From 471104f43b8a1c678dbcffc1d137f009fa881bc7 Mon Sep 17 00:00:00 2001
From: Henri Sivonen <hsivonen@hsivonen.fi>
Date: Thu, 19 Sep 2024 21:33:24 +0300
Subject: [PATCH] Restore PUA range check, add manual tests

---
 src/gb18030.rs | 34 +++++++++++++++++++++++++++++++---
 1 file changed, 31 insertions(+), 3 deletions(-)

diff --git a/src/gb18030.rs b/src/gb18030.rs
index 515d0e8..4933fac 100644
--- a/src/gb18030.rs
+++ b/src/gb18030.rs
@@ -388,8 +388,11 @@ fn gbk_encode_non_unified(bmp: u16) -> Option<(usize, usize)> {
         let offset = if other_trail < 0x3F { 0x40 } else { 0x41 };
         return Some((other_lead + (0x81 + 0x20), other_trail + offset));
     }
-    // CJK Radicals Supplement and U+9FBx ideographs in GBK_BOTTOM
-    if in_inclusive_range16(bmp, 0x2E81, 0x2ECA) || in_inclusive_range16(bmp, 0x9FB4, 0x9FBB) {
+    // CJK Radicals Supplement, PUA, and U+9FBx ideographs in GBK_BOTTOM
+    if in_inclusive_range16(bmp, 0x2E81, 0x2ECA)
+        || in_inclusive_range16(bmp, 0x9FB4, 0x9FBB)
+        || in_inclusive_range16(bmp, 0xE816, 0xE855)
+    {
         if let Some(pos) = position(&GBK_BOTTOM[21..], bmp) {
             let trail = pos + 16;
             let offset = if trail < 0x3F { 0x40 } else { 0x41 };
@@ -615,11 +618,18 @@ mod tests {
         decode_gb18030(b"\x81\x80", "\u{4E90}");
         decode_gb18030(b"\x81\xFE", "\u{4FA2}");
         decode_gb18030(b"\xFE\x40", "\u{FA0C}");
-        decode_gb18030(b"\xFE\x7E", "\u{E843}");
         decode_gb18030(b"\xFE\x7F", "\u{FFFD}\u{007F}");
         decode_gb18030(b"\xFE\x80", "\u{4723}");
         decode_gb18030(b"\xFE\xFE", "\u{E4C5}");
 
+        // Changes between GB18030-2005 and GB18030-2022
+        decode_gb18030(b"\xFE\x7E", "\u{9FB9}");
+        decode_gb18030(b"\xA6\xDD", "\u{FE14}");
+
+        // These mappings remain in place the GB18030-2005 way despite GB18030-2022
+        decode_gb18030(b"\x82\x35\x91\x32", "\u{9FB9}");
+        decode_gb18030(b"\x84\x31\x83\x30", "\u{FE14}");
+
         // The difference from the original GB18030
         decode_gb18030(b"\xA3\xA0", "\u{3000}");
         decode_gb18030(b"\xA1\xA1", "\u{3000}");
@@ -687,6 +697,15 @@ mod tests {
 
         // Edge cases
         encode_gb18030("\u{00F7}", b"\xA1\xC2");
+
+        // GB18030-2022
+        encode_gb18030("\u{9FB9}", b"\xFE\x7E");
+        encode_gb18030("\u{FE14}", b"\xA6\xDD");
+        encode_gb18030("\u{E843}", b"\xFE\x7E");
+        encode_gb18030("\u{E791}", b"\xA6\xDD");
+
+        // Non-change in GB18030-2022
+        encode_gb18030("\u{E817}", b"\xFE\x52");
     }
 
     #[test]
@@ -729,6 +748,15 @@ mod tests {
 
         // Edge cases
         encode_gbk("\u{00F7}", b"\xA1\xC2");
+
+        // GB18030-2022
+        encode_gb18030("\u{9FB9}", b"\xFE\x7E");
+        encode_gb18030("\u{FE14}", b"\xA6\xDD");
+        encode_gb18030("\u{E843}", b"\xFE\x7E");
+        encode_gb18030("\u{E791}", b"\xA6\xDD");
+
+        // Non-change in GB18030-2022
+        encode_gb18030("\u{E817}", b"\xFE\x52");
     }
 
     #[test]