[ISSUE #74]🚀Implement Small String Optimization (SSO) for CheetahString to reduce heap allocations for short strings (#75)

* [ISSUE #74]🚀Implement Small String Optimization (SSO) for CheetahString to reduce heap allocations for short strings

* [ISSUE #74]🚀Implement Small String Optimization (SSO) for CheetahString to reduce heap allocations for short strings

* [ISSUE #74]🚀Implement Small String Optimization (SSO) for CheetahString to reduce heap allocations for short strings

Author: mxsm

src/cheetah_string.rs

@@ -8,8 +8,6 @@ use std::ops::Deref;
 use std::str::FromStr;
 use std::sync::Arc;
 
-pub const EMPTY_STRING: &str = "";
-
 #[derive(Clone)]
 #[repr(transparent)]
 pub struct CheetahString {
@@ -19,7 +17,10 @@ pub struct CheetahString {
 impl Default for CheetahString {
     fn default() -> Self {
         CheetahString {
-            inner: InnerString::Empty,
+            inner: InnerString::Inline {
+                len: 0,
+                data: [0; INLINE_CAPACITY],
+            },
         }
     }
 }
@@ -176,11 +177,17 @@ impl From<CheetahString> for String {
     fn from(s: CheetahString) -> Self {
         match s {
             CheetahString {
-                inner: InnerString::ArcString(s),
-            } => s.as_ref().clone(),
+                inner: InnerString::Inline { len, data },
+            } => {
+                // SAFETY: Inline strings are always valid UTF-8
+                unsafe { String::from_utf8_unchecked(data[..len as usize].to_vec()) }
+            }
             CheetahString {
                 inner: InnerString::StaticStr(s),
             } => s.to_string(),
+            CheetahString {
+                inner: InnerString::ArcString(s),
+            } => s.as_ref().clone(),
             CheetahString {
                 inner: InnerString::ArcVecString(s),
             } => {
@@ -194,9 +201,6 @@ impl From<CheetahString> for String {
                 // SAFETY: Bytes variant should only be created from valid UTF-8 sources
                 unsafe { String::from_utf8_unchecked(b.to_vec()) }
             }
-            CheetahString {
-                inner: InnerString::Empty,
-            } => String::new(),
         }
     }
 }
@@ -242,7 +246,10 @@ impl CheetahString {
     #[inline]
     pub const fn empty() -> Self {
         CheetahString {
-            inner: InnerString::Empty,
+            inner: InnerString::Inline {
+                len: 0,
+                data: [0; INLINE_CAPACITY],
+            },
         }
     }
 
@@ -323,15 +330,41 @@ impl CheetahString {
 
     #[inline]
     pub fn from_slice(s: &str) -> Self {
-        CheetahString {
-            inner: InnerString::ArcString(Arc::new(s.to_owned())),
+        if s.len() <= INLINE_CAPACITY {
+            // Use inline storage for short strings
+            let mut data = [0u8; INLINE_CAPACITY];
+            data[..s.len()].copy_from_slice(s.as_bytes());
+            CheetahString {
+                inner: InnerString::Inline {
+                    len: s.len() as u8,
+                    data,
+                },
+            }
+        } else {
+            // Use Arc for long strings
+            CheetahString {
+                inner: InnerString::ArcString(Arc::new(s.to_owned())),
+            }
         }
     }
 
     #[inline]
     pub fn from_string(s: String) -> Self {
-        CheetahString {
-            inner: InnerString::ArcString(Arc::new(s)),
+        if s.len() <= INLINE_CAPACITY {
+            // Use inline storage for short strings
+            let mut data = [0u8; INLINE_CAPACITY];
+            data[..s.len()].copy_from_slice(s.as_bytes());
+            CheetahString {
+                inner: InnerString::Inline {
+                    len: s.len() as u8,
+                    data,
+                },
+            }
+        } else {
+            // Use Arc for long strings
+            CheetahString {
+                inner: InnerString::ArcString(Arc::new(s)),
+            }
         }
     }
     #[inline]
@@ -352,8 +385,13 @@ impl CheetahString {
     #[inline]
     pub fn as_str(&self) -> &str {
         match &self.inner {
-            InnerString::ArcString(s) => s.as_str(),
+            InnerString::Inline { len, data } => {
+                // SAFETY: Inline strings are only created from valid UTF-8 sources.
+                // The data is always valid UTF-8 up to len bytes.
+                unsafe { std::str::from_utf8_unchecked(&data[..*len as usize]) }
+            }
             InnerString::StaticStr(s) => s,
+            InnerString::ArcString(s) => s.as_str(),
             InnerString::ArcVecString(s) => {
                 // SAFETY: ArcVecString is only created from validated UTF-8 sources.
                 // All constructors ensure this invariant is maintained.
@@ -365,43 +403,42 @@ impl CheetahString {
                 // The from_bytes constructor ensures this invariant.
                 unsafe { std::str::from_utf8_unchecked(b.as_ref()) }
             }
-            InnerString::Empty => EMPTY_STRING,
         }
     }
 
     #[inline]
     pub fn as_bytes(&self) -> &[u8] {
         match &self.inner {
-            InnerString::ArcString(s) => s.as_bytes(),
+            InnerString::Inline { len, data } => &data[..*len as usize],
             InnerString::StaticStr(s) => s.as_bytes(),
+            InnerString::ArcString(s) => s.as_bytes(),
             InnerString::ArcVecString(s) => s.as_ref(),
             #[cfg(feature = "bytes")]
             InnerString::Bytes(b) => b.as_ref(),
-            InnerString::Empty => &[],
         }
     }
 
     #[inline]
     pub fn len(&self) -> usize {
         match &self.inner {
-            InnerString::ArcString(s) => s.len(),
+            InnerString::Inline { len, .. } => *len as usize,
             InnerString::StaticStr(s) => s.len(),
+            InnerString::ArcString(s) => s.len(),
             InnerString::ArcVecString(s) => s.len(),
             #[cfg(feature = "bytes")]
             InnerString::Bytes(b) => b.len(),
-            InnerString::Empty => 0,
         }
     }
 
     #[inline]
     pub fn is_empty(&self) -> bool {
         match &self.inner {
-            InnerString::ArcString(s) => s.is_empty(),
+            InnerString::Inline { len, .. } => *len == 0,
             InnerString::StaticStr(s) => s.is_empty(),
+            InnerString::ArcString(s) => s.is_empty(),
             InnerString::ArcVecString(s) => s.is_empty(),
             #[cfg(feature = "bytes")]
             InnerString::Bytes(b) => b.is_empty(),
-            InnerString::Empty => true,
         }
     }
 }
@@ -506,20 +543,35 @@ impl Borrow<str> for CheetahString {
     }
 }
 
+/// Maximum capacity for inline string storage (23 bytes + 1 byte for length = 24 bytes total)
+const INLINE_CAPACITY: usize = 23;
+
 /// The `InnerString` enum represents different types of string storage.
 ///
+/// This enum uses Small String Optimization (SSO) to avoid heap allocations for short strings.
+///
 /// Variants:
 ///
-/// * `ArcString(Arc<String>)` - A reference-counted string.
-/// * `StaticStr(&'static str)` - A static string slice.
+/// * `Inline` - Inline storage for strings <= 23 bytes (zero heap allocations).
+/// * `StaticStr(&'static str)` - A static string slice (zero heap allocations).
+/// * `ArcString(Arc<String>)` - A reference-counted string (one heap allocation).
+/// * `ArcVecString(Arc<Vec<u8>>)` - A reference-counted byte vector.
 /// * `Bytes(bytes::Bytes)` - A byte buffer (available when the "bytes" feature is enabled).
-/// * `Empty` - An empty string.
 #[derive(Clone)]
 pub(super) enum InnerString {
-    ArcString(Arc<String>),
+    /// Inline storage for short strings (up to 23 bytes).
+    /// Stores the length and data directly without heap allocation.
+    Inline {
+        len: u8,
+        data: [u8; INLINE_CAPACITY],
+    },
+    /// Static string slice with 'static lifetime.
     StaticStr(&'static str),
+    /// Reference-counted heap-allocated string.
+    ArcString(Arc<String>),
+    /// Reference-counted heap-allocated byte vector.
     ArcVecString(Arc<Vec<u8>>),
+    /// Bytes type integration (requires "bytes" feature).
     #[cfg(feature = "bytes")]
     Bytes(bytes::Bytes),
-    Empty,
 }

src/serde.rs

@@ -1,4 +1,4 @@
-use crate::cheetah_string::{InnerString, EMPTY_STRING};
+use crate::cheetah_string::InnerString;
 use crate::CheetahString;
 use serde::de::{Error, Unexpected, Visitor};
 use serde::{Deserialize, Deserializer, Serialize, Serializer};
@@ -9,12 +9,16 @@ impl Serialize for CheetahString {
         S: Serializer,
     {
         match &self.inner {
+            InnerString::Inline { len, data } => {
+                // Safety: InnerString::Inline guarantees that data[0..len] is valid UTF-8
+                let s = unsafe { std::str::from_utf8_unchecked(&data[..*len as usize]) };
+                serializer.serialize_str(s)
+            }
             InnerString::ArcString(s) => serializer.serialize_str(s.as_str()),
             InnerString::StaticStr(s) => serializer.serialize_str(s),
             InnerString::ArcVecString(s) => serializer.serialize_bytes(s),
             #[cfg(feature = "bytes")]
             InnerString::Bytes(bytes) => serializer.serialize_bytes(bytes.as_ref()),
-            InnerString::Empty => serializer.serialize_str(EMPTY_STRING),
         }
     }
 }