SVG: delegate entity expansion bounds, sniffing, and SVGZ to craft-svg

Cargo.lock

@@ -1079,6 +1079,21 @@ dependencies = [
  "unicode-normalization",
 ]
 
+[[package]]
+name = "craft-svg"
+version = "0.1.0"
+source = "git+https://github.com/storytold/craft-libs?rev=6d620e971e820b517b11a99cad040ebac670e1e2#6d620e971e820b517b11a99cad040ebac670e1e2"
+dependencies = [
+ "craft-xml",
+ "flate2",
+ "kurbo 0.13.1",
+ "roxmltree 0.21.2",
+ "thiserror 2.0.21",
+ "usvg 0.48.1",
+ "vectorcraft-geom 0.18.0",
+ "vectorcraft-pathops 0.18.0",
+]
+
 [[package]]
 name = "craft-ui"
 version = "0.1.0"
@@ -1168,6 +1183,15 @@ dependencies = [
  "windows-sys 0.61.2",
 ]
 
+[[package]]
+name = "craft-xml"
+version = "0.1.0"
+source = "git+https://github.com/storytold/craft-libs?rev=6d620e971e820b517b11a99cad040ebac670e1e2#6d620e971e820b517b11a99cad040ebac670e1e2"
+dependencies = [
+ "quick-xml 0.41.0",
+ "thiserror 2.0.21",
+]
+
 [[package]]
 name = "crc32fast"
 version = "1.5.3"
@@ -6408,7 +6432,7 @@ dependencies = [
  "serde_json",
  "vectorcraft-color",
  "vectorcraft-doc",
- "vectorcraft-geom",
+ "vectorcraft-geom 0.24.1",
 ]
 
 [[package]]
@@ -6423,8 +6447,8 @@ dependencies = [
  "vectorcraft-color",
  "vectorcraft-doc",
  "vectorcraft-effects",
- "vectorcraft-geom",
- "vectorcraft-pathops",
+ "vectorcraft-geom 0.24.1",
+ "vectorcraft-pathops 0.24.1",
  "vectorcraft-text",
 ]
 
@@ -6441,7 +6465,7 @@ dependencies = [
  "vectorcraft-doc",
  "vectorcraft-engine",
  "vectorcraft-format",
- "vectorcraft-geom",
+ "vectorcraft-geom 0.24.1",
  "vectorcraft-mcp",
  "vectorcraft-render",
  "vectorcraft-svg",
@@ -6478,7 +6502,7 @@ dependencies = [
  "thiserror 2.0.21",
  "tiff",
  "vectorcraft-color",
- "vectorcraft-geom",
+ "vectorcraft-geom 0.24.1",
  "zune-jpeg 0.5.15",
 ]
 
@@ -6493,8 +6517,8 @@ dependencies = [
  "svgtypes 0.16.1",
  "vectorcraft-color",
  "vectorcraft-doc",
- "vectorcraft-geom",
- "vectorcraft-pathops",
+ "vectorcraft-geom 0.24.1",
+ "vectorcraft-pathops 0.24.1",
  "vectorcraft-plugins",
  "vectorcraft-text",
  "vectorcraft-three-d",
@@ -6538,10 +6562,10 @@ dependencies = [
  "vectorcraft-doc",
  "vectorcraft-eps",
  "vectorcraft-format",
- "vectorcraft-geom",
+ "vectorcraft-geom 0.24.1",
  "vectorcraft-js",
  "vectorcraft-metafile",
- "vectorcraft-pathops",
+ "vectorcraft-pathops 0.24.1",
  "vectorcraft-pdf",
  "vectorcraft-plugins",
  "vectorcraft-render",
@@ -6576,9 +6600,9 @@ dependencies = [
  "vectorcraft-color",
  "vectorcraft-doc",
  "vectorcraft-effects",
- "vectorcraft-geom",
+ "vectorcraft-geom 0.24.1",
  "vectorcraft-metafile",
- "vectorcraft-pathops",
+ "vectorcraft-pathops 0.24.1",
  "vectorcraft-pdf",
  "vectorcraft-render",
  "vectorcraft-testkit",
@@ -6597,10 +6621,20 @@ dependencies = [
  "serde_json",
  "thiserror 2.0.21",
  "vectorcraft-doc",
- "vectorcraft-geom",
+ "vectorcraft-geom 0.24.1",
  "vectorcraft-testkit",
 ]
 
+[[package]]
+name = "vectorcraft-geom"
+version = "0.18.0"
+source = "git+https://github.com/storytold/vectorcraft.git?rev=6b3a6b8788ebcb420d69449199cbdf323888d27e#6b3a6b8788ebcb420d69449199cbdf323888d27e"
+dependencies = [
+ "kurbo 0.13.1",
+ "serde",
+ "thiserror 2.0.21",
+]
+
 [[package]]
 name = "vectorcraft-geom"
 version = "0.24.1"
@@ -6635,7 +6669,7 @@ dependencies = [
  "vectorcraft-doc",
  "vectorcraft-engine",
  "vectorcraft-format",
- "vectorcraft-geom",
+ "vectorcraft-geom 0.24.1",
  "vectorcraft-render",
  "vectorcraft-svg",
  "vectorcraft-testkit",
@@ -6652,10 +6686,21 @@ dependencies = [
  "vectorcraft-color",
  "vectorcraft-doc",
  "vectorcraft-effects",
- "vectorcraft-geom",
+ "vectorcraft-geom 0.24.1",
  "vectorcraft-text",
 ]
 
+[[package]]
+name = "vectorcraft-pathops"
+version = "0.18.0"
+source = "git+https://github.com/storytold/vectorcraft.git?rev=6b3a6b8788ebcb420d69449199cbdf323888d27e#6b3a6b8788ebcb420d69449199cbdf323888d27e"
+dependencies = [
+ "kurbo 0.13.1",
+ "linesweeper",
+ "thiserror 2.0.21",
+ "vectorcraft-geom 0.18.0",
+]
+
 [[package]]
 name = "vectorcraft-pathops"
 version = "0.24.1"
@@ -6664,7 +6709,7 @@ dependencies = [
  "linesweeper",
  "proptest",
  "thiserror 2.0.21",
- "vectorcraft-geom",
+ "vectorcraft-geom 0.24.1",
  "vectorcraft-testkit",
 ]
 
@@ -6697,7 +6742,7 @@ dependencies = [
  "vectorcraft-color",
  "vectorcraft-doc",
  "vectorcraft-effects",
- "vectorcraft-geom",
+ "vectorcraft-geom 0.24.1",
  "vectorcraft-testkit",
  "vectorcraft-text",
  "weezl 0.2.1",
@@ -6716,7 +6761,7 @@ dependencies = [
  "thiserror 2.0.21",
  "vectorcraft-color",
  "vectorcraft-doc",
- "vectorcraft-geom",
+ "vectorcraft-geom 0.24.1",
  "wasmi",
  "wat",
  "web-time",
@@ -6741,7 +6786,7 @@ dependencies = [
  "vectorcraft-color",
  "vectorcraft-doc",
  "vectorcraft-effects",
- "vectorcraft-geom",
+ "vectorcraft-geom 0.24.1",
  "vectorcraft-plugins",
  "vectorcraft-testkit",
  "vectorcraft-text",
@@ -6765,6 +6810,7 @@ name = "vectorcraft-svg"
 version = "0.24.1"
 dependencies = [
  "craft-docpath",
+ "craft-svg",
  "flate2",
  "image",
  "kurbo 0.13.1",
@@ -6781,8 +6827,8 @@ dependencies = [
  "vectorcraft-color",
  "vectorcraft-doc",
  "vectorcraft-effects",
- "vectorcraft-geom",
- "vectorcraft-pathops",
+ "vectorcraft-geom 0.24.1",
+ "vectorcraft-pathops 0.24.1",
  "vectorcraft-testkit",
  "vectorcraft-text",
 ]
@@ -6800,7 +6846,7 @@ dependencies = [
  "vectorcraft-doc",
  "vectorcraft-engine",
  "vectorcraft-format",
- "vectorcraft-geom",
+ "vectorcraft-geom 0.24.1",
  "vectorcraft-render",
  "vectorcraft-svg",
 ]
@@ -6820,7 +6866,7 @@ dependencies = [
  "unicode-script",
  "unicode-segmentation",
  "vectorcraft-doc",
- "vectorcraft-geom",
+ "vectorcraft-geom 0.24.1",
  "vello_cpu",
  "windows-registry",
  "write-fonts",
@@ -6830,8 +6876,8 @@ dependencies = [
 name = "vectorcraft-three-d"
 version = "0.24.1"
 dependencies = [
- "vectorcraft-geom",
- "vectorcraft-pathops",
+ "vectorcraft-geom 0.24.1",
+ "vectorcraft-pathops 0.24.1",
 ]
 
 [[package]]
@@ -6846,8 +6892,8 @@ dependencies = [
  "vectorcraft-color",
  "vectorcraft-doc",
  "vectorcraft-effects",
- "vectorcraft-geom",
- "vectorcraft-pathops",
+ "vectorcraft-geom 0.24.1",
+ "vectorcraft-pathops 0.24.1",
  "vectorcraft-text",
 ]
 
@@ -6860,7 +6906,7 @@ dependencies = [
  "serde",
  "serde_json",
  "thiserror 2.0.21",
- "vectorcraft-geom",
+ "vectorcraft-geom 0.24.1",
 ]
 
 [[package]]
@@ -6889,7 +6935,7 @@ dependencies = [
  "vectorcraft-engine",
  "vectorcraft-eps",
  "vectorcraft-format",
- "vectorcraft-geom",
+ "vectorcraft-geom 0.24.1",
  "vectorcraft-pdf",
  "vectorcraft-plugins",
  "vectorcraft-render",

crates/eps/src/ps.rs

@@ -21,9 +21,6 @@ VCdict begin
 end
 ";
 
-/// Longest line of encoded data (DSC readers take up to 255 characters).
-const LINE: usize = 76;
-
 /// `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();

crates/svg/Cargo.toml

@@ -9,6 +9,7 @@ description = "VectorCraft SVG import and export"
 workspace = true
 
 [dependencies]
+craft-svg = { workspace = true }
 # Which paths stored in documents may be opened.
 craft-docpath = { workspace = true }
 vectorcraft-geom = { workspace = true }

crates/svg/src/entities.rs

@@ -2,90 +2,10 @@
 //! replaces every `&name;` with the entity's text as it reads, with no total limit: a short file
 //! that repeats a large entity in an attribute allocates far more than it holds.
 
-use std::collections::HashMap;
-
-/// The most text the references to entities may expand to in all (bytes).
-const MAX_EXPANDED: usize = 16 << 20;
-
-/// How deep entities that refer to entities are followed (the reader's own limit is lower).
-const MAX_NESTING: u32 = 16;
-
-/// The longest entity name a reference is looked for with ([`references`]).
-const MAX_NAME: usize = 256;
-
-/// The internal general entities declared in `svg`: name → every value declared for it. Text that
-/// only looks like a declaration (in a comment, or in another entity's value) is kept too: the
-/// largest value counts, so it can only make the estimate larger.
-fn declared(svg: &str) -> HashMap<&str, Vec<&str>> {
-    let mut map: HashMap<&str, Vec<&str>> = HashMap::new();
-    for (at, _) in svg.match_indices("<!ENTITY") {
-        let rest = svg.get(at + "<!ENTITY".len()..).unwrap_or("").trim_start();
-        // A parameter entity (`%`) isn't referenced as `&name;`.
-        if rest.starts_with('%') {
-            continue;
-        }
-        let Some(end) = rest.find(char::is_whitespace) else { continue };
-        let (name, rest) = rest.split_at(end);
-        let rest = rest.trim_start();
-        let Some(quote) = rest.chars().next().filter(|q| matches!(q, '"' | '\'')) else { continue };
-        let Some(value) = rest.get(1..).and_then(|r| r.split_once(quote)).map(|(v, _)| v) else { continue };
-        map.entry(name).or_default().push(value);
-    }
-    map
-}
-
-/// The names of the `&name;` references in `text`.
-fn references(text: &str) -> impl Iterator<Item = &str> {
-    text.match_indices('&').filter_map(|(at, _)| {
-        let rest = text.get(at + 1..)?;
-        // Look no further than the longest name: a file of `&`s without `;` would be read quadratically.
-        let end = rest.bytes().take(MAX_NAME + 1).position(|b| b == b';')?;
-        let name = rest.get(..end)?;
-        (!name.is_empty() && name.len() <= MAX_NAME && !name.starts_with('#') && !name.contains(char::is_whitespace)).then_some(name)
-    })
-}
-
-/// What entity `name` expands to (bytes), references in its value followed.
-fn expanded<'a>(name: &'a str, entities: &HashMap<&'a str, Vec<&'a str>>, memo: &mut HashMap<&'a str, usize>, depth: u32) -> usize {
-    if let Some(n) = memo.get(name) {
-        return *n;
-    }
-    let Some(values) = entities.get(name) else { return 0 };
-    if depth >= MAX_NESTING {
-        return usize::MAX;
-    }
-    let mut most = 0usize;
-    for value in values {
-        let mut total = value.len();
-        for inner in references(value) {
-            total = total.saturating_add(expanded(inner, entities, memo, depth + 1));
-        }
-        most = most.max(total);
-    }
-    memo.insert(name, most);
-    most
-}
-
-/// Whether the references to the entities `svg` declares expand to more than [`MAX_EXPANDED`]
+/// Whether the references to the entities `svg` declares expand to more than the limit
 /// bytes in all. False for a file that declares none.
 pub(crate) fn expansion_too_large(svg: &str) -> bool {
-    let entities = declared(svg);
-    if entities.is_empty() {
-        return false;
-    }
-    // A reference to a longer name isn't looked for: refuse such a declaration instead.
-    if entities.keys().any(|name| name.len() > MAX_NAME) {
-        return true;
-    }
-    let mut memo = HashMap::new();
-    let mut total = 0usize;
-    for name in references(svg) {
-        total = total.saturating_add(expanded(name, &entities, &mut memo, 0));
-        if total > MAX_EXPANDED {
-            return true;
-        }
-    }
-    false
+    craft_svg::guard::entities_expand_past(svg, craft_svg::guard::MAX_ENTITY_EXPANSION)
 }
 
 #[cfg(test)]
@@ -118,7 +38,7 @@ mod tests {
         assert!(expansion_too_large(&commented));
         let in_value = format!("<!DOCTYPE svg [<!ENTITY c \"<!ENTITY a 'x'>\"><!ENTITY a \"{big}\">]><svg id=\"{refs}\"/>");
         assert!(expansion_too_large(&in_value));
-        let long = "n".repeat(MAX_NAME + 1);
+        let long = "n".repeat(257);
         let long_name = format!("<!DOCTYPE svg [<!ENTITY {long} \"x\">]><svg id=\"&{long};\"/>");
         assert!(expansion_too_large(&long_name));
     }

crates/svg/src/lib.rs

@@ -458,21 +458,17 @@ pub(crate) fn body_hash(parts: &[&str]) -> String {
 
 /// Does `bytes` start like gzip data (an SVGZ file)?
 pub fn is_svgz(bytes: &[u8]) -> bool {
-    bytes.starts_with(&[0x1f, 0x8b])
+    craft_svg::sniff::is_svgz(bytes)
 }
 
 /// Compress SVG text as SVGZ (gzip).
 pub fn compress(svg: &str) -> Vec<u8> {
-    compress_bytes(svg.as_bytes())
+    craft_svg::svgz::compress(svg.as_bytes())
 }
 
 /// Compress an SVG file (in any encoding, see [`Output::bytes`]) as SVGZ (gzip).
 pub fn compress_bytes(svg: &[u8]) -> Vec<u8> {
-    use std::io::Write as _;
-    let mut gz = flate2::write::GzEncoder::new(Vec::with_capacity(svg.len() / 4), flate2::Compression::default());
-    // Writing to a Vec cannot fail.
-    let _ = gz.write_all(svg);
-    gz.finish().unwrap_or_default()
+    craft_svg::svgz::compress(svg)
 }
 
 /// Most bytes an SVGZ file may unpack to (a guard against decompression bombs).
@@ -487,29 +483,26 @@ pub fn text_of(bytes: &[u8]) -> Result<std::borrow::Cow<'_, str>, SvgError> {
 /// [`text_of`], SVGZ bytes unpacking within what `unpacked`, the bytes SVGZ data unpacked to so
 /// far, leaves of [`MAX_SVGZ`]. What they unpack to is added to it, whether or not it decodes.
 pub(crate) fn text_within<'a>(bytes: &'a [u8], unpacked: &std::sync::atomic::AtomicU64) -> Result<std::borrow::Cow<'a, str>, SvgError> {
-    use std::io::Read as _;
     use std::sync::atomic::Ordering;
     if !is_svgz(bytes) {
-        return match decode(bytes)? {
-            Decoded::Utf8(skip) => std::str::from_utf8(bytes.get(skip..).unwrap_or_default()).map(Into::into).map_err(|_| not_utf8()),
-            Decoded::Text(t) => Ok(t.into()),
-        };
+        return craft_svg::svgz::decode(bytes).map_err(|_| not_utf8());
     }
-    let most = MAX_SVGZ.saturating_sub(unpacked.load(Ordering::Relaxed));
-    let mut raw = Vec::new();
-    let read = flate2::read::GzDecoder::new(bytes).take(most.saturating_add(1)).read_to_end(&mut raw);
-    // Read before a failure is kept: it was unpacked too.
-    unpacked.fetch_add(raw.len() as u64, Ordering::Relaxed);
-    read.map_err(|e| SvgError::Parse(format!("not a valid SVGZ file: {e}")))?;
-    if raw.len() as u64 > most {
-        return Err(SvgError::TooBig(format!("the SVGZ file unpacks to more than {} MB", most >> 20)));
-    }
-    match decode(&raw)? {
-        Decoded::Utf8(skip) => {
-            raw.drain(..skip);
-            String::from_utf8(raw).map(Into::into).map_err(|_| not_utf8())
+    let most = usize::try_from(MAX_SVGZ.saturating_sub(unpacked.load(Ordering::Relaxed))).unwrap_or(usize::MAX);
+    match craft_svg::svgz::inflate(bytes, most) {
+        Ok(raw) => {
+            unpacked.fetch_add(raw.len() as u64, Ordering::Relaxed);
+            craft_svg::svgz::decode(&raw)
+                .map(|t| std::borrow::Cow::Owned(t.into_owned()))
+                .map_err(|_| not_utf8())
         }
-        Decoded::Text(t) => Ok(t.into()),
+        Err(craft_svg::svgz::Error::TooLarge(_)) => {
+            Err(SvgError::TooBig(format!("the SVGZ file unpacks to more than {} MB", most >> 20)))
+        }
+        Err(craft_svg::svgz::Error::Corrupt(e)) => {
+            Err(SvgError::Parse(format!("not a valid SVGZ file: {e}")))
+        }
+        Err(craft_svg::svgz::Error::Encoding) => Err(not_utf8()),
+        Err(e) => Err(SvgError::Parse(e.to_string())),
     }
 }
 
@@ -517,44 +510,6 @@ fn not_utf8() -> SvgError {
     SvgError::Parse("SVG is not UTF-8, UTF-16 or ISO 8859-1".into())
 }
 
-/// How SVG bytes are read.
-enum Decoded {
-    /// UTF-8 after this many bytes (a byte order mark).
-    Utf8(usize),
-    /// Text decoded from another encoding.
-    Text(String),
-}
-
-/// The encoding of SVG bytes: a byte order mark (UTF-8 or UTF-16), else the XML declaration's
-/// (ISO 8859-1 by its names; anything else is read as UTF-8).
-fn decode(bytes: &[u8]) -> Result<Decoded, SvgError> {
-    let utf16 = |rest: &[u8], unit: fn([u8; 2]) -> u16| {
-        let (units, odd) = rest.as_chunks::<2>();
-        if !odd.is_empty() {
-            return Err(not_utf8());
-        }
-        char::decode_utf16(units.iter().map(|u| unit(*u))).collect::<Result<String, _>>().map(Decoded::Text).map_err(|_| not_utf8())
-    };
-    match bytes {
-        [0xef, 0xbb, 0xbf, ..] => Ok(Decoded::Utf8(3)),
-        [0xfe, 0xff, rest @ ..] => utf16(rest, u16::from_be_bytes),
-        [0xff, 0xfe, rest @ ..] => utf16(rest, u16::from_le_bytes),
-        _ if declared_latin1(bytes) => Ok(Decoded::Text(bytes.iter().map(|b| char::from(*b)).collect())),
-        _ => Ok(Decoded::Utf8(0)),
-    }
-}
-
-/// Does the XML declaration at the start of `bytes` name ISO 8859-1?
-fn declared_latin1(bytes: &[u8]) -> bool {
-    let head = bytes.get(..bytes.len().min(200)).unwrap_or_default();
-    let Some(decl) = head.strip_prefix(b"<?xml").and_then(|d| d.split(|b| *b == b'>').next()) else { return false };
-    let decl = String::from_utf8_lossy(decl).to_ascii_lowercase();
-    let Some(at) = decl.find("encoding") else { return false };
-    let value = decl.get(at + "encoding".len()..).unwrap_or("").trim_start().trim_start_matches('=').trim_start();
-    let name = value.trim_start_matches(['"', '\'']).split(['"', '\'', ' ', '?']).next().unwrap_or("");
-    matches!(name, "iso-8859-1" | "iso8859-1" | "iso_8859-1" | "latin1" | "latin-1" | "l1")
-}
-
 /// Import an SVG document.
 pub fn import(svg: &str) -> Result<Document, SvgError> {
     import_with_report(svg).map(|(d, _)| d)