v1.3.0

[minirang]

Description

• String indexing/slicing/length() are now UTF-8 codepoint-based, not byte-based.
  Previously "안녕하세요"[1] sliced into the middle of a multi-byte UTF-8 sequence and
  produced corrupted output — any non-ASCII string indexing was broken. Fixed via a small
  UTF-8 boundary scanner (engine/interpreter/Utf8.h); contains/starts_with/ends_with/
  + were already byte-safe (UTF-8 is self-synchronizing) and needed no change. The regex
  engine remains byte-oriented by design/scope — see docs/IMPLEMENTATION_NOTES.md sections
  14-15 for the exact boundaries and reasoning.

• to_number/to_str/to_boolean are now core builtins — no use DLC:convert needed.
  use DLC:convert still works (harmlessly re-registers the same functions).

• Fixed the Makefile using Windows-only batch syntax (@if exist ... del) for clean/wasm,
  which failed with a shell syntax error on Linux/macOS (and in CI). Restored portable
  rm -f/mkdir -p. Also fixed wasm-clean deleting the whole npm/dist/ directory,
  including the hand-written cuffscript.d.ts, instead of just the compiled outputs.

• Restored docs/EXTENDING.md, which the changelog referenced but was missing from the repo.

---

This discussion was created from the release v1.3.0.