Start work on optomizing tab map

We currently iterate over each character when looking for tab bytes even
though chunks keeps a bitmask that represents each tab position. This
commit is the first step in using the bitmask

Co-authored-by: Remco Smits <djsmits12@gmail.com>
Co-authored-by: Cole Miller <m@cole-miller.net>

crates/language/src/buffer.rs

@@ -28,6 +28,7 @@ use gpui::{
     App, AppContext as _, Context, Entity, EventEmitter, HighlightStyle, SharedString, StyledText,
     Task, TaskLabel, TextStyle,
 };
+
 use lsp::{LanguageServerId, NumberOrString};
 use parking_lot::Mutex;
 use schemars::JsonSchema;
@@ -485,6 +486,8 @@ pub struct Chunk<'a> {
     pub is_unnecessary: bool,
     /// Whether this chunk of text was originally a tab character.
     pub is_tab: bool,
+    /// A bitset of which characters are tabs in this string.
+    pub tabs: u128,
     /// Whether this chunk of text was originally a tab character.
     pub is_inlay: bool,
     /// Whether to underline the corresponding text range in the editor.
@@ -4579,7 +4582,7 @@ impl<'a> Iterator for BufferChunks<'a> {
         }
         self.diagnostic_endpoints = diagnostic_endpoints;
 
-        if let Some(chunk) = self.chunks.peek() {
+        if let Some((chunk, tabs)) = self.chunks.peek_tabs() {
             let chunk_start = self.range.start;
             let mut chunk_end = (self.chunks.offset() + chunk.len())
                 .min(next_capture_start)
@@ -4594,6 +4597,8 @@ impl<'a> Iterator for BufferChunks<'a> {
 
             let slice =
                 &chunk[chunk_start - self.chunks.offset()..chunk_end - self.chunks.offset()];
+            let tabs = tabs >> (chunk_start - self.chunks.offset());
+
             self.range.start = chunk_end;
             if self.range.start == self.chunks.offset() + chunk.len() {
                 self.chunks.next().unwrap();
@@ -4605,6 +4610,7 @@ impl<'a> Iterator for BufferChunks<'a> {
                 underline: self.underline,
                 diagnostic_severity: self.current_diagnostic_severity(),
                 is_unnecessary: self.current_code_is_unnecessary(),
+                tabs,
                 ..Chunk::default()
             })
         } else {

crates/language/src/buffer_tests.rs

@@ -3279,6 +3279,28 @@ fn test_contiguous_ranges() {
     );
 }
 
+#[test]
+fn test_buffer_chunks_tabs() {
+    let buffer = text::Buffer::new(0, BufferId::new(1).unwrap(), "\ta\tbc");
+    let mut iter = buffer.as_rope().chunks();
+
+    while let Some((str, tabs)) = iter.peek_tabs() {
+        dbg!(str, format!("{:b}", tabs));
+        iter.next();
+    }
+    dbg!("---");
+
+    let buffer = text::Buffer::new(0, BufferId::new(1).unwrap(), "\ta\tbc");
+    let mut iter = buffer.as_rope().chunks();
+    iter.seek(3);
+
+    while let Some((str, tabs)) = iter.peek_tabs() {
+        dbg!(str, format!("{:b}", tabs));
+        iter.next();
+    }
+    assert!(false)
+}
+
 #[gpui::test(iterations = 500)]
 fn test_trailing_whitespace_ranges(mut rng: StdRng) {
     // Generate a random multi-line string containing