Fix string.drop_start on JavaScript with multi-byte characters

The JavaScript target reused unsafe_byte_slice with Erlang byte
offsets, but JS strings are indexed by UTF-16 code units, so
drop_start returned wrong results for multi-byte input. Add a
target-specific implementation that slices off the grapheme
prefix using code-unit lengths.

Author: jtdowney

CHANGELOG.md

@@ -1,5 +1,10 @@
 # Changelog
 
+## Unreleased
+
+- Fixed a bug where `string.drop_start` would return incorrect results on
+  JavaScript when the string contained multi-byte characters.
+
 ## v1.0.4 - 2026-05-30
 
 - Fix a bug where dicts and sets with hash collisions but equal entries would

src/gleam/string.gleam

@@ -200,7 +200,6 @@ pub fn slice(from string: String, at_index idx: Int, length len: Int) -> String
 fn grapheme_slice(string: String, index: Int, length: Int) -> String
 
 @external(erlang, "binary", "part")
-@external(javascript, "../gleam_stdlib.mjs", "string_byte_slice")
 fn unsafe_byte_slice(string: String, index: Int, length: Int) -> String
 
 /// Drops contents of the first `String` that occur before the second `String`.
@@ -227,6 +226,7 @@ pub fn crop(from string: String, before substring: String) -> String
 /// assert drop_start(from: "The Lone Gunmen", up_to: 2) == "e Lone Gunmen"
 /// ```
 ///
+@external(javascript, "../gleam_stdlib.mjs", "string_drop_start")
 pub fn drop_start(from string: String, up_to num_graphemes: Int) -> String {
   case num_graphemes <= 0 {
     True -> string

src/gleam_stdlib.mjs

@@ -199,8 +199,13 @@ export function length(data) {
   return data.length;
 }
 
-export function string_byte_slice(string, index, length) {
-  return string.slice(index, index + length);
+export function string_drop_start(string, num_graphemes) {
+  if (num_graphemes <= 0) {
+    return string;
+  }
+
+  const prefix = string_grapheme_slice(string, 0, num_graphemes);
+  return string.slice(prefix.length);
 }
 
 export function string_grapheme_slice(string, idx, len) {

test/gleam/string_test.gleam

@@ -471,6 +471,19 @@ pub fn drop_start_3499_test() {
   assert string.drop_start("\r]", 1) == "]"
 }
 
+pub fn drop_start_multibyte_test() {
+  // https://github.com/gleam-lang/stdlib/issues/924
+  assert string.drop_start("广州abcdefghijklmn", 0) == "广州abcdefghijklmn"
+  assert string.drop_start("广州abcdefghijklmn", 1) == "州abcdefghijklmn"
+  assert string.drop_start("广州abcdefghijklmn", 2) == "abcdefghijklmn"
+  assert string.drop_start("广州abcdefghijklmn", 3) == "bcdefghijklmn"
+}
+
+pub fn drop_start_grapheme_cluster_test() {
+  assert string.drop_start("👶🏿abc", 1) == "abc"
+  assert string.drop_start("e\u{0301}abc", 1) == "abc"
+}
+
 pub fn drop_end_basic_test() {
   assert string.drop_end("gleam", up_to: 2) == "gle"
 }