Failsafe when splitting surrogate pairs - fixes #248

src/structs/ContentString.js

@@ -51,6 +51,17 @@ export class ContentString {
   splice (offset) {
     const right = new ContentString(this.str.slice(offset))
     this.str = this.str.slice(0, offset)
+
+    // Prevent encoding invalid documents because of splitting of surrogate pairs: https://github.com/yjs/yjs/issues/248
+    const firstCharCode = this.str.charCodeAt(offset - 1)
+    if (firstCharCode >= 0xD800 && firstCharCode <= 0xDBFF) {
+      // Last character of the left split is the start of a surrogate utf16/ucs2 pair.
+      // We don't support splitting of surrogate pairs because this may lead to invalid documents.
+      // Replace the invalid character with a unicode replacement character (<28> / U+FFFD)
+      this.str = this.str.slice(0, offset - 1) + '<27>'
+      // replace right as well
+      right.str = '<27>' + right.str.slice(1)
+    }
     return right
   }