bech32: add new EncodeM and DecodeGeneric functions for bech32

In this commit, we add two new package level functions: `EncodeM`, and
`DecodeGeneric`. The new encode method is intended to allow callers to
specify that they want to use the new bech32m checksum. This should be
used when encoding segwit addresses with version 1 and beyond. The new
`DecodeGeneric` function allows a caller to decode a bech32 and bech32m
string with a single function. A new return value is added which is the
version of the returned bech32 string, which allows callers to perform
additional segwit addr validation (v1+ should use bech32m etc).

We opted to add new functions rather than modifying the existing
functions to not cause a breaking API change, as most uses in the wild
can just use the existing functions, and only taproot related logic/code
needs to worry about the new methods.

A series of tests have been added to ensure that `DecodeGeneric`
extracts the proper bech version, and we've also adopted the bech32m
tests from BIP 350.

bech32/bech32.go

@@ -123,8 +123,11 @@ func bech32Polymod(hrp string, values, checksum []byte) int {
 // and 126), otherwise the results are undefined.
 //
 // For more details on the checksum calculation, please refer to BIP 173.
-func writeBech32Checksum(hrp string, data []byte, bldr *strings.Builder) {
+func writeBech32Checksum(hrp string, data []byte, bldr *strings.Builder,
-	polymod := bech32Polymod(hrp, data, nil) ^ 1
+	version Version) {
+
+	bech32Const := int(VersionToConsts[version])
+	polymod := bech32Polymod(hrp, data, nil) ^ bech32Const
 	for i := 0; i < 6; i++ {
 		b := byte((polymod >> uint(5*(5-i))) & 31)
 
@@ -137,38 +140,47 @@ func writeBech32Checksum(hrp string, data []byte, bldr *strings.Builder) {
 
 // bech32VerifyChecksum verifies whether the bech32 string specified by the
 // provided hrp and payload data (encoded as 5 bits per element byte slice) has
-// the correct checksum suffix.
+// the correct checksum suffix. The version of bech32 used (bech32 OG, or
+// bech32m) is also returned to allow the caller to perform proper address
+// validation (segwitv0 should use bech32, v1+ should use bech32m).
 //
 // Data MUST have more than 6 elements, otherwise this function panics.
 //
 // For more details on the checksum verification, please refer to BIP 173.
-func bech32VerifyChecksum(hrp string, data []byte) bool {
+func bech32VerifyChecksum(hrp string, data []byte) (Version, bool) {
 	checksum := data[len(data)-6:]
 	values := data[:len(data)-6]
 	polymod := bech32Polymod(hrp, values, checksum)
-	return polymod == 1
+
+	// Before BIP-350, we'd always check this against a static constant of
+	// 1 to know if the checksum was computed properly. As we want to
+	// generically support decoding for bech32m as well as bech32, we'll
+	// look up the returned value and compare it to the set of defined
+	// constants.
+	bech32Version, ok := ConstsToVersion[ChecksumConst(polymod)]
+	if ok {
+		return bech32Version, true
+	}
+
+	return VersionUnknown, false
 }
 
-// DecodeNoLimit decodes a bech32 encoded string, returning the human-readable
+// DecodeNoLimit is a bech32 checksum version aware arbitrary string length
-// part and the data part excluding the checksum.  This function does NOT
+// decoder. This function will return the version of the decoded checksum
-// validate against the BIP-173 maximum length allowed for bech32 strings and
+// constant so higher level validation can be performed to ensure the correct
-// is meant for use in custom applications (such as lightning network payment
+// version of bech32 was used when encoding.
-// requests), NOT on-chain addresses.
+func decodeNoLimit(bech string) (string, []byte, Version, error) {
-//
-// Note that the returned data is 5-bit (base32) encoded and the human-readable
-// part will be lowercase.
-func DecodeNoLimit(bech string) (string, []byte, error) {
 	// The minimum allowed size of a bech32 string is 8 characters, since it
 	// needs a non-empty HRP, a separator, and a 6 character checksum.
 	if len(bech) < 8 {
-		return "", nil, ErrInvalidLength(len(bech))
+		return "", nil, VersionUnknown, ErrInvalidLength(len(bech))
 	}
 
 	// Only	ASCII characters between 33 and 126 are allowed.
 	var hasLower, hasUpper bool
 	for i := 0; i < len(bech); i++ {
 		if bech[i] < 33 || bech[i] > 126 {
-			return "", nil, ErrInvalidCharacter(bech[i])
+			return "", nil, VersionUnknown, ErrInvalidCharacter(bech[i])
 		}
 
 		// The characters must be either all lowercase or all uppercase. Testing
@@ -176,7 +188,7 @@ func DecodeNoLimit(bech string) (string, []byte, error) {
 		hasLower = hasLower || (bech[i] >= 97 && bech[i] <= 122)
 		hasUpper = hasUpper || (bech[i] >= 65 && bech[i] <= 90)
 		if hasLower && hasUpper {
-			return "", nil, ErrMixedCase{}
+			return "", nil, VersionUnknown, ErrMixedCase{}
 		}
 	}
 
@@ -191,7 +203,7 @@ func DecodeNoLimit(bech string) (string, []byte, error) {
 	// last 6 characters of the string (since checksum cannot contain '1').
 	one := strings.LastIndexByte(bech, '1')
 	if one < 1 || one+7 > len(bech) {
-		return "", nil, ErrInvalidSeparatorIndex(one)
+		return "", nil, VersionUnknown, ErrInvalidSeparatorIndex(one)
 	}
 
 	// The human-readable part is everything before the last '1'.
@@ -202,12 +214,13 @@ func DecodeNoLimit(bech string) (string, []byte, error) {
 	// 'charset'.
 	decoded, err := toBytes(data)
 	if err != nil {
-		return "", nil, err
+		return "", nil, VersionUnknown, err
 	}
 
 	// Verify if the checksum (stored inside decoded[:]) is valid, given the
 	// previously decoded hrp.
-	if !bech32VerifyChecksum(hrp, decoded) {
+	bech32Version, ok := bech32VerifyChecksum(hrp, decoded)
+	if !ok {
 		// Invalid checksum. Calculate what it should have been, so that the
 		// error contains this information.
 
@@ -215,21 +228,42 @@ func DecodeNoLimit(bech string) (string, []byte, error) {
 		actual := bech[len(bech)-6:]
 		payload := decoded[:len(decoded)-6]
 
-		// Calculate the expected checksum, given the hrp and payload data.
+		// Calculate the expected checksum, given the hrp and payload
+		// data. We'll actually compute _both_ possibly valid checksum
+		// to further aide in debugging.
 		var expectedBldr strings.Builder
 		expectedBldr.Grow(6)
-		writeBech32Checksum(hrp, payload, &expectedBldr)
+		writeBech32Checksum(hrp, payload, &expectedBldr, Version0)
-		expected := expectedBldr.String()
+		expectedVersion0 := expectedBldr.String()
+
+		var b strings.Builder
+		b.Grow(6)
+		writeBech32Checksum(hrp, payload, &expectedBldr, VersionM)
+		expectedVersionM := expectedBldr.String()
 
 		err = ErrInvalidChecksum{
-			Expected: expected,
+			Expected:  expectedVersion0,
+			ExpectedM: expectedVersionM,
 			Actual:    actual,
 		}
-		return "", nil, err
+		return "", nil, VersionUnknown, err
 	}
 
 	// We exclude the last 6 bytes, which is the checksum.
-	return hrp, decoded[:len(decoded)-6], nil
+	return hrp, decoded[:len(decoded)-6], bech32Version, nil
+}
+
+// DecodeNoLimit decodes a bech32 encoded string, returning the human-readable
+// part and the data part excluding the checksum.  This function does NOT
+// validate against the BIP-173 maximum length allowed for bech32 strings and
+// is meant for use in custom applications (such as lightning network payment
+// requests), NOT on-chain addresses.
+//
+// Note that the returned data is 5-bit (base32) encoded and the human-readable
+// part will be lowercase.
+func DecodeNoLimit(bech string) (string, []byte, error) {
+	hrp, data, _, err := decodeNoLimit(bech)
+	return hrp, data, err
 }
 
 // Decode decodes a bech32 encoded string, returning the human-readable part and
@@ -243,16 +277,31 @@ func Decode(bech string) (string, []byte, error) {
 		return "", nil, ErrInvalidLength(len(bech))
 	}
 
-	return DecodeNoLimit(bech)
+	hrp, data, _, err := decodeNoLimit(bech)
+	return hrp, data, err
 }
 
-// Encode encodes a byte slice into a bech32 string with the given
+// DecodeGeneric is identical to the existing Decode method, but will also
-// human-readable part (HRP).  The HRP will be converted to lowercase if needed
+// return bech32 version that matches the decoded checksum. This method should
-// since mixed cased encodings are not permitted and lowercase is used for
+// be used when decoding segwit addresses, as it enables additional
-// checksum purposes.  Note that the bytes must each encode 5 bits (base32).
+// verification to ensure the proper checksum is used.
-func Encode(hrp string, data []byte) (string, error) {
+func DecodeGeneric(bech string) (string, []byte, Version, error) {
-	// The resulting bech32 string is the concatenation of the lowercase hrp,
+	// The maximum allowed length for a bech32 string is 90.
-	// the separator 1, data and the 6-byte checksum.
+	if len(bech) > 90 {
+		return "", nil, VersionUnknown, ErrInvalidLength(len(bech))
+	}
+
+	return decodeNoLimit(bech)
+}
+
+// encodeGeneric is the base bech32 encoding function that is aware of the
+// existence of the checksum versions. This method is private, as the Encode
+// and EncodeM methods are intended to be used instead.
+func encodeGeneric(hrp string, data []byte,
+	version Version) (string, error) {
+
+	// The resulting bech32 string is the concatenation of the lowercase
+	// hrp, the separator 1, data and the 6-byte checksum.
 	hrp = strings.ToLower(hrp)
 	var bldr strings.Builder
 	bldr.Grow(len(hrp) + 1 + len(data) + 6)
@@ -268,11 +317,26 @@ func Encode(hrp string, data []byte) (string, error) {
 	}
 
 	// Calculate and write the checksum of the data.
-	writeBech32Checksum(hrp, data, &bldr)
+	writeBech32Checksum(hrp, data, &bldr, version)
 
 	return bldr.String(), nil
 }
 
+// Encode encodes a byte slice into a bech32 string with the given
+// human-readable part (HRP).  The HRP will be converted to lowercase if needed
+// since mixed cased encodings are not permitted and lowercase is used for
+// checksum purposes.  Note that the bytes must each encode 5 bits (base32).
+func Encode(hrp string, data []byte) (string, error) {
+	return encodeGeneric(hrp, data, Version0)
+}
+
+// EncodeM is the exactly same as the Encode method, but it uses the new
+// bech32m constant instead of the original one. It should be used whenever one
+// attempts to encode a segwit address of v1 and beyond.
+func EncodeM(hrp string, data []byte) (string, error) {
+	return encodeGeneric(hrp, data, VersionM)
+}
+
 // ConvertBits converts a byte slice where each byte is encoding fromBits bits,
 // to a byte slice where each byte is encoding toBits bits.
 func ConvertBits(data []byte, fromBits, toBits uint8, pad bool) ([]byte, error) {

bech32/bech32_test.go

@@ -22,11 +22,12 @@ func TestBech32(t *testing.T) {
 		expectedError error
 	}{
 		{"A12UEL5L", nil},
+		{"a12uel5l", nil},
 		{"an83characterlonghumanreadablepartthatcontainsthenumber1andtheexcludedcharactersbio1tt5tgs", nil},
 		{"abcdef1qpzry9x8gf2tvdw0s3jn54khce6mua7lmqqqxw", nil},
 		{"11qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqc8247j", nil},
 		{"split1checkupstagehandshakeupstreamerranterredcaperred2y9e3w", nil},
-		{"split1checkupstagehandshakeupstreamerranterredcaperred2y9e2w", ErrInvalidChecksum{"2y9e3w", "2y9e2w"}},              // invalid checksum
+		{"split1checkupstagehandshakeupstreamerranterredcaperred2y9e2w", ErrInvalidChecksum{"2y9e3w", "2y9e3wlc445v", "2y9e2w"}}, // invalid checksum
 		{"s lit1checkupstagehandshakeupstreamerranterredcaperredp8hs2p", ErrInvalidCharacter(' ')},                               // invalid character (space) in hrp
 		{"spl\x7Ft1checkupstagehandshakeupstreamerranterredcaperred2y9e3w", ErrInvalidCharacter(127)},                            // invalid character (DEL) in hrp
 		{"split1cheo2y9e2w", ErrNonCharsetChar('o')},                                                                             // invalid character (o) in data part
@@ -44,7 +45,7 @@ func TestBech32(t *testing.T) {
 		{"x1b4n0q5v", ErrNonCharsetChar(98)},
 		{"li1dgmt3", ErrInvalidSeparatorIndex(2)},
 		{"de1lg7wt\xff", ErrInvalidCharacter(0xff)},
-		{"A1G7SGD8", ErrInvalidChecksum{"2uel5l", "g7sgd8"}},
+		{"A1G7SGD8", ErrInvalidChecksum{"2uel5l", "2uel5llqfn3a", "g7sgd8"}},
 		{"10a06t8", ErrInvalidLength(7)},
 		{"1qzzfhee", ErrInvalidSeparatorIndex(0)},
 		{"a12UEL5L", ErrMixedCase{}},
@@ -86,6 +87,127 @@ func TestBech32(t *testing.T) {
 	}
 }
 
+// TestBech32M tests that the following set of strings, based on the test
+// vectors in BIP-350 are either valid or invalid using the new bech32m
+// checksum algo. Some of these strings are similar to the set of above test
+// vectors, but end up with different checksums.
+func TestBech32M(t *testing.T) {
+	tests := []struct {
+		str           string
+		expectedError error
+	}{
+		{"A1LQFN3A", nil},
+		{"a1lqfn3a", nil},
+		{"an83characterlonghumanreadablepartthatcontainsthetheexcludedcharactersbioandnumber11sg7hg6", nil},
+		{"abcdef1l7aum6echk45nj3s0wdvt2fg8x9yrzpqzd3ryx", nil},
+		{"11llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllludsr8", nil},
+		{"split1checkupstagehandshakeupstreamerranterredcaperredlc445v", nil},
+		{"?1v759aa", nil},
+
+		// Additional test vectors used in bitcoin core
+		{"\x201xj0phk", ErrInvalidCharacter('\x20')},
+		{"\x7f1g6xzxy", ErrInvalidCharacter('\x7f')},
+		{"\x801vctc34", ErrInvalidCharacter('\x80')},
+		{"an84characterslonghumanreadablepartthatcontainsthetheexcludedcharactersbioandnumber11d6pts4", ErrInvalidLength(91)},
+		{"qyrz8wqd2c9m", ErrInvalidSeparatorIndex(-1)},
+		{"1qyrz8wqd2c9m", ErrInvalidSeparatorIndex(0)},
+		{"y1b0jsk6g", ErrNonCharsetChar(98)},
+		{"lt1igcx5c0", ErrNonCharsetChar(105)},
+		{"in1muywd", ErrInvalidSeparatorIndex(2)},
+		{"mm1crxm3i", ErrNonCharsetChar(105)},
+		{"au1s5cgom", ErrNonCharsetChar(111)},
+		{"M1VUXWEZ", ErrInvalidChecksum{"mzl49c", "mzl49cw70eq6", "vuxwez"}},
+		{"16plkw9", ErrInvalidLength(7)},
+		{"1p2gdwpf", ErrInvalidSeparatorIndex(0)},
+
+		{" 1nwldj5", ErrInvalidCharacter(' ')},
+		{"\x7f" + "1axkwrx", ErrInvalidCharacter(0x7f)},
+		{"\x801eym55h", ErrInvalidCharacter(0x80)},
+	}
+
+	for i, test := range tests {
+		str := test.str
+		hrp, decoded, err := Decode(str)
+		if test.expectedError != err {
+			t.Errorf("%d: (%v) expected decoding error %v "+
+				"instead got %v", i, str, test.expectedError,
+				err)
+			continue
+		}
+
+		if err != nil {
+			// End test case here if a decoding error was expected.
+			continue
+		}
+
+		// Check that it encodes to the same string, using bech32 m.
+		encoded, err := EncodeM(hrp, decoded)
+		if err != nil {
+			t.Errorf("encoding failed: %v", err)
+		}
+
+		if encoded != strings.ToLower(str) {
+			t.Errorf("expected data to encode to %v, but got %v",
+				str, encoded)
+		}
+
+		// Flip a bit in the string an make sure it is caught.
+		pos := strings.LastIndexAny(str, "1")
+		flipped := str[:pos+1] + string((str[pos+1] ^ 1)) + str[pos+2:]
+		_, _, err = Decode(flipped)
+		if err == nil {
+			t.Error("expected decoding to fail")
+		}
+	}
+}
+
+// TestBech32DecodeGeneric tests that given a bech32 string, or a bech32m
+// string, the proper checksum version is returned so that callers can perform
+// segwit addr validation.
+func TestBech32DecodeGeneric(t *testing.T) {
+	tests := []struct {
+		str     string
+		version Version
+	}{
+		{"A1LQFN3A", VersionM},
+		{"a1lqfn3a", VersionM},
+		{"an83characterlonghumanreadablepartthatcontainsthetheexcludedcharactersbioandnumber11sg7hg6", VersionM},
+		{"abcdef1l7aum6echk45nj3s0wdvt2fg8x9yrzpqzd3ryx", VersionM},
+		{"11llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllllludsr8", VersionM},
+		{"split1checkupstagehandshakeupstreamerranterredcaperredlc445v", VersionM},
+		{"?1v759aa", VersionM},
+
+		{"A12UEL5L", Version0},
+		{"a12uel5l", Version0},
+		{"an83characterlonghumanreadablepartthatcontainsthenumber1andtheexcludedcharactersbio1tt5tgs", Version0},
+		{"abcdef1qpzry9x8gf2tvdw0s3jn54khce6mua7lmqqqxw", Version0},
+		{"11qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqc8247j", Version0},
+		{"split1checkupstagehandshakeupstreamerranterredcaperred2y9e3w", Version0},
+
+		{"BC1QW508D6QEJXTDG4Y5R3ZARVARY0C5XW7KV8F3T4", Version0},
+		{"tb1qrp33g0q5c5txsp9arysrx4k6zdkfs4nce4xj0gdcccefvpysxf3q0sl5k7", Version0},
+		{"bc1pw508d6qejxtdg4y5r3zarvary0c5xw7kw508d6qejxtdg4y5r3zarvary0c5xw7kt5nd6y", VersionM},
+		{"BC1SW50QGDZ25J", VersionM},
+		{"bc1zw508d6qejxtdg4y5r3zarvaryvaxxpcs", VersionM},
+		{"tb1qqqqqp399et2xygdj5xreqhjjvcmzhxw4aywxecjdzew6hylgvsesrxh6hy", Version0},
+		{"tb1pqqqqp399et2xygdj5xreqhjjvcmzhxw4aywxecjdzew6hylgvsesf3hn0c", VersionM},
+		{"bc1p0xlxvlhemja6c4dqv22uapctqupfhlxm9h8z3k2e72q4k9hcz7vqzk5jj0", VersionM},
+	}
+	for i, test := range tests {
+		_, _, version, err := DecodeGeneric(test.str)
+		if err != nil {
+			t.Errorf("%d: (%v) unexpected error during "+
+				"decoding: %v", i, test.str, err)
+			continue
+		}
+
+		if version != test.version {
+			t.Errorf("(%v): invalid version: expected %v, got %v",
+				test.str, test.version, version)
+		}
+	}
+}
+
 // TestMixedCaseEncode ensures mixed case HRPs are converted to lowercase as
 // expected when encoding and that decoding the produced encoding when converted
 // to all uppercase produces the lowercase HRP and original data.
@@ -242,7 +364,7 @@ func TestBech32Base256(t *testing.T) {
 	}, {
 		name:    "same as previous but with checksum invalidated",
 		encoded: "split1checkupstagehandshakeupstreamerranterredcaperred2y9e2w",
-		err:     ErrInvalidChecksum{"2y9e3w", "2y9e2w"},
+		err:     ErrInvalidChecksum{"2y9e3w", "2y9e3wlc445v", "2y9e2w"},
 	}, {
 		name:    "hrp with invalid character (space)",
 		encoded: "s lit1checkupstagehandshakeupstreamerranterredcaperredp8hs2p",