encoding: require unique alphabet for base32 and base64

In order for decoding to faithfully reproduce the encoded input,
the symbols must be unique (i.e., provide a bijective mapping).
Thus, reject duplicate symbols in NewEncoding.

As a minor optimization, modify WithPadding to use the decodeMap
to quickly check whether the padding character is used in O(1)
instead of O(32) or O(64).

Change-Id: I5631f6ff9335c35d59d020dc0e307e3520786fbc
Reviewed-on: https://go-review.googlesource.com/c/go/+/520335
Reviewed-by: Dmitri Shuralyov <[email protected]>
Auto-Submit: Joseph Tsai <[email protected]>
TryBot-Result: Gopher Robot <[email protected]>
Reviewed-by: Ian Lance Taylor <[email protected]>
Run-TryBot: Joseph Tsai <[email protected]>

Author: dsnet

src/encoding/base32/base32.go

@@ -20,8 +20,8 @@ import (
 // introduced for SASL GSSAPI and standardized in RFC 4648.
 // The alternate "base32hex" encoding is used in DNSSEC.
 type Encoding struct {
-	encode    [32]byte
-	decodeMap [256]byte
+	encode    [32]byte   // mapping of symbol index to symbol byte value
+	decodeMap [256]uint8 // mapping of symbol byte value to symbol index
 	padChar   rune
 }
 
@@ -45,14 +45,19 @@ const (
 		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff" +
 		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff" +
 		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff"
+	invalidIndex = '\xff'
 )
 
 const encodeStd = "ABCDEFGHIJKLMNOPQRSTUVWXYZ234567"
 const encodeHex = "0123456789ABCDEFGHIJKLMNOPQRSTUV"
 
-// NewEncoding returns a new Encoding defined by the given alphabet,
-// which must be a 32-byte string. The alphabet is treated as sequence
-// of byte values without any special treatment for multi-byte UTF-8.
+// NewEncoding returns a new padded Encoding defined by the given alphabet,
+// which must be a 32-byte string that contains unique byte values and
+// does not contain the padding character or CR / LF ('\r', '\n').
+// The alphabet is treated as a sequence of byte values
+// without any special treatment for multi-byte UTF-8.
+// The resulting Encoding uses the default padding character ('='),
+// which may be changed or disabled via WithPadding.
 func NewEncoding(encoder string) *Encoding {
 	if len(encoder) != 32 {
 		panic("encoding alphabet is not 32-bytes long")
@@ -64,7 +69,16 @@ func NewEncoding(encoder string) *Encoding {
 	copy(e.decodeMap[:], decodeMapInitialize)
 
 	for i := 0; i < len(encoder); i++ {
-		e.decodeMap[encoder[i]] = byte(i)
+		// Note: While we document that the alphabet cannot contain
+		// the padding character, we do not enforce it since we do not know
+		// if the caller intends to switch the padding from StdPadding later.
+		switch {
+		case encoder[i] == '\n' || encoder[i] == '\r':
+			panic("encoding alphabet contains newline character")
+		case e.decodeMap[encoder[i]] != invalidIndex:
+			panic("encoding alphabet includes duplicate symbols")
+		}
+		e.decodeMap[encoder[i]] = uint8(i)
 	}
 	return e
 }
@@ -85,16 +99,12 @@ var HexEncoding = NewEncoding(encodeHex)
 // Padding characters above '\x7f' are encoded as their exact byte value
 // rather than using the UTF-8 representation of the codepoint.
 func (enc Encoding) WithPadding(padding rune) *Encoding {
-	if padding < NoPadding || padding == '\r' || padding == '\n' || padding > 0xff {
+	switch {
+	case padding < NoPadding || padding == '\r' || padding == '\n' || padding > 0xff:
 		panic("invalid padding")
+	case padding != NoPadding && enc.decodeMap[byte(padding)] != invalidIndex:
+		panic("padding contained in alphabet")
 	}
-
-	for i := 0; i < len(enc.encode); i++ {
-		if rune(enc.encode[i]) == padding {
-			panic("padding contained in alphabet")
-		}
-	}
-
 	enc.padChar = padding
 	return &enc
 }

src/encoding/base64/base64.go

@@ -22,8 +22,8 @@ import (
 // (RFC 1421).  RFC 4648 also defines an alternate encoding, which is
 // the standard encoding with - and _ substituted for + and /.
 type Encoding struct {
-	encode    [64]byte
-	decodeMap [256]byte
+	encode    [64]byte   // mapping of symbol index to symbol byte value
+	decodeMap [256]uint8 // mapping of symbol byte value to symbol index
 	padChar   rune
 	strict    bool
 }
@@ -48,34 +48,40 @@ const (
 		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff" +
 		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff" +
 		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff"
+	invalidIndex = '\xff'
 )
 
 const encodeStd = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/"
 const encodeURL = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789-_"
 
 // NewEncoding returns a new padded Encoding defined by the given alphabet,
-// which must be a 64-byte string that does not contain the padding character
-// or CR / LF ('\r', '\n'). The alphabet is treated as sequence of byte values
+// which must be a 64-byte string that contains unique byte values and
+// does not contain the padding character or CR / LF ('\r', '\n').
+// The alphabet is treated as a sequence of byte values
 // without any special treatment for multi-byte UTF-8.
 // The resulting Encoding uses the default padding character ('='),
 // which may be changed or disabled via WithPadding.
 func NewEncoding(encoder string) *Encoding {
 	if len(encoder) != 64 {
 		panic("encoding alphabet is not 64-bytes long")
 	}
-	for i := 0; i < len(encoder); i++ {
-		if encoder[i] == '\n' || encoder[i] == '\r' {
-			panic("encoding alphabet contains newline character")
-		}
-	}
 
 	e := new(Encoding)
 	e.padChar = StdPadding
 	copy(e.encode[:], encoder)
 	copy(e.decodeMap[:], decodeMapInitialize)
 
 	for i := 0; i < len(encoder); i++ {
-		e.decodeMap[encoder[i]] = byte(i)
+		// Note: While we document that the alphabet cannot contain
+		// the padding character, we do not enforce it since we do not know
+		// if the caller intends to switch the padding from StdPadding later.
+		switch {
+		case encoder[i] == '\n' || encoder[i] == '\r':
+			panic("encoding alphabet contains newline character")
+		case e.decodeMap[encoder[i]] != invalidIndex:
+			panic("encoding alphabet includes duplicate symbols")
+		}
+		e.decodeMap[encoder[i]] = uint8(i)
 	}
 	return e
 }
@@ -88,16 +94,12 @@ func NewEncoding(encoder string) *Encoding {
 // Padding characters above '\x7f' are encoded as their exact byte value
 // rather than using the UTF-8 representation of the codepoint.
 func (enc Encoding) WithPadding(padding rune) *Encoding {
-	if padding < NoPadding || padding == '\r' || padding == '\n' || padding > 0xff {
+	switch {
+	case padding < NoPadding || padding == '\r' || padding == '\n' || padding > 0xff:
 		panic("invalid padding")
+	case padding != NoPadding && enc.decodeMap[byte(padding)] != invalidIndex:
+		panic("padding contained in alphabet")
 	}
-
-	for i := 0; i < len(enc.encode); i++ {
-		if rune(enc.encode[i]) == padding {
-			panic("padding contained in alphabet")
-		}
-	}
-
 	enc.padChar = padding
 	return &enc
 }