feat(lsp): properly handle UTF-16 char offsets

The LSP requires all char indices to be in UTF-16 byte offsets by
default. But Rust uses UTF-8, so the right offset has to be calculated
accordingly.

crates/stef-lsp/src/compile.rs

@@ -11,6 +11,8 @@ use stef_parser::{
 };
 use tower_lsp::lsp_types::{self as lsp, Diagnostic};
 
+use crate::utf16;
+
 pub fn compile(schema: &str) -> std::result::Result<Schema<'_>, Diagnostic> {
     stef_parser::Schema::parse(schema, None).map_err(|e| match &e.cause {
         ParseSchemaCause::Parser(_) => {
@@ -187,13 +189,10 @@ fn get_range(schema: &str, location: Range<usize>) -> lsp::Range {
     let start_char = schema[..location.start]
         .lines()
         .last()
-        .map_or(0, |line| line.chars().count());
+        .map_or(0, utf16::len);
 
     let end_line = schema[..location.end].lines().count().saturating_sub(1);
-    let end_char = schema[..location.end]
-        .lines()
-        .last()
-        .map_or(0, |line| line.chars().count());
+    let end_char = schema[..location.end].lines().last().map_or(0, utf16::len);
 
     lsp::Range::new(
         lsp::Position::new(start_line as u32, start_char as u32),

crates/stef-lsp/src/main.rs

@@ -1,7 +1,5 @@
 #![allow(missing_docs)]
 
-mod compile;
-
 use std::collections::HashMap;
 
 use ouroboros::self_referencing;
@@ -20,6 +18,9 @@ use tower_lsp::{
 use tracing::debug;
 use tracing_subscriber::EnvFilter;
 
+mod compile;
+mod utf16;
+
 #[derive(Debug)]
 struct Backend {
     client: Client,

crates/stef-lsp/src/utf16.rs

@@ -0,0 +1,10 @@
+//! Utilities for handling UTF-16 encoding of strings.
+//!
+//! As the LSP by default uses UTF-16 and Rust strings are encoded as UTF-8, special handling is
+//! required to calculate the correct information. Mostly these are adjustments to character
+//! offsets.
+
+/// Get the UTF-16 byte count for the code line.
+pub fn len(line: &str) -> usize {
+    line.encode_utf16().count()
+}