Release 0.2.1

Signed-off-by: Colin Walters <walters@verbum.org>

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "canon-json"
-version = "0.2.0"
+version = "0.2.1"
 edition = "2021"
 description = "serde_json Formatter to serialize as RFC 8785 canonical JSON"
 license = "MIT OR Apache-2.0"
@@ -20,3 +20,6 @@ cap-std = "3.4.4"
 sha2 = "0.10.9"
 # For cross-integration testing
 olpc-cjson = "0.1"
+cjson = "0.1.2"
+# For round trip testing
+serde_json = { version = "1.0", features = ["float_roundtrip"] }

src/lib.rs

@@ -3,6 +3,7 @@
 
 #![doc = include_str!("../README.md")]
 #![forbid(unsafe_code)]
+#![deny(missing_docs)]
 
 mod floatformat;
 
@@ -12,7 +13,7 @@ use std::io::{Error, ErrorKind, Result, Write};
 use serde::Serialize;
 use serde_json::ser::{CharEscape, CompactFormatter, Formatter, Serializer};
 
-/// A [`Formatter`] that produces canonical JSON.
+/// A [`Formatter`] that produces canonical (RFC 8785) JSON.
 ///
 /// See the [crate-level documentation](../index.html) for more detail.
 ///
@@ -51,8 +52,11 @@ impl ObjectKey {
         let val = serde_json::Value::String(s);
         let mut s = Serializer::new(w);
         val.serialize(&mut s).map_err(|e| {
-            let kind = e.io_error_kind().unwrap();
-            Error::new(kind, "I/O error")
+            if let Some(kind) = e.io_error_kind() {
+                Error::new(kind, "I/O error")
+            } else {
+                Error::new(ErrorKind::Other, e.to_string())
+            }
         })
     }
 }
@@ -351,6 +355,42 @@ impl Formatter for CanonicalFormatter {
     }
 }
 
+/// A helper trait to write canonical JSON.
+pub trait CanonJsonSerialize {
+    /// Serialize the given data structure as JSON into the I/O stream.
+    fn to_canon_json_writer<W>(&self, writer: W) -> Result<()>
+    where
+        W: Write;
+    /// Serialize the given data structure as a JSON byte vector.
+    fn to_canon_json_vec(&self) -> Result<Vec<u8>>;
+    /// Serialize the given data structure as a String.
+    fn to_canon_json_string(&self) -> Result<String>;
+}
+
+impl<S> CanonJsonSerialize for S
+where
+    S: Serialize,
+{
+    fn to_canon_json_writer<W>(&self, writer: W) -> Result<()>
+    where
+        W: Write,
+    {
+        let mut ser = Serializer::with_formatter(writer, CanonicalFormatter::new());
+        Ok(self.serialize(&mut ser)?)
+    }
+
+    fn to_canon_json_vec(&self) -> Result<Vec<u8>> {
+        let mut buf = Vec::new();
+        self.to_canon_json_writer(&mut buf)?;
+        Ok(buf)
+    }
+
+    fn to_canon_json_string(&self) -> Result<String> {
+        String::from_utf8(self.to_canon_json_vec()?)
+            .map_err(|err| Error::new(ErrorKind::InvalidData, err))
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;
@@ -358,8 +398,7 @@ mod tests {
     use std::{cmp::Ordering, io::Result};
 
     use proptest::prelude::*;
-    use serde::Serialize;
-    use serde_json::{Number, Serializer};
+    use serde_json::Number;
     use sha2::{Digest, Sha256};
     use similar_asserts::assert_eq;
 
@@ -390,10 +429,7 @@ mod tests {
     macro_rules! encode {
         ($($tt:tt)+) => {
             (|v: serde_json::Value| -> Result<Vec<u8>> {
-                let mut buf = Vec::new();
-                let mut ser = Serializer::with_formatter(&mut buf, CanonicalFormatter::new());
-                v.serialize(&mut ser)?;
-                Ok(buf)
+                v.to_canon_json_vec()
             })(serde_json::json!($($tt)+))
         };
     }
@@ -501,10 +537,6 @@ mod tests {
             i128: i128::MIN,
         };
 
-        let mut buf = Vec::new();
-        let mut ser = Serializer::with_formatter(&mut buf, CanonicalFormatter::new());
-        value.serialize(&mut ser).unwrap();
-
         let expected = [
             123, 34, 105, 49, 50, 56, 34, 58, 45, 49, 55, 48, 49, 52, 49, 49, 56, 51, 52, 54, 48,
             52, 54, 57, 50, 51, 49, 55, 51, 49, 54, 56, 55, 51, 48, 51, 55, 49, 53, 56, 56, 52, 49,
@@ -513,7 +545,7 @@ mod tests {
             55, 54, 56, 50, 49, 49, 52, 53, 53, 125,
         ];
 
-        assert_eq!(buf, expected);
+        assert_eq!(value.to_canon_json_vec().unwrap(), expected);
     }
 
     #[test]
@@ -524,24 +556,39 @@ mod tests {
         assert_eq!(&buf, &expected);
     }
 
-    fn arbitrary_json() -> impl Strategy<Value = serde_json::Value> {
+    /// As it says, generate arbitrary JSON. This is based on
+    /// https://proptest-rs.github.io/proptest/proptest/tutorial/recursive.html
+    ///
+    /// We support controlling the regex for keys, and whether or not floating point values are emitted.
+    fn arbitrary_json(
+        keyspace: &'static str,
+        allow_fp: bool,
+    ) -> impl Strategy<Value = serde_json::Value> {
         use serde_json::Value;
-        const S: &str = ".*";
         let leaf = prop_oneof![
             Just(Value::Null),
-            any::<u32>().prop_map(|v| Value::Number(Number::from_u128(v.into()).unwrap())),
+            any::<f64>().prop_filter_map("valid f64 for JSON", move |v| {
+                let n = if allow_fp && v.fract() != 0.0 {
+                    Number::from_f64(v).unwrap()
+                } else {
+                    // Constrain to values clearly lower than
+                    // the https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Number/MAX_SAFE_INTEGER
+                    Number::from_u128(v as u32 as u128).unwrap()
+                };
+                Some(Value::Number(n))
+            }),
             any::<bool>().prop_map(Value::Bool),
-            S.prop_map(Value::String),
+            keyspace.prop_map(Value::String),
         ];
         leaf.prop_recursive(
             8,   // 8 levels deep
             256, // Shoot for maximum size of 256 nodes
             10,  // We put up to 10 items per collection
-            |inner| {
+            move |inner| {
                 prop_oneof![
                     // Take the inner strategy and make the two recursive cases.
                     prop::collection::vec(inner.clone(), 0..10).prop_map(Value::Array),
-                    prop::collection::hash_map(S, inner, 0..10)
+                    prop::collection::hash_map(keyspace, inner, 0..10)
                         .prop_map(|v| { v.into_iter().collect() }),
                 ]
             },
@@ -550,7 +597,7 @@ mod tests {
 
     proptest! {
         #[test]
-        fn roundtrip_rfc8785(v in arbitrary_json()) {
+        fn roundtrip_rfc8785(v in arbitrary_json(".*", true)) {
             let buf = encode!(&v).unwrap();
             let v2: serde_json::Value = serde_json::from_slice(&buf)
                 .map_err(|e| format!("Failed to parse {v:?} -> {}: {e}", String::from_utf8_lossy(&buf))).unwrap();
@@ -560,11 +607,7 @@ mod tests {
 
     fn verify(input: &str, expected: &str) {
         let input: serde_json::Value = serde_json::from_str(input).unwrap();
-        let mut buf = Vec::new();
-        let mut ser = Serializer::with_formatter(&mut buf, CanonicalFormatter::new());
-        input.serialize(&mut ser).unwrap();
-        let buf = String::from_utf8(buf).unwrap();
-        assert_eq!(expected, &buf);
+        assert_eq!(expected, input.to_canon_json_string().unwrap());
     }
 
     #[test]
@@ -651,17 +694,44 @@ mod tests {
         Ok(())
     }
 
+    // Regex that excludes basically everything except printable ASCII
+    // because we know that e.g. olpc-cjson bombs on control characters,
+    // and also because it does NFC orering that will cause non-equivalency
+    // for some whitespace etc.
+    const ASCII_ALPHANUMERIC: &str = r"[a-zA-Z0-9]*";
+
     proptest! {
+        // Verify strict equivalency with printable ASCII only keys
         #[test]
-        fn crosscheck_olpc_cjson(v in arbitrary_json()) {
-            use olpc_cjson::CanonicalFormatter;
-
+        fn crosscheck_olpc_cjson_ascii(v in arbitrary_json(ASCII_ALPHANUMERIC, false)) {
+            let canon_json = String::from_utf8(encode!(&v).unwrap()).unwrap();
             let mut olpc_cjson_serialized = Vec::new();
-            let mut ser = serde_json::Serializer::with_formatter(&mut olpc_cjson_serialized, CanonicalFormatter::new());
-            prop_assume!(v.serialize(&mut ser).is_ok());
+            let mut ser = serde_json::Serializer::with_formatter(&mut olpc_cjson_serialized, olpc_cjson::CanonicalFormatter::new());
+            v.serialize(&mut ser).unwrap();
+            assert_eq!(canon_json, String::from_utf8(olpc_cjson_serialized).unwrap());
+        }
+    }
 
+    proptest! {
+        // Verify strict equivalency with printable ASCII only keys
+        #[test]
+        fn crosscheck_cjson_ascii(v in arbitrary_json(ASCII_ALPHANUMERIC, false)) {
+            let canon_json = String::from_utf8(encode!(&v).unwrap()).unwrap();
+            let cjson = String::from_utf8(cjson::to_vec(&v).unwrap()).unwrap();
+            assert_eq!(canon_json, cjson);
+        }
+
+        // Verify equivalency (after sorting) with non-ASCII keys
+        #[test]
+        fn crosscheck_cjson(v in arbitrary_json(".*", false)) {
             let buf = encode!(&v).unwrap();
-            assert_eq!(buf, olpc_cjson_serialized);
+            let self_reparsed = serde_json::from_slice::<serde_json::Value>(&buf).unwrap();
+            let buf = cjson::to_vec(&v).unwrap();
+            let cjson_reparsed = serde_json::from_slice::<serde_json::Value>(&buf).unwrap();
+            // As above with olpc-cjson, this relies on the fact that serde_json
+            // sorts object keys by default.
+            assert_eq!(self_reparsed, v);
+            assert_eq!(cjson_reparsed, v);
         }
     }
 }