From 00187950437a10952b82353405e5ba4b4515fb29 Mon Sep 17 00:00:00 2001
From: Luke Shumaker <lukeshu@lukeshu.com>
Date: Thu, 16 Feb 2023 19:06:46 -0700
Subject: reencode: Don't normalize the capitalization of \uXXXX hex escapes

---
 internal/jsonstring/encode_string.go | 60 ++++++++++++++++++++++++++----------
 1 file changed, 44 insertions(+), 16 deletions(-)

(limited to 'internal/jsonstring/encode_string.go')

diff --git a/internal/jsonstring/encode_string.go b/internal/jsonstring/encode_string.go
index 2488cb2..1416b3e 100644
--- a/internal/jsonstring/encode_string.go
+++ b/internal/jsonstring/encode_string.go
@@ -31,22 +31,49 @@ type BackslashEscapeMode uint8
 const (
 	BackslashEscapeNone BackslashEscapeMode = iota
 	BackslashEscapeShort
-	BackslashEscapeUnicode
 	BackslashEscapeRawByte
+
+	// It is significant to the implementation that if X=binary-0
+	// and x=binary-1, then these "BackslashEscapeUnicode"
+	// constants are counting in-order from 0 to 15.
+
+	BackslashEscapeUnicodeXXXX
+	BackslashEscapeUnicodeXXXx
+	BackslashEscapeUnicodeXXxX
+	BackslashEscapeUnicodeXXxx
+	BackslashEscapeUnicodeXxXX
+	BackslashEscapeUnicodeXxXx
+	BackslashEscapeUnicodeXxxX
+	BackslashEscapeUnicodeXxxx
+	BackslashEscapeUnicodexXXX
+	BackslashEscapeUnicodexXXx
+	BackslashEscapeUnicodexXxX
+	BackslashEscapeUnicodexXxx
+	BackslashEscapeUnicodexxXX
+	BackslashEscapeUnicodexxXx
+	BackslashEscapeUnicodexxxX
+	BackslashEscapeUnicodexxxx
+
+	BackslashEscapeUnicodeMin = BackslashEscapeUnicodeXXXX
+	BackslashEscapeUnicodeMax = BackslashEscapeUnicodexxxx
+
+	BackslashEscapeUnicode = BackslashEscapeUnicodexxxx // back-compat
 )
 
 // BackslashEscaper is describe in the main lowmemjson package docs.
 type BackslashEscaper = func(rune, BackslashEscapeMode) BackslashEscapeMode
 
-func WriteStringUnicodeEscape(w io.Writer, c rune) error {
-	const alphabet = "0123456789abcdef"
+func WriteStringUnicodeEscape(w io.Writer, c rune, mode BackslashEscapeMode) error {
+	const alphabet = "0123456789ABCDEF"
+	_mode := byte(mode - BackslashEscapeUnicodeMin)
 	buf := [6]byte{
 		'\\',
 		'u',
-		alphabet[(c>>12)&0xf],
-		alphabet[(c>>8)&0xf],
-		alphabet[(c>>4)&0xf],
-		alphabet[(c>>0)&0xf],
+		// The 0b0010_0000 bit is the ASCII "lowercase bit".
+		alphabet[(c>>12)&0xf] | ((_mode << 2) & 0b0010_0000),
+		alphabet[(c>>8)&0xf] | ((_mode << 3) & 0b0010_0000),
+		alphabet[(c>>4)&0xf] | ((_mode << 4) & 0b0010_0000),
+		alphabet[(c>>0)&0xf] | ((_mode << 5) & 0b0010_0000),
 	}
 	_, err := noescape.Write(w, buf[:])
 	return err
@@ -84,7 +111,7 @@ func WriteStringChar(w fastio.AllWriter, c rune, escape BackslashEscapeMode) err
 			case '\b', '\f', '\n', '\r', '\t': // short-escape if possible
 				return writeStringShortEscape(w, c)
 			default:
-				return WriteStringUnicodeEscape(w, c)
+				return WriteStringUnicodeEscape(w, c, BackslashEscapeUnicode)
 			}
 		case c == '"' || c == '\\': // override, gotta escape these
 			return writeStringShortEscape(w, c)
@@ -100,14 +127,6 @@ func WriteStringChar(w fastio.AllWriter, c rune, escape BackslashEscapeMode) err
 			_, err := w.WriteRune(c)
 			return err
 		}
-	case BackslashEscapeUnicode:
-		switch {
-		case c > 0xFFFF: // override, can't escape these (TODO: unless we use UTF-16 surrogates?)
-			_, err := w.WriteRune(c)
-			return err
-		default: // obey
-			return WriteStringUnicodeEscape(w, c)
-		}
 	case BackslashEscapeRawByte:
 		switch {
 		case c < utf8.RuneSelf:
@@ -118,6 +137,15 @@ func WriteStringChar(w fastio.AllWriter, c rune, escape BackslashEscapeMode) err
 			return w.WriteByte(byte(c))
 		}
 	default:
+		if BackslashEscapeUnicodeMin <= escape && escape <= BackslashEscapeUnicodeMax {
+			switch {
+			case c > 0xFFFF: // override, can't escape these (TODO: unless we use UTF-16 surrogates?)
+				_, err := w.WriteRune(c)
+				return err
+			default: // obey
+				return WriteStringUnicodeEscape(w, c, escape)
+			}
+		}
 		panic(fmt.Errorf("escaper returned an invalid escape mode=%d", escape))
 	}
 }
-- 
cgit v1.2.3-2-g168b