karlos-font-helper/src/sff.rs

#[derive(Debug)]
struct Table {
    first_codepoint: u32,
    num_codepoints: u32,
    offset: u32,
}

#[derive(Debug, Copy, Clone, PartialEq, Eq)]
pub enum Encoding {
    BitPerPix1,
    BitPerPix2,
    BitPerPix4,
    BitPerPix8,
}

struct Access {
    byte_index: usize,
    shift: u8,
    mask: u8,
}

impl Encoding {
    fn into_code(self) -> u16 {
        match self {
            Encoding::BitPerPix1 => 0,
            Encoding::BitPerPix2 => 1,
            Encoding::BitPerPix4 => 2,
            Encoding::BitPerPix8 => 3,
        }
    }

    fn access_from_index(self, px_index: usize) -> Access {
        match self {
            Encoding::BitPerPix1 => Access {
                byte_index: px_index / 8,
                shift: (7 - (px_index % 8)) as u8,
                mask: 0x01,
            },
            Encoding::BitPerPix2 => Access {
                byte_index: px_index / 4,
                shift: ((3 - (px_index % 4)) << 1) as u8,
                mask: 0x03,
            },
            Encoding::BitPerPix4 => Access {
                byte_index: px_index / 2,
                shift: ((1 - (px_index % 2)) << 2) as u8,
                mask: 0x0f,
            },
            Encoding::BitPerPix8 => Access {
                byte_index: px_index,
                shift: 0,
                mask: 0xff,
            },
        }
    }

    /// Compresses standard brightness data (0-255) into a suitable form for sff.
    fn compress(self, input: &[u8]) -> Vec<u8> {
        let mut out = Vec::new();
        match self {
            Encoding::BitPerPix1 => todo!(),
            Encoding::BitPerPix2 => todo!(),
            Encoding::BitPerPix4 => {
                assert!(input.len() % 2 == 0); // TODO not strictly necessary
                for c in 0..input.len() / 2 {
                    let one = input[2 * c] >> 4;
                    let two = input[2 * c + 1] >> 4;
                    out.push((one << 4) | two);
                }
            }
            Encoding::BitPerPix8 => Vec::extend_from_slice(&mut out, input),
        }
        out
    }
}

fn get_u16(input: &[u8], byte_offset: usize) -> u16 {
    assert!(byte_offset % 2 == 0);
    input[byte_offset] as u16 | (input[byte_offset + 1] as u16) << 8
}

fn get_u32(input: &[u8], byte_offset: usize) -> u32 {
    assert!(byte_offset % 4 == 0);
    input[byte_offset] as u32
        | (input[byte_offset + 1] as u32) << 8
        | (input[byte_offset + 2] as u32) << 16
        | (input[byte_offset + 3] as u32) << 24
}

#[derive(Debug)]
struct FileDesc<'a> {
    input: &'a [u8],
    char_width: usize,
    char_height: usize,
    encoding: Encoding,
    tables: Vec<Table>,
}

impl<'a> FileDesc<'a> {
    fn bytes_per_codepoint(&self) -> usize {
        let px_per_codepoint = self.char_width * self.char_height;
        match self.encoding {
            Encoding::BitPerPix1 => px_per_codepoint >> 3,
            Encoding::BitPerPix2 => px_per_codepoint >> 2,
            Encoding::BitPerPix4 => px_per_codepoint >> 1,
            Encoding::BitPerPix8 => px_per_codepoint,
        }
    }

    fn find_codepoint(&self, codepoint: u32) -> Option<&[u8]> {
        self.tables
            .iter()
            .find(|tab| {
                tab.first_codepoint <= codepoint
                    && tab.first_codepoint + tab.num_codepoints > codepoint
            })
            .map(|tab| {
                let off = tab.offset as usize
                    + (codepoint - tab.first_codepoint) as usize * self.bytes_per_codepoint();
                &self.input[off..off + self.bytes_per_codepoint()]
            })
    }

    // input should be result of find_codepoint here
    fn get_pixel_value(&self, input: &[u8], row: u8, col: u8) -> Option<u8> {
        if row as usize >= self.char_height || col as usize >= self.char_width {
            return None;
        }
        let px_index = row as usize * self.char_width + col as usize;
        let Access {
            byte_index,
            shift,
            mask,
        } = self.encoding.access_from_index(px_index);
        Some((input[byte_index] >> shift) & mask)
    }
}

fn decode(input: &[u8]) -> FileDesc<'_> {
    if input[0] != b's' || input[1] != b'f' || input[2] != b'f' || input[3] != b'\0' {
        panic!("invalid header magic");
    }
    if get_u32(input, 4) != 1 {
        panic!("invalid version");
    }
    let char_width = get_u16(input, 0x8) as usize;
    let char_height = get_u16(input, 0xa) as usize;
    let encoding = match get_u16(input, 0xc) {
        0 => Encoding::BitPerPix1,
        1 => Encoding::BitPerPix2,
        2 => Encoding::BitPerPix4,
        3 => Encoding::BitPerPix8,
        _ => unreachable!(),
    };
    let num_tables = get_u16(input, 0xe);
    let mut tables = Vec::new();
    let mut coff = 0x10;
    for _ in 0..num_tables {
        let tab = Table {
            first_codepoint: get_u32(input, coff),
            num_codepoints: get_u32(input, coff + 4),
            offset: get_u32(input, coff + 8),
        };
        tables.push(tab);
        coff += 3 * 4;
    }

    FileDesc {
        input,
        char_width,
        char_height,
        encoding,
        tables,
    }
}

/// data holds unencoded values, from 0 to 255. They are scaled down/rounded internally.
#[derive(Debug)]
pub struct TableOut {
    pub first_codepoint: u32,
    pub num_codepoints: u32,
    pub data: Vec<Vec<u8>>,
}

#[derive(Debug)]
pub struct FileDescOut {
    pub char_width: usize,
    pub char_height: usize,
    pub encoding: Encoding,
    pub tables: Vec<TableOut>,
}

fn push_u16(v: &mut Vec<u8>, value: u16) {
    v.push((value & 0xff) as u8);
    v.push(((value >> 8) & 0xff) as u8);
}

fn push_u32(v: &mut Vec<u8>, value: u32) {
    v.push((value & 0xff) as u8);
    v.push(((value >> 8) & 0xff) as u8);
    v.push(((value >> 16) & 0xff) as u8);
    v.push(((value >> 24) & 0xff) as u8);
}

fn put_u16(output: &mut [u8], byte_offset: usize, value: u16) {
    assert!(byte_offset % 2 == 0);
    output[byte_offset] = (value & 0xff) as u8;
    output[byte_offset + 1] = ((value >> 8) & 0xff) as u8;
}

fn put_u32(output: &mut [u8], byte_offset: usize, value: u32) {
    assert!(byte_offset % 4 == 0);
    output[byte_offset] = (value & 0xff) as u8;
    output[byte_offset + 1] = ((value >> 8) & 0xff) as u8;
    output[byte_offset + 2] = ((value >> 16) & 0xff) as u8;
    output[byte_offset + 3] = ((value >> 24) & 0xff) as u8;
}

impl FileDescOut {
    pub fn encode(&self) -> Vec<u8> {
        let mut v = vec![b's', b'f', b'f', b'\0'];
        push_u32(&mut v, 1);
        assert!(self.char_width < 1 << 16);
        assert!(self.char_height < 1 << 16);
        push_u16(&mut v, self.char_width as u16);
        push_u16(&mut v, self.char_height as u16);
        push_u16(&mut v, self.encoding.into_code());
        push_u16(&mut v, self.tables.len().try_into().unwrap());
        for _ in 0..self.tables.len() {
            // first, reserve space for the table headers
            push_u32(&mut v, 0);
            push_u32(&mut v, 0);
            push_u32(&mut v, 0);
        }
        for (i, table) in self.tables.iter().enumerate() {
            let header_off = 0x10 + i * 3 * 4;
            put_u32(&mut v[..], header_off, table.first_codepoint);
            put_u32(&mut v[..], header_off + 4, table.num_codepoints);
            let curr_len = v.len() as u32;
            put_u32(&mut v[..], header_off + 8, curr_len);
            for glyph_data in table.data.iter() {
                // TODO ugly clone
                v.extend_from_slice(&self.encoding.compress(glyph_data)[..]);
            }
        }
        v
    }
}