perf: to_writer use utf8 offset and WithIndices use OnceCell (#62)

* perf: to_writer use utf8 offset

* perf: WithIndices use OnceCell

src/replace_source.rs

@@ -45,8 +45,6 @@ pub struct ReplaceSource<T> {
 struct Replacement {
   start: u32,
   end: u32,
-  char_start: OnceCell<u32>,
-  char_end: OnceCell<u32>,
   content: String,
   name: Option<String>,
 }
@@ -79,26 +77,10 @@ impl Replacement {
     Self {
       start,
       end,
-      char_start: OnceCell::new(),
-      char_end: OnceCell::new(),
       content,
       name,
     }
   }
-
-  pub fn char_start(&self, inner_source_code: &str) -> u32 {
-    *self.char_start.get_or_init(|| {
-      str_indices::chars::from_byte_idx(inner_source_code, self.start as usize)
-        as u32
-    })
-  }
-
-  pub fn char_end(&self, inner_source_code: &str) -> u32 {
-    *self.char_end.get_or_init(|| {
-      str_indices::chars::from_byte_idx(inner_source_code, self.end as usize)
-        as u32
-    })
-  }
 }
 
 impl<T> ReplaceSource<T> {
@@ -158,30 +140,26 @@ impl<T: Source + Hash + PartialEq + Eq + 'static> Source for ReplaceSource<T> {
     self.sort_replacement();
 
     let inner_source_code = self.get_inner_source_code();
-    let inner_source_code_with_indices = WithIndices::new(inner_source_code);
 
     // mut_string_push_str is faster that vec join
     // concatenate strings benchmark, see https://github.com/hoodie/concatenation_benchmarks-rs
     let mut source_code = String::new();
     let mut inner_pos = 0;
     for replacement in self.replacements.lock().iter() {
-      if inner_pos < replacement.char_start(inner_source_code) {
-        let end_pos = (replacement.char_start(inner_source_code) as usize)
-          .min(inner_source_code.len());
-        source_code.push_str(
-          inner_source_code_with_indices.substring(inner_pos as usize, end_pos),
-        );
+      if inner_pos < replacement.start {
+        let end_pos = (replacement.start as usize).min(inner_source_code.len());
+        source_code.push_str(&inner_source_code[inner_pos as usize..end_pos]);
       }
       source_code.push_str(&replacement.content);
       #[allow(clippy::manual_clamp)]
       {
         inner_pos = inner_pos
-          .max(replacement.char_end(inner_source_code))
+          .max(replacement.end)
           .min(inner_source_code.len() as u32);
       }
     }
     source_code.push_str(
-      inner_source_code_with_indices.substring(inner_pos as usize, usize::MAX),
+      &inner_source_code[inner_pos as usize..inner_source_code.len()],
     );
 
     source_code.into()

src/with_indices.rs

@@ -1,20 +1,17 @@
+use once_cell::sync::OnceCell;
+
 #[derive(Debug, Clone)]
 pub struct WithIndices<T: AsRef<str>> {
   /// line is a string reference
   pub line: T,
   /// the byte position of each `char` in `line` string slice .
-  pub indices_indexes: Box<[u32]>,
+  pub indices_indexes: OnceCell<Box<[u32]>>,
 }
 
 impl<T: AsRef<str>> WithIndices<T> {
   pub fn new(line: T) -> Self {
     Self {
-      indices_indexes: line
-        .as_ref()
-        .char_indices()
-        .map(|(i, _)| i as u32)
-        .collect::<Vec<_>>()
-        .into_boxed_slice(),
+      indices_indexes: OnceCell::new(),
       line,
     }
   }
@@ -26,17 +23,26 @@ impl<T: AsRef<str>> WithIndices<T> {
       return "";
     }
 
+    let indices_indexes = self.indices_indexes.get_or_init(|| {
+      self
+        .line
+        .as_ref()
+        .char_indices()
+        .map(|(i, _)| i as u32)
+        .collect::<Vec<_>>()
+        .into_boxed_slice()
+    });
+
     let str_len = self.line.as_ref().len() as u32;
-    let start = *self.indices_indexes.get(start_index).unwrap_or(&str_len);
-    let end = *self.indices_indexes.get(end_index).unwrap_or(&str_len);
+    let start =
+      indices_indexes.get(start_index).unwrap_or(&str_len).clone() as usize;
+    let end =
+      indices_indexes.get(end_index).unwrap_or(&str_len).clone() as usize;
     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
-        .as_ref()
-        .get_unchecked(start as usize..end as usize)
+      self.line.as_ref().get_unchecked(start..end)
     }
   }
 }