Review PR 1109 string character offset handling

packages/openapi-generator/package.json

@@ -22,7 +22,7 @@
     "test:target": "c8 --all --src src node --require @swc-node/register"
   },
   "dependencies": {
-    "@swc/core": "1.5.7",
+    "@swc/core": "1.10.6",
     "cmd-ts": "0.13.0",
     "comment-parser": "1.4.1",
     "fp-ts": "2.16.9",
@@ -39,8 +39,8 @@
     "typescript": "4.7.4"
   },
   "optionalDependencies": {
-    "@swc/core-linux-x64-gnu": "1.5.7",
-    "@swc/core-darwin-arm64": "1.5.7"
+    "@swc/core-linux-x64-gnu": "1.10.6",
+    "@swc/core-darwin-arm64": "1.10.6"
   },
   "publishConfig": {
     "access": "public"

packages/openapi-generator/src/comments.ts

@@ -1,13 +1,46 @@
 import { parse as parseComment, Block } from 'comment-parser';
 import { Schema } from './ir';
 
+/**
+ * Convert a UTF-8 byte offset to a JavaScript string character offset.
+ * SWC (written in Rust) uses byte offsets, but JavaScript strings use
+ * UTF-16 code unit offsets. This function handles the conversion by
+ * iterating through the string and accumulating byte lengths.
+ *
+ * @param str The source string
+ * @param byteOffset The byte offset to convert
+ * @returns The corresponding character offset
+ */
+function byteOffsetToCharOffset(str: string, byteOffset: number): number {
+  let charCount = 0;
+  let byteCount = 0;
+
+  for (const char of str) {
+    const charBytes = Buffer.byteLength(char, 'utf8');
+    if (byteCount + charBytes > byteOffset) break;
+    byteCount += charBytes;
+    charCount++;
+  }
+
+  return charCount;
+}
+
 export function leadingComment(
   src: string,
   srcSpanStart: number,
   start: number,
   end: number,
 ): Block[] {
-  let commentString = src.slice(start - srcSpanStart, end - srcSpanStart).trim();
+  // SWC uses byte offsets, but JavaScript strings use character offsets.
+  // When there are multibyte UTF-8 characters (e.g., À, 日, 😀), we need to
+  // convert byte offsets to character offsets for correct string slicing.
+  const startByteOffset = start - srcSpanStart;
+  const endByteOffset = end - srcSpanStart;
+
+  const startCharOffset = byteOffsetToCharOffset(src, startByteOffset);
+  const endCharOffset = byteOffsetToCharOffset(src, endByteOffset);
+
+  let commentString = src.slice(startCharOffset, endCharOffset).trim();
 
   if (commentString.includes(' * ') && !/\/\*\*([\s\S]*?)\*\//.test(commentString)) {
     // The comment block seems to be JSDoc but was sliced incorrectly