perf: use byte_slice_unchecked (#149)

* perf: faster rope comparison

* perf: faster `WithIndices`

* chore: clippy

Author: h-a-n-a

src/helpers.rs

@@ -1260,6 +1260,10 @@ pub trait SourceText<'a>: Default + Clone + ToString {
   /// Returns a slice of the text specified by the byte range.
   fn byte_slice(&self, range: Range<usize>) -> Self;
 
+  /// Returns a slice of the text specified by the byte range without bounds checking.
+  #[allow(unsafe_code)]
+  unsafe fn byte_slice_unchecked(&self, range: Range<usize>) -> Self;
+
   /// Returns true if the text is empty.
   fn is_empty(&self) -> bool;
 
@@ -1293,6 +1297,11 @@ impl<'a> SourceText<'a> for Rope<'a> {
     self.byte_slice(range)
   }
 
+  #[allow(unsafe_code)]
+  unsafe fn byte_slice_unchecked(&self, range: Range<usize>) -> Self {
+    self.byte_slice_unchecked(range)
+  }
+
   #[inline]
   fn is_empty(&self) -> bool {
     self.is_empty()
@@ -1330,6 +1339,11 @@ impl<'a> SourceText<'a> for &'a str {
     self.get(range).unwrap_or_default()
   }
 
+  #[allow(unsafe_code)]
+  unsafe fn byte_slice_unchecked(&self, range: Range<usize>) -> Self {
+    self.get_unchecked(range)
+  }
+
   #[inline]
   fn is_empty(&self) -> bool {
     (*self).is_empty()

src/rope.rs

@@ -2,7 +2,6 @@
 
 use std::{
   borrow::Cow,
-  cell::RefCell,
   collections::VecDeque,
   hash::Hash,
   ops::{Bound, RangeBounds},
@@ -734,6 +733,18 @@ impl PartialEq<Rope<'_>> for Rope<'_> {
       return false;
     }
 
+    if let (
+      Rope {
+        repr: Repr::Light(s),
+      },
+      Rope {
+        repr: Repr::Light(other),
+      },
+    ) = (self, other)
+    {
+      return s == other;
+    }
+
     let chunks = match &self.repr {
       Repr::Light(s) => &[(*s, 0)][..],
       Repr::Full(data) => &data[..],
@@ -743,22 +754,63 @@ impl PartialEq<Rope<'_>> for Rope<'_> {
       Repr::Full(data) => &data[..],
     };
 
-    let mut cur = 0;
-    let other_chunk_index = RefCell::new(0);
-    let mut other_chunk_byte_index = 0;
-    let other_chunk = || other_chunks[*other_chunk_index.borrow()].0.as_bytes();
-    for (chunk, start_pos) in chunks.iter() {
-      let chunk = chunk.as_bytes();
-      while (cur - start_pos) < chunk.len() {
-        if other_chunk_byte_index >= other_chunk().len() {
-          other_chunk_byte_index = 0;
-          *other_chunk_index.borrow_mut() += 1;
+    let total_bytes = self.len();
+    let mut byte_idx = 0;
+
+    let mut chunks_idx = 0;
+    let mut in_chunk_byte_idx = 0;
+
+    let mut other_chunks_idx = 0;
+    let mut in_other_chunk_byte_idx = 0;
+
+    loop {
+      if byte_idx == total_bytes {
+        break;
+      }
+
+      let &(chunk, _) = &chunks[chunks_idx];
+      let chunk_len = chunk.len();
+      let &(other_chunk, _) = &other_chunks[other_chunks_idx];
+      let other_chunk_len = other_chunk.len();
+
+      let chunk_remaining = chunk_len - in_chunk_byte_idx;
+      let other_chunk_remaining = other_chunk_len - in_other_chunk_byte_idx;
+
+      match chunk_remaining.cmp(&other_chunk_remaining) {
+        std::cmp::Ordering::Less => {
+          if other_chunk
+            [in_other_chunk_byte_idx..in_other_chunk_byte_idx + chunk_remaining]
+            != chunk[in_chunk_byte_idx..]
+          {
+            return false;
+          }
+          in_other_chunk_byte_idx += chunk_remaining;
+          chunks_idx += 1;
+          in_chunk_byte_idx = 0;
+          byte_idx += chunk_remaining;
         }
-        if chunk[cur - start_pos] == other_chunk()[other_chunk_byte_index] {
-          cur += 1;
-          other_chunk_byte_index += 1;
-        } else {
-          return false;
+        std::cmp::Ordering::Equal => {
+          if chunk[in_chunk_byte_idx..]
+            != other_chunk[in_other_chunk_byte_idx..]
+          {
+            return false;
+          }
+          chunks_idx += 1;
+          other_chunks_idx += 1;
+          in_chunk_byte_idx = 0;
+          in_other_chunk_byte_idx = 0;
+          byte_idx += chunk_remaining;
+        }
+        std::cmp::Ordering::Greater => {
+          if chunk[in_chunk_byte_idx..in_chunk_byte_idx + other_chunk_remaining]
+            != other_chunk[in_other_chunk_byte_idx..]
+          {
+            return false;
+          }
+          in_chunk_byte_idx += other_chunk_remaining;
+          other_chunks_idx += 1;
+          in_other_chunk_byte_idx = 0;
+          byte_idx += other_chunk_remaining;
         }
       }
     }
@@ -1061,6 +1113,16 @@ mod tests {
     b.add("fghi");
 
     assert_eq!(a, b);
+
+    let mut a = Rope::new();
+    a.add("abc");
+
+    let mut b = Rope::new();
+    b.add("a");
+    b.add("b");
+    b.add("c");
+
+    assert_eq!(a, b);
   }
 
   #[test]

src/with_indices.rs

@@ -39,7 +39,14 @@ where
     let str_len = self.line.len();
     let start = *indices_indexes.get(start_index).unwrap_or(&str_len);
     let end = *indices_indexes.get(end_index).unwrap_or(&str_len);
-    self.line.byte_slice(start..end)
+
+    #[allow(unsafe_code)]
+    unsafe {
+      // SAFETY: Since `indices` iterates over the `CharIndices` of `self`, we can guarantee
+      // that the indices obtained from it will always be within the bounds of `self` and they
+      // will always lie on UTF-8 sequence boundaries.
+      self.line.byte_slice_unchecked(start..end)
+    }
   }
 }