pub fn advances(
    data: &[u8],
    hmtx_at: usize,
    hhea_at: usize,
    count: usize,
) -> Result<Vec<f64>, Error> {
    let metric_count = read_u16(data, hhea_at + 34)? as usize;

    if metric_count == 0 {
        return Err(Error::new("hhea の numberOfHMetrics が 0 です", hhea_at + 34));
    }

    let mut found = Vec::with_capacity(count);

    for index in 0..count {
        let at = hmtx_at + index.min(metric_count - 1) * 4;
        found.push(read_u16(data, at)? as f64);
    }

    Ok(found)
}


/// cmap の中から、いちばん広く読める部分表を選ぶ
fn best_subtable(data: &[u8], at: usize) -> Result<usize, Error> {
    let count = read_u16(data, at + 2)? as usize;
    let mut best: Option<(u8, usize)> = None;

    for index in 0..count {
        let record = at + 4 + index * 8;
        let platform = read_u16(data, record)?;
        let encoding = read_u16(data, record + 2)?;
        let offset = at + read_u32(data, record + 4)? as usize;
        let format = read_u16(data, offset)?;
        let score = match (platform, encoding, format) {
            (3, 10, 12) => 5,
            (0, _, 12) => 4,
            (3, 1, 4) => 3,
            (0, _, 4) => 2,
            (_, _, 4) | (_, _, 12) => 1,
            _ => continue,
        };

        if best.is_none_or(|(current, _)| score > current) {
            best = Some((score, offset));
        }
    }

    match best {
        Some((_, offset)) => Ok(offset),
        None => Err(Error::new("読める cmap の部分表がありません", at)),
    }
}


/// 対応づけを入れすぎないための上限。字形は 65536 までしかないので、これを超えるのは壊れた表である
const MAP_LIMIT: usize = 300_000;


/// 文字から字形番号への対応
pub fn character_map(data: &[u8], at: usize) -> Result<HashMap<char, u16>, Error> {
    let offset = best_subtable(data, at)?;

    match read_u16(data, offset)? {
        4 => segment_map(data, offset),
        12 => group_map(data, offset),
        format => Err(Error::new(format!("cmap の形式 {format} は読めません"), offset)),
    }
}


/// cmap の形式 4
fn segment_map(data: &[u8], at: usize) -> Result<HashMap<char, u16>, Error> {
    let segment_count = read_u16(data, at + 6)? as usize / 2;
    let end_at = at + 14;
    let start_at = end_at + segment_count * 2 + 2;
    let delta_at = start_at + segment_count * 2;
    let range_at = delta_at + segment_count * 2;
    let mut found = HashMap::new();

    for index in 0..segment_count {
        let end = read_u16(data, end_at + index * 2)? as u32;
        let start = read_u16(data, start_at + index * 2)? as u32;
        let delta = read_i16(data, delta_at + index * 2)? as i32;
        let range = read_u16(data, range_at + index * 2)? as usize;

        if start > end || start == 0xffff {
            continue;
        }

        for code in start..=end {
            if found.len() >= MAP_LIMIT {
                return Err(Error::new("cmap の対応づけが多すぎます", at));
            }

            let glyph = if range == 0 {
                ((code as i32 + delta) & 0xffff) as u16
            } else {
                let entry = range_at + index * 2 + range + (code - start) as usize * 2;
                let raw = read_u16(data, entry)?;

                if raw == 0 {
                    continue;
                }

                ((raw as i32 + delta) & 0xffff) as u16
            };

            if glyph == 0 {
                continue;
            }

            if let Some(character) = char::from_u32(code) {
                found.insert(character, glyph);
            }
        }
    }

    Ok(found)
}


/// cmap の形式 12
fn group_map(data: &[u8], at: usize) -> Result<HashMap<char, u16>, Error> {
    let count = read_u32(data, at + 12)? as usize;
    let mut found = HashMap::new();

    for index in 0..count {
        let group = at + 16 + index * 12;
        let start = read_u32(data, group)?;
        let end = read_u32(data, group + 4)?;
        let glyph = read_u32(data, group + 8)?;

        if start > end {
            continue;
        }

        for step in 0..=(end - start) {
            if found.len() >= MAP_LIMIT {
                return Err(Error::new("cmap の対応づけが多すぎます", at));
            }

            let number = glyph + step;

            if number == 0 || number > 0xffff {
                continue;
            }

            if let Some(character) = char::from_u32(start + step) {
                found.insert(character, number as u16);
            }
        }
    }

    Ok(found)
}


/// name から書体の名前を取り出す。typographic family（16 番）を先に見て、なければ family（1 番）
pub fn family_name(data: &[u8], at: usize) -> Result<String, Error> {
    let count = read_u16(data, at + 2)? as usize;
    let storage = at + read_u16(data, at + 4)? as usize;
    let mut best: Option<(u8, String)> = None;

    for index in 0..count {
        let record = at + 6 + index * 12;
        let platform = read_u16(data, record)?;
        let encoding = read_u16(data, record + 2)?;
        let name_id = read_u16(data, record + 6)?;

        if name_id != 1 && name_id != 16 {
            continue;
        }

        let length = read_u16(data, record + 8)? as usize;
        let offset = storage + read_u16(data, record + 10)? as usize;
        let slice = match data.get(offset..offset + length) {
            Some(slice) => slice,
            None => continue,
        };
        let text = match (platform, encoding) {
            (3, _) | (0, _) => decode_utf16(slice),
            (1, 0) => slice.iter().map(|byte| *byte as char).collect(),
            _ => continue,
        };

        if text.is_empty() {
            continue;
        }

        let score = if name_id == 16 {2} else {1};

        if best.as_ref().is_none_or(|(current, _)| score > *current) {
            best = Some((score, text));
        }
    }

    match best {
        Some((_, text)) => Ok(text),
        None => Err(Error::new("name テーブルに書体の名前がありません", at)),
    }
}


fn decode_utf16(slice: &[u8]) -> String {
    let units: Vec<u16> = slice
        .chunks_exact(2)
        .map(|pair| u16::from_be_bytes([pair[0], pair[1]]))
        .collect();

    String::from_utf16_lossy(&units)
}


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

    fn sample() -> Vec<u8> {
        std::fs::read("font/latinmodern-math.otf").expect("試験用のフォントを読めません")
    }

    #[test]
    fn reads_numbers() {
        let data = [0x12u8, 0x34, 0xff, 0xfe];
        assert_eq!(read_u16(&data, 0).unwrap(), 0x1234);
        assert_eq!(read_i16(&data, 2).unwrap(), -2);
        assert_eq!(read_u32(&data, 0).unwrap(), 0x1234fffe);
        assert!(read_u16(&data, 3).is_err());
    }

    #[test]
    fn finds_the_tables_of_the_sample_font() {
        let data = sample();
        assert!(find(&data, "CFF ").unwrap().is_some());
        assert!(find(&data, "MATH").unwrap().is_some());
        assert!(find(&data, "glyf").unwrap().is_none());
        assert!(require(&data, "nope").is_err());
    }

    #[test]
    fn reads_head_and_the_glyph_count() {
        let data = sample();
        let (head_at, _) = require(&data, "head").unwrap();
        assert_eq!(head(&data, head_at).unwrap().units_per_em, 1000.0);
        let (maxp_at, _) = require(&data, "maxp").unwrap();
        assert!(glyph_count(&data, maxp_at).unwrap() > 1000);
    }

    #[test]
    fn maps_characters_to_glyphs() {
        let data = sample();
        let (cmap_at, _) = require(&data, "cmap").unwrap();
        let map = character_map(&data, cmap_at).unwrap();
        assert!(map.contains_key(&'x'));
        assert!(map.contains_key(&'\u{221a}'));
        assert!(map.contains_key(&'\u{2211}'));
        assert_ne!(map[&'x'], map[&'y']);
    }

    #[test]
    fn reads_the_advances() {
        let data = sample();
        let (hmtx_at, _) = require(&data, "hmtx").unwrap();
        let (hhea_at, _) = require(&data, "hhea").unwrap();
        let (maxp_at, _) = require(&data, "maxp").unwrap();
        let count = glyph_count(&data, maxp_at).unwrap();
        let width = advances(&data, hmtx_at, hhea_at, count).unwrap();
        assert_eq!(width.len(), count);
        assert!(width.iter().any(|value| *value > 0.0));
    }

    #[test]
    fn reads_the_family_name() {
        let data = sample();
        let (name_at, _) = require(&data, "name").unwrap();
        assert_eq!(family_name(&data, name_at).unwrap(), "Latin Modern Math");
    }
}