113 lines
3.8 KiB
Rust

/// TextIndex provides conversions between byte offsets (used in the core)
/// and LSP positions (line, column in UTF-16 units).
///
/// Notes:
/// - `line_starts` stores byte offsets for the start of each line.
/// - We keep a copy of the text to allow conversions without external dependencies.
/// - The LSP column is counted in UTF-16 units, excluding the end-of-line `\n`.
#[derive(Clone, Debug)]
pub struct TextIndex {
text: String,
line_starts: Vec<u32>,
}
impl TextIndex {
/// Builds the index from the file's current textual content.
pub fn new(text: &str) -> Self {
let mut line_starts = Vec::with_capacity(128);
line_starts.push(0);
for (byte, ch) in text.char_indices() {
if ch == '\n' {
// the start of the next line is the byte after the '\n'
line_starts.push((byte + 1) as u32);
}
}
Self {
text: text.to_string(),
line_starts,
}
}
/// Number of lines (0-based; empty lines count).
#[inline]
pub fn line_count(&self) -> u32 {
self.line_starts.len() as u32
}
/// Converts a byte offset (within the file) to (line, UTF-16 column) in LSP format.
///
/// For offsets exactly at end-of-line, the column will be the line's UTF-16 length.
pub fn byte_to_lsp(&self, byte: u32) -> (u32, u32) {
let byte = byte.min(self.text.len() as u32);
let line = match self.line_starts.binary_search(&byte) {
Ok(i) => i as u32,
Err(i) => (i.saturating_sub(1)) as u32,
};
let (line_start, line_end) = self.line_bounds(line);
let rel = byte.saturating_sub(line_start as u32) as usize;
let slice = &self.text[line_start..line_end];
let mut utf16_col: u32 = 0;
for (i, ch) in slice.char_indices() {
if i >= rel { break; }
utf16_col += ch.len_utf16() as u32;
}
(line, utf16_col)
}
/// Converts (line, UTF-16 column) to a byte offset.
///
/// - Lines outside the range are clamped to [0, last].
/// - Columns larger than the line's UTF-16 length return end-of-line.
pub fn lsp_to_byte(&self, line: u32, utf16_col: u32) -> u32 {
let line = line.min(self.line_count().saturating_sub(1));
let (line_start, line_end) = self.line_bounds(line);
let slice = &self.text[line_start..line_end];
let mut acc: u32 = 0;
for (i, ch) in slice.char_indices() {
if acc >= utf16_col {
return (line_start + i) as u32;
}
acc += ch.len_utf16() as u32;
}
// If the target column is after the last character, return end-of-line.
line_end as u32
}
#[inline]
fn line_bounds(&self, line: u32) -> (usize, usize) {
let start = *self
.line_starts
.get(line as usize)
.unwrap_or(self.line_starts.last().unwrap());
let next = self.line_starts.get(line as usize + 1).copied();
// If there is a next line, `next` points to the byte after the current line's '\n',
// so the content ends at `next - 1`. Otherwise (last line), the content ends at `text.len()`.
let end = match next {
Some(next_start) => next_start.saturating_sub(1),
None => self.text.len() as u32,
};
(start as usize, end as usize)
}
}
#[cfg(test)]
mod tests_internal {
use super::*;
#[test]
fn line_bounds_basic() {
let s = "ab\ncd\n";
let idx = TextIndex::new(s);
assert_eq!(idx.line_count(), 3);
// line 0: "ab"
assert_eq!(idx.line_bounds(0), (0, 2));
// line 1: "cd"
assert_eq!(idx.line_bounds(1), (3, 5));
// line 2: final empty line
assert_eq!(idx.line_bounds(2), (6, 6));
}
}