Merge pull request #1212 from storytold/codex/craft-eps-fontscan

Extract embedded PostScript font scanning to craft-eps

crates/eps/src/embedded.rs

@@ -0,0 +1,131 @@
+//! Embedded Type 1 and CFF font programs in PostScript / EPS source.
+
+/// A font program embedded in the PostScript source of an EPS file.
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct EmbeddedFontProgram {
+    /// The PostScript `/FontName` when one is present in the source. CFF `StartData` blocks store
+    /// their name inside the CFF Name INDEX, which the font parser reads from `program`.
+    pub name: Option<String>,
+    /// The bytes of the Type 1 program or CFF font set.
+    pub program: Vec<u8>,
+    /// Whether `program` is a CFF font set rather than a Type 1 program.
+    pub cff: bool,
+}
+
+/// Embedded font programs in a PostScript section, in source order within each format.
+///
+/// Type 1 programs are recognized from their `eexec` and `cleartomark` delimiters and are named
+/// by `/FontName`. CFF `StartData` blocks are returned without a name because the name is inside
+/// the binary CFF data. Malformed or truncated blocks are skipped, as in the original EPS reader.
+pub fn embedded_font_programs(ps: &[u8]) -> Vec<EmbeddedFontProgram> {
+    let mut out = Vec::new();
+    let mut heads = positions(ps, b"%!PS-AdobeFont");
+    heads.extend(positions(ps, b"%!FontType1"));
+    heads.sort_unstable();
+    let names = positions(ps, b"/FontName");
+    let mut from = 0;
+    let mut used = 0;
+    while let Some(ee) = find_from(ps, b"eexec", from) {
+        from = ee + 5;
+        let last_before = |v: &[usize]| v[..v.partition_point(|&p| p < ee)].last().copied().filter(|&p| p >= used);
+        // The fallback look-back stops at the previous font's end, so it cannot pick up its header.
+        let start = last_before(&heads).or_else(|| last_before(&names).map(|p| p.saturating_sub(512).max(used)));
+        let Some(start) = start else { continue };
+        let end = find_from(ps, b"cleartomark", ee).map_or(ps.len(), |p| p + 11);
+        if let Some(name) = name_after(&ps[start..ee], b"/FontName") {
+            out.push(EmbeddedFontProgram { name: Some(name), program: ps[start..end].to_vec(), cff: false });
+        }
+        from = end;
+        used = end;
+    }
+
+    let mut from = 0;
+    while let Some(sd) = find_from(ps, b"StartData", from) {
+        from = sd + 9;
+        let head = String::from_utf8_lossy(&ps[sd.saturating_sub(64)..sd]).into_owned();
+        let Some(len) = head.split_whitespace().last().and_then(|t| t.parse::<usize>().ok()) else { continue };
+        let s = sd + 10;
+        let Some(data) = ps.get(s..s.saturating_add(len)) else { continue };
+        out.push(EmbeddedFontProgram { name: None, program: data.to_vec(), cff: true });
+        from = s + len;
+    }
+    out
+}
+
+fn find_from(hay: &[u8], needle: &[u8], from: usize) -> Option<usize> {
+    hay.get(from..)?.windows(needle.len()).position(|w| w == needle).map(|p| p + from)
+}
+
+fn positions(hay: &[u8], needle: &[u8]) -> Vec<usize> {
+    let mut out = vec![];
+    let mut from = 0;
+    while let Some(p) = find_from(hay, needle, from) {
+        out.push(p);
+        from = p + needle.len();
+    }
+    out
+}
+
+fn name_after(b: &[u8], key: &[u8]) -> Option<String> {
+    let p = find_from(b, key, 0)? + key.len();
+    let rest = &b[p..];
+    let s = rest.iter().position(|c| *c == b'/')? + 1;
+    let e = rest[s..].iter().position(|c| c.is_ascii_whitespace() || matches!(c, b'/' | b'[' | b'(' | b'{')).map_or(rest.len(), |q| s + q);
+    (e > s).then(|| String::from_utf8_lossy(&rest[s..e]).into_owned())
+}
+
+#[cfg(test)]
+mod tests {
+    #![allow(clippy::unwrap_used, clippy::expect_used, clippy::panic)]
+
+    use super::embedded_font_programs;
+
+    #[test]
+    fn finds_type1_font_programs_by_name() {
+        let ps = b"%!PS-Adobe-3.0 EPSF-3.0\n%!PS-AdobeFont-1.0: Test\n/FontName /Test def\ncurrentfile eexec\n0000\ncleartomark\n";
+        let fonts = embedded_font_programs(ps);
+        assert_eq!(fonts.len(), 1);
+        assert_eq!(fonts[0].name.as_deref(), Some("Test"));
+        assert!(!fonts[0].cff);
+        assert!(String::from_utf8_lossy(&fonts[0].program).contains("cleartomark"));
+    }
+
+    #[test]
+    fn finds_cff_startdata_without_interpreting_its_name_index() {
+        let ps = b"%!PS-Adobe-3.0 EPSF-3.0\n10 StartData\n0123456789\n";
+        let fonts = embedded_font_programs(ps);
+        assert_eq!(fonts.len(), 1);
+        assert_eq!(fonts[0].name, None);
+        assert!(fonts[0].cff);
+        assert_eq!(fonts[0].program, b"0123456789");
+    }
+
+    #[test]
+    fn ignores_truncated_startdata() {
+        let fonts = embedded_font_programs(b"%!PS-Adobe-3.0 EPSF-3.0\n100 StartData\nshort");
+        assert!(fonts.is_empty());
+    }
+
+    #[test]
+    fn many_headerless_eexec_words_do_not_hide_later_font() {
+        let mut ps = b"%!PS-Adobe-3.0 EPSF-3.0\n".to_vec();
+        ps.extend_from_slice(b"eexec ".repeat(20_000).as_slice());
+        ps.extend_from_slice(b"\n%!PS-AdobeFont-1.0: Real\n/FontName /Real def\ncurrentfile eexec\n0000\ncleartomark\n");
+        let fonts = embedded_font_programs(&ps);
+        assert_eq!(fonts.len(), 1);
+        assert_eq!(fonts[0].name.as_deref(), Some("Real"));
+    }
+
+    #[test]
+    fn headerless_font_does_not_reuse_previous_font_header() {
+        let first = b"%!PS-AdobeFont-1.0: A\n/FontName /A def\ncurrentfile eexec\n0000\ncleartomark\n";
+        let second = b"/FontName /B def\ncurrentfile eexec\n0000\ncleartomark\n";
+        let mut ps = b"%!PS-Adobe-3.0 EPSF-3.0\n".to_vec();
+        ps.extend_from_slice(first);
+        ps.extend_from_slice(second);
+        let fonts = embedded_font_programs(&ps);
+        assert_eq!(fonts.len(), 2);
+        let b = fonts.iter().find(|font| font.name.as_deref() == Some("B")).unwrap();
+        assert!(!String::from_utf8_lossy(&b.program).contains("AdobeFont"));
+    }
+}

crates/eps/src/lib.rs

@@ -18,6 +18,7 @@
 pub mod ascii85;
 mod container;
 mod dsc;
+mod embedded;
 #[cfg(feature = "paths")]
 pub mod paths;
 mod preview;
@@ -25,6 +26,7 @@ mod rectangles;
 
 pub use container::{DOS_HEADER, DOS_MAGIC, Sections, sections, wrap};
 pub use dsc::{Comments, MAX_HEADER, MAX_TRAILER};
+pub use embedded::{EmbeddedFontProgram, embedded_font_programs};
 pub use preview::{Bitmap, MAX_EPSI, epsi};
 pub use rectangles::{MAX_SIMPLE_RECTANGLE_SAVE_DEPTH, MAX_SIMPLE_RECTANGLE_TOKENS, MAX_SIMPLE_RECTANGLES, Rectangle, simple_rectangles};