From 2bf8bce04a271c0adefd420db9e7c2e07a93c761 Mon Sep 17 00:00:00 2001
From: Yushion Yang <2685310462@qq.com>
Date: Fri, 23 Oct 2026 18:00:36 +0800
Subject: [PATCH] Keep surrogate pairs intact when wrapping help text

---
 CHANGELOG.md      |  5 +++++
 lib/textwrap.js   | 15 +++++++++++++--
 test/test_port.js | 35 +++++++++++++++++++++++++++++++++++
 3 files changed, 53 insertions(+), 2 deletions(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index ad9424d..d923b06 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -6,6 +6,11 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 
+## [Unreleased]
+### Fixed
+- Kept UTF-16 surrogate pairs together when wrapping long help text.
+
+
 ## [3.0.3] - 2026-10-16
 ### Security
 - Fixed ReDoS in `int` type conversion. Not critical for typical CLI use:
diff --git a/lib/textwrap.js b/lib/textwrap.js
index e3248f8..9027bc4 100644
--- a/lib/textwrap.js
+++ b/lib/textwrap.js
@@ -164,8 +164,19 @@ class TextWrapper {
         // If we're allowed to break long words, then do so: put as much
         // of the next chunk onto the current line as will fit.
         if (this.break_long_words) {
-            cur_line.push(reversed_chunks[reversed_chunks.length - 1].slice(0, space_left))
-            reversed_chunks[reversed_chunks.length - 1] = reversed_chunks[reversed_chunks.length - 1].slice(space_left)
+            const chunk = reversed_chunks[reversed_chunks.length - 1]
+            space_left = Math.floor(space_left)
+            const high = chunk.charCodeAt(space_left - 1)
+            const low = chunk.charCodeAt(space_left)
+            if (high >= 0xd800 && high <= 0xdbff && low >= 0xdc00 && low <= 0xdfff) {
+                // Keep surrogate pairs together. Allow a whole pair on an
+                // empty line when the available width is only one.
+                space_left += space_left === 1 && cur_len === 0 ? 1 : -1
+            }
+            if (space_left === 0) return
+            cur_line.push(chunk.slice(0, space_left))
+            reversed_chunks[reversed_chunks.length - 1] = chunk.slice(space_left)
+            if (!reversed_chunks[reversed_chunks.length - 1]) reversed_chunks.pop()
 
         // Otherwise, we have to preserve the long word intact.  Only add
         // it to the current line if there's nothing already there --
diff --git a/test/test_port.js b/test/test_port.js
index af1925b..24aa673 100644
--- a/test/test_port.js
+++ b/test/test_port.js
@@ -6,6 +6,7 @@
 const assert = require('assert')
 const { describe, it } = require('node:test')
 const argparse = require('../')
+const textwrap = require('../lib/textwrap')
 
 
 // Number conversion: CPython calls the int() / float() builtins, the port
@@ -81,3 +82,37 @@ for (const type of ['int', 'float']) {
         }
     })
 }
+
+
+describe('UTF-16 line wrapping', () => {
+    for (const [text, width, expected] of [
+        ['😀😀😀', 3, ['😀', '😀', '😀']],
+        ['a😀😀b', 2, ['a', '😀', '😀', 'b']],
+        ['😀😀', 1, ['😀', '😀']],
+        ['😀😀', 1.5, ['😀', '😀']],
+        ['ab 😀😀😀', 4, ['ab', '😀😀', '😀']],
+        ['a😀b', 3, ['a😀', 'b']],
+        ['abcd', 2, ['ab', 'cd']]
+    ]) {
+        it(`keeps characters intact at width ${width}: ${text}`, () => {
+            assert.deepStrictEqual(textwrap.wrap(text, { width }), expected)
+        })
+    }
+
+    it('keeps a help description well formed', () => {
+        class NarrowFormatter extends argparse.HelpFormatter {
+            constructor (options) {
+                super({ ...options, width: 23 })
+            }
+        }
+        const parser = new argparse.ArgumentParser({
+            prog: 'demo',
+            add_help: false,
+            color: false,
+            formatter_class: NarrowFormatter,
+            description: '😀'.repeat(20)
+        })
+        assert.strictEqual(parser.format_help(),
+            'usage: demo\n\n' + '😀'.repeat(11) + '\n' + '😀'.repeat(9) + '\n')
+    })
+})