vectorcraft/crates/eps/src/ps.rs

//! PostScript writing primitives: numbers, strings, paths, the prolog's short operator names, and
//! the data encodings images and embedded data use (ASCII85, run-length, Flate).

use std::fmt::Write;
use std::io::{Read, Write as _};

use kurbo::PathEl;
use vectorcraft_geom::{Affine, BezPath, FillRule};

/// The procedures the page uses, in `VCdict`.
pub(crate) const PROLOG: &str = "/VCdict 64 dict def
VCdict begin
/bd {bind def} bind def
/m {moveto} bd /l {lineto} bd /c {curveto} bd /h {closepath} bd
/f {fill} bd /f* {eofill} bd /S {stroke} bd
/W {clip newpath} bd /W* {eoclip newpath} bd
/q {gsave} bd /Q {grestore} bd /cm {concat} bd
/g {setgray} bd /rg {setrgbcolor} bd /k {setcmykcolor} bd
/w {setlinewidth} bd /J {setlinecap} bd /j {setlinejoin} bd /M {setmiterlimit} bd /d {setdash} bd
/op {setoverprint} bd
end
";

/// `v` as PostScript writes it: at most four decimals, no trailing zeros (0 when not finite).
pub(crate) fn num(v: f64) -> String {
    let mut s = String::new();
    push_num(&mut s, v);
    s
}

/// Append [`num`]`(v)` to `s`.
pub(crate) fn push_num(s: &mut String, v: f64) {
    push_num_places(s, v, 4);
}

/// The value [`num`]`(v)` stands for, as a reader parses it back.
pub(crate) fn written(v: f64) -> f64 {
    num(v).parse().unwrap_or(0.0)
}

/// Append `v` to `s` with at most `places` decimals (up to 9), no trailing zeros (0 when not
/// finite).
pub(crate) fn push_num_places(s: &mut String, v: f64, places: u32) {
    let places = places.min(9);
    let scale = 10f64.powi(places as i32);
    // Past where the scaling overflows, values have no fractional digits to round.
    let v = if !v.is_finite() {
        0.0
    } else if (v * scale).is_finite() {
        (v * scale).round() / scale
    } else {
        v
    };
    if v == v.trunc() && v.abs() < 1e15 {
        let _ = write!(s, "{}", v as i64);
        return;
    }
    let start = s.len();
    let _ = write!(s, "{v:.*}", places as usize);
    let keep = s[start..].trim_end_matches('0').trim_end_matches('.').len();
    s.truncate(start + keep);
}

/// `vs` as numbers separated by spaces, then a space.
pub(crate) fn push_nums(s: &mut String, vs: &[f64]) {
    for v in vs {
        push_num(s, *v);
        s.push(' ');
    }
}

/// `text` as a PostScript string: `(` `)` `\` escaped, anything outside printable ASCII as octal
/// escapes of its UTF-8 bytes.
pub(crate) fn string(text: &str) -> String {
    let mut s = String::with_capacity(text.len() + 2);
    s.push('(');
    for b in text.bytes() {
        match b {
            b'(' | b')' | b'\\' => {
                s.push('\\');
                s.push(char::from(b));
            }
            0x20..=0x7e => s.push(char::from(b)),
            _ => {
                let _ = write!(s, "\\{b:03o}");
            }
        }
    }
    s.push(')');
    s
}

/// `bytes` as a PostScript string: `(` `)` `\` escaped, carriage returns, line feeds and tabs as
/// `\r` `\n` `\t`, other bytes outside printable ASCII as octal escapes.
pub(crate) fn bytes(bytes: &[u8]) -> String {
    let mut s = String::with_capacity(bytes.len() + 2);
    s.push('(');
    for &b in bytes {
        match b {
            b'(' | b')' | b'\\' => {
                s.push('\\');
                s.push(char::from(b));
            }
            b'\r' => s.push_str("\\r"),
            b'\n' => s.push_str("\\n"),
            b'\t' => s.push_str("\\t"),
            0x20..=0x7e => s.push(char::from(b)),
            _ => {
                let _ = write!(s, "\\{b:03o}");
            }
        }
    }
    s.push(')');
    s
}

/// A matrix operand `[a b c d e f]`.
pub(crate) fn matrix(a: Affine) -> String {
    let mut s = String::from("[");
    push_nums(&mut s, &a.as_coeffs());
    s.pop();
    s.push(']');
    s
}

/// Append path `bp` (`m`, `l`, `c`, `h`; quadratics raised to cubics), one segment a line.
pub(crate) fn push_path(s: &mut String, bp: &BezPath) {
    let mut last = kurbo::Point::ZERO;
    let mut start = kurbo::Point::ZERO;
    for el in bp.elements() {
        match *el {
            PathEl::MoveTo(p) => {
                push_nums(s, &[p.x, p.y]);
                s.push_str("m\n");
                (last, start) = (p, p);
            }
            PathEl::LineTo(p) => {
                push_nums(s, &[p.x, p.y]);
                s.push_str("l\n");
                last = p;
            }
            PathEl::QuadTo(a, p) => {
                let c1 = last + (a - last) * (2.0 / 3.0);
                let c2 = p + (a - p) * (2.0 / 3.0);
                push_nums(s, &[c1.x, c1.y, c2.x, c2.y, p.x, p.y]);
                s.push_str("c\n");
                last = p;
            }
            PathEl::CurveTo(a, b, p) => {
                push_nums(s, &[a.x, a.y, b.x, b.y, p.x, p.y]);
                s.push_str("c\n");
                last = p;
            }
            PathEl::ClosePath => {
                s.push_str("h\n");
                last = start;
            }
        }
    }
}

/// The fill operator for `rule`.
pub(crate) fn fill_op(rule: FillRule) -> &'static str {
    match rule {
        FillRule::NonZero => "f\n",
        FillRule::EvenOdd => "f*\n",
    }
}

/// The clip operator for `rule` (it also ends the path).
pub(crate) fn clip_op(rule: FillRule) -> &'static str {
    match rule {
        FillRule::NonZero => "W\n",
        FillRule::EvenOdd => "W*\n",
    }
}

/// `bytes` in ASCII85 with its `~>` end, in lines of about [`LINE`] characters, none starting
/// with `%` (DSC readers would take it for a comment).
pub(crate) use craft_eps::ascii85::encode as ascii85;

/// Most bytes [`ascii85_decode`] makes: what one filter or data section may hold.
pub(crate) const MAX_ASCII85: usize = 1 << 28;

/// ASCII85 `text` (whitespace ignored, up to `~>`) decoded; `None` when it is malformed or longer
/// than [`MAX_ASCII85`].
pub(crate) fn ascii85_decode(text: &str) -> Option<Vec<u8>> {
    ascii85_decode_within(text, MAX_ASCII85)
}

/// [`ascii85_decode`] of at most `max` bytes: `z` is four bytes for one, so the length of the
/// text doesn't bound what it becomes (#1715).
pub(crate) fn ascii85_decode_within(text: &str, max: usize) -> Option<Vec<u8>> {
    craft_eps::ascii85::decode(text.as_bytes(), max)
}

/// Run-length encoding (PackBits: what PostScript's `RunLengthDecode` reads), shared with TIFF.
pub(crate) use vectorcraft_render::encode::tiff::packbits;

/// `data` compressed for `FlateDecode` (zlib).
pub(crate) fn deflate(data: &[u8]) -> Vec<u8> {
    let mut e = flate2::write::ZlibEncoder::new(Vec::with_capacity(data.len() / 2), flate2::Compression::default());
    // Writing to a Vec doesn't fail.
    let _ = e.write_all(data);
    e.finish().unwrap_or_default()
}

/// zlib `data` inflated, when it is valid and at most `max` bytes.
pub(crate) fn inflate(data: &[u8], max: u64) -> Option<Vec<u8>> {
    let mut out = vec![];
    flate2::read::ZlibDecoder::new(data).take(max + 1).read_to_end(&mut out).ok()?;
    (out.len() as u64 <= max).then_some(out)
}

#[cfg(test)]
mod tests {
    use super::{ascii85, ascii85_decode, num, push_num_places};

    #[test]
    fn ascii85_first_line_never_starts_with_percent() {
        let data = [0x0C, 0x80, 0, 0, 1, 2, 3, 4];
        let s = ascii85(&data);
        assert!(s.lines().all(|l| !l.starts_with('%')), "got {s:?}");
        assert_eq!(ascii85_decode(&s).as_deref(), Some(&data[..]));
    }

    #[test]
    fn huge_finite_numbers_are_valid_postscript() {
        for v in [1.5e305, -1.5e305, f64::MAX, f64::MIN] {
            let s = num(v);
            assert!(!s.is_empty() && s.chars().all(|c| c.is_ascii_digit() || c == '-' || c == '.'), "{v}: got {s:?}");
        }
    }

    #[test]
    fn numbers_keep_the_decimals_asked_for() {
        let places = |v: f64, n: u32| {
            let mut s = String::new();
            push_num_places(&mut s, v, n);
            s
        };
        assert_eq!(num(0.123_456_78), "0.1235");
        assert_eq!(places(0.123_456_78, 7), "0.1234568");
        assert_eq!(places(f64::from(0.4f32), 7), "0.4");
        assert_eq!(places(2.5, 7), "2.5");
        assert_eq!(places(f64::NAN, 7), "0");
    }
}