Merge pull request #105 from oschwald/greg/optimize

Performance improvements

Author: oschwald

CHANGELOG.md

@@ -1,5 +1,13 @@
 # Change Log
 
+## 0.27.1 - 2025-12-18
+
+- Performance improvement: Skipped UTF-8 validation for map keys during
+  deserialization. This significantly speeds up full record decoding by
+  treating keys as raw bytes when matching against struct fields.
+- Performance improvement: Optimized tree traversal by reducing bounds checks
+  during node reading.
+
 ## 0.27.0 - 2025-11-28
 
 This release includes significant API changes. See [UPGRADING.md](UPGRADING.md)

src/decoder.rs

@@ -446,19 +446,26 @@ impl<'de> Decoder<'de> {
         }
     }
 
-    /// Reads a string directly, following pointers if needed.
-    pub(crate) fn read_string(&mut self) -> DecodeResult<&'de str> {
+    /// Reads a string's bytes directly, following pointers if needed.
+    /// Does NOT validate UTF-8.
+    pub(crate) fn read_str_as_bytes(&mut self) -> DecodeResult<&'de [u8]> {
         let (size, type_num) = self.size_and_type();
         if type_num == TYPE_POINTER {
             // Pointer
             let new_ptr = self.decode_pointer(size);
             let saved_ptr = self.current_ptr;
             self.current_ptr = new_ptr;
-            let result = self.read_string();
+            let result = self.read_str_as_bytes();
             self.current_ptr = saved_ptr;
             result
         } else if type_num == TYPE_STRING {
-            self.decode_string(size)
+            let new_offset = self.current_ptr + size;
+            if new_offset > self.buf.len() {
+                return Err(self.invalid_db_error("string length exceeds buffer"));
+            }
+            let bytes = &self.buf[self.current_ptr..new_offset];
+            self.current_ptr = new_offset;
+            Ok(bytes)
         } else {
             Err(self.invalid_db_error(&format!("expected string, got type {type_num}")))
         }
@@ -595,10 +602,23 @@ impl<'de: 'a, 'a> de::Deserializer<'de> for &'a mut Decoder<'de> {
         visitor.visit_enum(EnumAccessor { de: self })
     }
 
+    fn deserialize_identifier<V>(self, visitor: V) -> DecodeResult<V::Value>
+    where
+        V: Visitor<'de>,
+    {
+        let (_, type_num) = self.peek_type()?;
+        if type_num == TYPE_STRING {
+            let bytes = self.read_str_as_bytes()?;
+            visitor.visit_borrowed_bytes(bytes)
+        } else {
+            self.decode_any(visitor)
+        }
+    }
+
     forward_to_deserialize_any! {
         bool i8 i16 i32 i64 i128 u8 u16 u32 u64 u128 f32 f64 char str string
         bytes byte_buf unit unit_struct newtype_struct seq tuple
-        tuple_struct map struct identifier
+        tuple_struct map struct
     }
 }
 

src/reader.rs

@@ -450,9 +450,8 @@ impl<'de, S: AsRef<[u8]>> Reader<S> {
         let val = match self.metadata.record_size {
             24 => {
                 let offset = base_offset + index * 3;
-                (buf[offset] as usize) << 16
-                    | (buf[offset + 1] as usize) << 8
-                    | buf[offset + 2] as usize
+                let bytes = &buf[offset..offset + 3];
+                (bytes[0] as usize) << 16 | (bytes[1] as usize) << 8 | bytes[2] as usize
             }
             28 => {
                 let middle = if index != 0 {
@@ -461,17 +460,19 @@ impl<'de, S: AsRef<[u8]>> Reader<S> {
                     (buf[base_offset + 3] & 0xF0) >> 4
                 };
                 let offset = base_offset + index * 4;
+                let bytes = &buf[offset..offset + 3];
                 (middle as usize) << 24
-                    | (buf[offset] as usize) << 16
-                    | (buf[offset + 1] as usize) << 8
-                    | buf[offset + 2] as usize
+                    | (bytes[0] as usize) << 16
+                    | (bytes[1] as usize) << 8
+                    | bytes[2] as usize
             }
             32 => {
                 let offset = base_offset + index * 4;
-                (buf[offset] as usize) << 24
-                    | (buf[offset + 1] as usize) << 16
-                    | (buf[offset + 2] as usize) << 8
-                    | buf[offset + 3] as usize
+                let bytes = &buf[offset..offset + 4];
+                (bytes[0] as usize) << 24
+                    | (bytes[1] as usize) << 16
+                    | (bytes[2] as usize) << 8
+                    | bytes[3] as usize
             }
             s => {
                 return Err(MaxMindDbError::invalid_database(format!(

src/reader_test.rs

@@ -1212,6 +1212,7 @@ fn test_ignored_any() {
     let _ = env_logger::try_init();
 
     // Struct that only reads some fields, ignoring others via IgnoredAny
+    #[allow(dead_code)]
     #[derive(Deserialize, Debug)]
     struct PartialRead {
         utf8_string: String,

src/result.rs

@@ -252,9 +252,10 @@ impl<'a, S: AsRef<[u8]>> LookupResult<'a, S> {
                     let size = decoder.consume_map_header().map_err(with_path)?;
 
                     let mut found = false;
+                    let key_bytes = key.as_bytes();
                     for _ in 0..size {
-                        let k = decoder.read_string().map_err(with_path)?;
-                        if k == key {
+                        let k = decoder.read_str_as_bytes().map_err(with_path)?;
+                        if k == key_bytes {
                             found = true;
                             break;
                         } else {