started with aztec code

aztec/highlevel.go

@@ -0,0 +1,171 @@
+package aztec
+
+import (
+	"github.com/boombuler/barcode/utils"
+)
+
+func highlevelEncode(data []byte) *utils.BitList {
+	states := stateSlice{initialState}
+
+	for index := 0; index < len(data); index++ {
+		pairCode := 0
+		nextChar := byte(0)
+		if index+1 < len(data) {
+			nextChar = data[index+1]
+		}
+
+		switch cur := data[index]; {
+		case cur == '\r' && nextChar == '\n':
+			pairCode = 2
+		case cur == '.' && nextChar == ' ':
+			pairCode = 3
+		case cur == ',' && nextChar == ' ':
+			pairCode = 4
+		case cur == ':' && nextChar == ' ':
+			pairCode = 5
+		}
+		if pairCode > 0 {
+			// We have one of the four special PUNCT pairs.  Treat them specially.
+			// Get a new set of states for the two new characters.
+			states = updateStateListForPair(states, data, index, pairCode)
+			index++
+		} else {
+			// Get a new set of states for the new character.
+			states = updateStateListForChar(states, data, index)
+		}
+	}
+	minBitCnt := int((^uint(0)) >> 1)
+	var result *state = nil
+	for _, s := range states {
+		if s.bitCount < minBitCnt {
+			minBitCnt = s.bitCount
+			result = s
+		}
+	}
+	if result != nil {
+		return result.toBitList(data)
+	} else {
+		return new(utils.BitList)
+	}
+}
+
+func simplifyStates(states stateSlice) stateSlice {
+	var result stateSlice = nil
+	for _, newState := range states {
+		add := true
+		var newResult stateSlice = nil
+
+		for _, oldState := range result {
+			if add && oldState.isBetterThanOrEqualTo(newState) {
+				add = false
+			}
+			if !(add && newState.isBetterThanOrEqualTo(oldState)) {
+				newResult = append(newResult, oldState)
+			}
+		}
+
+		if add {
+			result = append(newResult, newState)
+		} else {
+			result = newResult
+		}
+
+	}
+
+	return result
+}
+
+// We update a set of states for a new character by updating each state
+// for the new character, merging the results, and then removing the
+// non-optimal states.
+func updateStateListForChar(states stateSlice, data []byte, index int) stateSlice {
+	var result stateSlice = nil
+	for _, s := range states {
+		if r := updateStateForChar(s, data, index); len(r) > 0 {
+			result = append(result, r...)
+		}
+	}
+	return simplifyStates(result)
+}
+
+// Return a set of states that represent the possible ways of updating this
+// state for the next character.  The resulting set of states are added to
+// the "result" list.
+func updateStateForChar(s *state, data []byte, index int) stateSlice {
+	var result stateSlice = nil
+	ch := data[index]
+	charInCurrentTable := charMap[s.mode][ch] > 0
+
+	var stateNoBinary *state = nil
+	for mode := mode_upper; mode <= mode_punct; mode++ {
+		charInMode := charMap[mode][ch]
+		if charInMode > 0 {
+			if stateNoBinary == nil {
+				// Only create stateNoBinary the first time it's required.
+				stateNoBinary = s.endBinaryShift(index)
+			}
+			// Try generating the character by latching to its mode
+			if !charInCurrentTable || mode == s.mode || mode == mode_digit {
+				// If the character is in the current table, we don't want to latch to
+				// any other mode except possibly digit (which uses only 4 bits).  Any
+				// other latch would be equally successful *after* this character, and
+				// so wouldn't save any bits.
+				res := stateNoBinary.latchAndAppend(mode, charInMode)
+				result = append(result, res)
+			}
+			// Try generating the character by switching to its mode.
+			if _, ok := shiftTable[s.mode][mode]; !charInCurrentTable && ok {
+				// It never makes sense to temporarily shift to another mode if the
+				// character exists in the current mode.  That can never save bits.
+				res := stateNoBinary.shiftAndAppend(mode, charInMode)
+				result = append(result, res)
+			}
+		}
+	}
+	if s.bShiftByteCount > 0 || charMap[s.mode][ch] == 0 {
+		// It's never worthwhile to go into binary shift mode if you're not already
+		// in binary shift mode, and the character exists in your current mode.
+		// That can never save bits over just outputting the char in the current mode.
+		res := s.addBinaryShiftChar(index)
+		result = append(result, res)
+	}
+	return result
+}
+
+// We update a set of states for a new character by updating each state
+// for the new character, merging the results, and then removing the
+// non-optimal states.
+func updateStateListForPair(states stateSlice, data []byte, index int, pairCode int) stateSlice {
+	var result stateSlice = nil
+	for _, s := range states {
+		if r := updateStateForPair(s, data, index, pairCode); len(r) > 0 {
+			result = append(result, r...)
+		}
+	}
+	return simplifyStates(result)
+}
+
+func updateStateForPair(s *state, data []byte, index int, pairCode int) stateSlice {
+	var result stateSlice
+	stateNoBinary := s.endBinaryShift(index)
+	// Possibility 1.  Latch to MODE_PUNCT, and then append this code
+	result = append(result, stateNoBinary.latchAndAppend(mode_punct, pairCode))
+	if s.mode != mode_punct {
+		// Possibility 2.  Shift to MODE_PUNCT, and then append this code.
+		// Every state except MODE_PUNCT (handled above) can shift
+		result = append(result, stateNoBinary.shiftAndAppend(mode_punct, pairCode))
+	}
+	if pairCode == 3 || pairCode == 4 {
+		// both characters are in DIGITS.  Sometimes better to just add two digits
+		digitState := stateNoBinary.
+			latchAndAppend(mode_digit, 16-pairCode). // period or comma in DIGIT
+			latchAndAppend(mode_digit, 1)            // space in DIGIT
+		result = append(result, digitState)
+	}
+	if s.bShiftByteCount > 0 {
+		// It only makes sense to do the characters as binary if we're already
+		// in binary mode.
+		result = append(result, s.addBinaryShiftChar(index).addBinaryShiftChar(index+1))
+	}
+	return result
+}

aztec/highlevel_test.go

@@ -0,0 +1,132 @@
+package aztec
+
+import (
+	"bytes"
+	"strings"
+	"testing"
+
+	"github.com/boombuler/barcode/utils"
+)
+
+func bitStr(bl *utils.BitList) string {
+	buf := new(bytes.Buffer)
+
+	for i := 0; i < bl.Len(); i++ {
+		if bl.GetBit(i) {
+			buf.WriteRune('X')
+		} else {
+			buf.WriteRune('.')
+		}
+	}
+	return buf.String()
+}
+
+func testHighLevelEncodeString(t *testing.T, s, expectedBits string) {
+	bits := highlevelEncode([]byte(s))
+	result := bitStr(bits)
+	expectedBits = strings.Replace(expectedBits, " ", "", -1)
+
+	if result != expectedBits {
+		t.Errorf("invalid result for highlevelEncode(%q). Got:\n%s", s, result)
+	}
+}
+func testHighLevelEncodeStringCnt(t *testing.T, s string, expectedBitCnt int) {
+	bits := highlevelEncode([]byte(s))
+
+	if bits.Len() != expectedBitCnt {
+		t.Errorf("invalid result for highlevelEncode(%q). Got %d, expected %d bits", s, bits.Len(), expectedBitCnt)
+	}
+}
+
+func Test_HighLevelEncode(t *testing.T) {
+	testHighLevelEncodeString(t, "A. b.",
+		// 'A'  P/S   '. ' L/L    b    D/L    '.'
+		"...X. ..... ...XX XXX.. ...XX XXXX. XX.X")
+	testHighLevelEncodeString(t, "Lorem ipsum.",
+		// 'L'  L/L   'o'   'r'   'e'   'm'   ' '   'i'   'p'   's'   'u'   'm'   D/L   '.'
+		".XX.X XXX.. X.... X..XX ..XX. .XXX. ....X .X.X. X...X X.X.. X.XX. .XXX. XXXX. XX.X")
+	testHighLevelEncodeString(t, "Lo. Test 123.",
+		// 'L'  L/L   'o'   P/S   '. '  U/S   'T'   'e'   's'   't'    D/L   ' '  '1'  '2'  '3'  '.'
+		".XX.X XXX.. X.... ..... ...XX XXX.. X.X.X ..XX. X.X.. X.X.X  XXXX. ...X ..XX .X.. .X.X XX.X")
+	testHighLevelEncodeString(t, "Lo...x",
+		// 'L'  L/L   'o'   D/L   '.'  '.'  '.'  U/L  L/L   'x'
+		".XX.X XXX.. X.... XXXX. XX.X XX.X XX.X XXX. XXX.. XX..X")
+	testHighLevelEncodeString(t, ". x://abc/.",
+		//P/S   '. '  L/L   'x'   P/S   ':'   P/S   '/'   P/S   '/'   'a'   'b'   'c'   P/S   '/'   D/L   '.'
+		"..... ...XX XXX.. XX..X ..... X.X.X ..... X.X.. ..... X.X.. ...X. ...XX ..X.. ..... X.X.. XXXX. XX.X")
+	// Uses Binary/Shift rather than Lower/Shift to save two bits.
+	testHighLevelEncodeString(t, "ABCdEFG",
+		//'A'   'B'   'C'   B/S    =1    'd'     'E'   'F'   'G'
+		"...X. ...XX ..X.. XXXXX ....X .XX..X.. ..XX. ..XXX .X...")
+
+	testHighLevelEncodeStringCnt(t,
+		// Found on an airline boarding pass.  Several stretches of Binary shift are
+		// necessary to keep the bitcount so low.
+		"09  UAG    ^160MEUCIQC0sYS/HpKxnBELR1uB85R20OoqqwFGa0q2uEi"+
+			"Ygh6utAIgLl1aBVM4EOTQtMQQYH9M2Z3Dp4qnA/fwWuQ+M8L3V8U=",
+		823)
+}
+
+func Test_HighLevelEncodeBinary(t *testing.T) {
+	// binary short form single byte
+	testHighLevelEncodeString(t, "N\u0000N",
+		// 'N'  B/S    =1   '\0'      N
+		".XXXX XXXXX ....X ........ .XXXX") // Encode "N" in UPPER
+
+	testHighLevelEncodeString(t, "N\u0000n",
+		// 'N'  B/S    =2   '\0'       'n'
+		".XXXX XXXXX ...X. ........ .XX.XXX.") // Encode "n" in BINARY
+
+	// binary short form consecutive bytes
+	testHighLevelEncodeString(t, "N\x00\x80 A",
+		// 'N'  B/S    =2    '\0'    \u0080   ' '  'A'
+		".XXXX XXXXX ...X. ........ X....... ....X ...X.")
+
+	// binary skipping over single character
+	testHighLevelEncodeString(t, "\x00a\xFF\x80 A",
+		// B/S  =4    '\0'      'a'     '\3ff'   '\200'   ' '   'A'
+		"XXXXX ..X.. ........ .XX....X XXXXXXXX X....... ....X ...X.")
+
+	// getting into binary mode from digit mode
+	testHighLevelEncodeString(t, "1234\u0000",
+		//D/L   '1'  '2'  '3'  '4'  U/L  B/S    =1    \0
+		"XXXX. ..XX .X.. .X.X .XX. XXX. XXXXX ....X ........")
+
+	// Create a string in which every character requires binary
+	sb := new(bytes.Buffer)
+	for i := 0; i <= 3000; i++ {
+		sb.WriteByte(byte(128 + (i % 30)))
+	}
+
+	// Test the output generated by Binary/Switch, particularly near the
+	// places where the encoding changes: 31, 62, and 2047+31=2078
+	for _, i := range []int{1, 2, 3, 10, 29, 30, 31, 32, 33, 60, 61, 62, 63, 64, 2076, 2077, 2078, 2079, 2080, 2100} {
+		// This is the expected length of a binary string of length "i"
+		expectedLength := (8 * i)
+		switch {
+		case i <= 31:
+			expectedLength += 10
+		case i <= 62:
+			expectedLength += 20
+		case i <= 2078:
+			expectedLength += 21
+		default:
+			expectedLength += 31
+		}
+		data := string(sb.Bytes()[:i])
+
+		// Verify that we are correct about the length.
+		testHighLevelEncodeStringCnt(t, data, expectedLength)
+		if i != 1 && i != 32 && i != 2079 {
+			// The addition of an 'a' at the beginning or end gets merged into the binary code
+			// in those cases where adding another binary character only adds 8 or 9 bits to the result.
+			// So we exclude the border cases i=1,32,2079
+			// A lower case letter at the beginning will be merged into binary mode
+			testHighLevelEncodeStringCnt(t, "a"+string(sb.Bytes()[:i-1]), expectedLength)
+			// A lower case letter at the end will also be merged into binary mode
+			testHighLevelEncodeStringCnt(t, string(sb.Bytes()[:i-1])+"a", expectedLength)
+		}
+		// A lower case letter at both ends will enough to latch us into LOWER.
+		testHighLevelEncodeStringCnt(t, "a"+data+"b", expectedLength+15)
+	}
+}

aztec/state.go

@@ -0,0 +1,264 @@
+package aztec
+
+import (
+	"fmt"
+
+	"github.com/boombuler/barcode/utils"
+)
+
+type encodingMode byte
+
+const (
+	mode_upper encodingMode = iota // 5 bits
+	mode_lower                     // 5 bits
+	mode_digit                     // 4 bits
+	mode_mixed                     // 5 bits
+	mode_punct                     // 5 bits
+)
+
+var (
+	// The Latch Table shows, for each pair of Modes, the optimal method for
+	// getting from one mode to another.  In the worst possible case, this can
+	// be up to 14 bits.  In the best possible case, we are already there!
+	// The high half-word of each entry gives the number of bits.
+	// The low half-word of each entry are the actual bits necessary to change
+	latchTable = map[encodingMode]map[encodingMode]int{
+		mode_upper: {
+			mode_upper: 0,
+			mode_lower: (5 << 16) + 28,
+			mode_digit: (5 << 16) + 30,
+			mode_mixed: (5 << 16) + 29,
+			mode_punct: (10 << 16) + (29 << 5) + 30,
+		},
+		mode_lower: {
+			mode_upper: (9 << 16) + (30 << 4) + 14,
+			mode_lower: 0,
+			mode_digit: (5 << 16) + 30,
+			mode_mixed: (5 << 16) + 29,
+			mode_punct: (10 << 16) + (29 << 5) + 30,
+		},
+		mode_digit: {
+			mode_upper: (4 << 16) + 14,
+			mode_lower: (9 << 16) + (14 << 5) + 28,
+			mode_digit: 0,
+			mode_mixed: (9 << 16) + (14 << 5) + 29,
+			mode_punct: (14 << 16) + (14 << 10) + (29 << 5) + 30,
+		},
+		mode_mixed: {
+			mode_upper: (5 << 16) + 29,
+			mode_lower: (5 << 16) + 28,
+			mode_digit: (10 << 16) + (29 << 5) + 30,
+			mode_mixed: 0,
+			mode_punct: (5 << 16) + 30,
+		},
+		mode_punct: {
+			mode_upper: (5 << 16) + 31,
+			mode_lower: (10 << 16) + (31 << 5) + 28,
+			mode_digit: (10 << 16) + (31 << 5) + 30,
+			mode_mixed: (10 << 16) + (31 << 5) + 29,
+			mode_punct: 0,
+		},
+	}
+	// A map showing the available shift codes.  (The shifts to BINARY are not shown)
+	shiftTable = map[encodingMode]map[encodingMode]int{
+		mode_upper: {
+			mode_punct: 0,
+		},
+		mode_lower: {
+			mode_punct: 0,
+			mode_upper: 28,
+		},
+		mode_mixed: {
+			mode_punct: 0,
+		},
+		mode_digit: {
+			mode_punct: 0,
+			mode_upper: 15,
+		},
+	}
+	charMap map[encodingMode][]int
+)
+
+type state struct {
+	mode            encodingMode
+	tokens          token
+	bShiftByteCount int
+	bitCount        int
+}
+type stateSlice []*state
+
+var initialState *state = &state{
+	mode:            mode_upper,
+	tokens:          nil,
+	bShiftByteCount: 0,
+	bitCount:        0,
+}
+
+func init() {
+	charMap = make(map[encodingMode][]int)
+	charMap[mode_upper] = make([]int, 256)
+	charMap[mode_lower] = make([]int, 256)
+	charMap[mode_digit] = make([]int, 256)
+	charMap[mode_mixed] = make([]int, 256)
+	charMap[mode_punct] = make([]int, 256)
+
+	charMap[mode_upper][' '] = 1
+	for c := 'A'; c <= 'Z'; c++ {
+		charMap[mode_upper][int(c)] = int(c - 'A' + 2)
+	}
+
+	charMap[mode_lower][' '] = 1
+	for c := 'a'; c <= 'z'; c++ {
+		charMap[mode_lower][c] = int(c - 'a' + 2)
+	}
+	charMap[mode_digit][' '] = 1
+	for c := '0'; c <= '9'; c++ {
+		charMap[mode_digit][c] = int(c - '0' + 2)
+	}
+	charMap[mode_digit][','] = 12
+	charMap[mode_digit]['.'] = 13
+
+	mixedTable := []int{
+		0, ' ', 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,
+		11, 12, 13, 27, 28, 29, 30, 31, '@', '\\', '^',
+		'_', '`', '|', '~', 127,
+	}
+	for i, v := range mixedTable {
+		charMap[mode_mixed][v] = i
+	}
+
+	punctTable := []int{
+		0, '\r', 0, 0, 0, 0, '!', '\'', '#', '$', '%', '&', '\'',
+		'(', ')', '*', '+', ',', '-', '.', '/', ':', ';', '<', '=', '>', '?',
+		'[', ']', '{', '}',
+	}
+	for i, v := range punctTable {
+		if v > 0 {
+			charMap[mode_punct][v] = i
+		}
+	}
+}
+
+func (em encodingMode) BitCount() byte {
+	if em == mode_digit {
+		return 4
+	}
+	return 5
+}
+
+// Create a new state representing this state with a latch to a (not
+// necessary different) mode, and then a code.
+func (s *state) latchAndAppend(mode encodingMode, value int) *state {
+	bitCount := s.bitCount
+	tokens := s.tokens
+
+	if mode != s.mode {
+		latch := latchTable[s.mode][mode]
+		tokens = newSimpleToken(tokens, latch&0xFFFF, byte(latch>>16))
+		bitCount += latch >> 16
+	}
+	tokens = newSimpleToken(tokens, value, mode.BitCount())
+	return &state{
+		mode:            mode,
+		tokens:          tokens,
+		bShiftByteCount: 0,
+		bitCount:        bitCount + int(mode.BitCount()),
+	}
+}
+
+// Create a new state representing this state, with a temporary shift
+// to a different mode to output a single value.
+func (s *state) shiftAndAppend(mode encodingMode, value int) *state {
+	tokens := s.tokens
+
+	// Shifts exist only to UPPER and PUNCT, both with tokens size 5.
+	tokens = newSimpleToken(tokens, shiftTable[s.mode][mode], s.mode.BitCount())
+	tokens = newSimpleToken(tokens, value, 5)
+
+	return &state{
+		mode:            s.mode,
+		tokens:          tokens,
+		bShiftByteCount: 0,
+		bitCount:        s.bitCount + int(s.mode.BitCount()) + 5,
+	}
+}
+
+// Create a new state representing this state, but an additional character
+// output in Binary Shift mode.
+func (s *state) addBinaryShiftChar(index int) *state {
+	tokens := s.tokens
+	mode := s.mode
+	bitCnt := s.bitCount
+	if s.mode == mode_punct || s.mode == mode_digit {
+		latch := latchTable[s.mode][mode_upper]
+		tokens = newSimpleToken(tokens, latch&0xFFFF, byte(latch>>16))
+		bitCnt += latch >> 16
+		mode = mode_upper
+	}
+	deltaBitCount := 8
+	if s.bShiftByteCount == 0 || s.bShiftByteCount == 31 {
+		deltaBitCount = 18
+	} else if s.bShiftByteCount == 62 {
+		deltaBitCount = 9
+	}
+	result := &state{
+		mode:            mode,
+		tokens:          tokens,
+		bShiftByteCount: s.bShiftByteCount + 1,
+		bitCount:        bitCnt + deltaBitCount,
+	}
+	if result.bShiftByteCount == 2047+31 {
+		// The string is as long as it's allowed to be.  We should end it.
+		result = result.endBinaryShift(index + 1)
+	}
+
+	return result
+}
+
+// Create the state identical to this one, but we are no longer in
+// Binary Shift mode.
+func (s *state) endBinaryShift(index int) *state {
+	if s.bShiftByteCount == 0 {
+		return s
+	}
+	tokens := newShiftToken(s.tokens, index-s.bShiftByteCount, s.bShiftByteCount)
+	return &state{
+		mode:            s.mode,
+		tokens:          tokens,
+		bShiftByteCount: 0,
+		bitCount:        s.bitCount,
+	}
+}
+
+// Returns true if "this" state is better (or equal) to be in than "that"
+// state under all possible circumstances.
+func (this *state) isBetterThanOrEqualTo(other *state) bool {
+	mySize := this.bitCount + (latchTable[this.mode][other.mode] >> 16)
+
+	if other.bShiftByteCount > 0 && (this.bShiftByteCount == 0 || this.bShiftByteCount > other.bShiftByteCount) {
+		mySize += 10 // Cost of entering Binary Shift mode.
+	}
+	return mySize <= other.bitCount
+}
+
+func (s *state) toBitList(text []byte) *utils.BitList {
+	tokens := make([]token, 0)
+	se := s.endBinaryShift(len(text))
+
+	for t := se.tokens; t != nil; t = t.prev() {
+		tokens = append(tokens, t)
+	}
+	res := new(utils.BitList)
+	for i := len(tokens) - 1; i >= 0; i-- {
+		tokens[i].appendTo(res, text)
+	}
+	return res
+}
+
+func (s *state) String() string {
+	tokens := make([]token, 0)
+	for t := s.tokens; t != nil; t = t.prev() {
+		tokens = append([]token{t}, tokens...)
+	}
+	return fmt.Sprintf("M:%d bits=%d bytes=%d: %v", s.mode, s.bitCount, s.bShiftByteCount, tokens)
+}

aztec/token.go

@@ -0,0 +1,75 @@
+package aztec
+
+import (
+	"fmt"
+
+	"github.com/boombuler/barcode/utils"
+)
+
+type token interface {
+	fmt.Stringer
+	prev() token
+	appendTo(bits *utils.BitList, text []byte)
+}
+
+type simpleToken struct {
+	token
+	value    int
+	bitCount byte
+}
+
+type binaryShiftToken struct {
+	token
+	bShiftStart   int
+	bShiftByteCnt int
+}
+
+func newSimpleToken(prev token, value int, bitCount byte) token {
+	return &simpleToken{prev, value, bitCount}
+}
+func newShiftToken(prev token, bShiftStart int, bShiftCnt int) token {
+	return &binaryShiftToken{prev, bShiftStart, bShiftCnt}
+}
+
+func (st *simpleToken) prev() token {
+	return st.token
+}
+func (st *simpleToken) appendTo(bits *utils.BitList, text []byte) {
+	bits.AddBits(st.value, st.bitCount)
+}
+func (st *simpleToken) String() string {
+	value := st.value & ((1 << st.bitCount) - 1)
+	value |= 1 << st.bitCount
+	return "<" + fmt.Sprintf("%b", value)[1:] + ">"
+}
+
+func (bst *binaryShiftToken) prev() token {
+	return bst.token
+}
+func (bst *binaryShiftToken) appendTo(bits *utils.BitList, text []byte) {
+	for i := 0; i < bst.bShiftByteCnt; i++ {
+		if i == 0 || (i == 31 && bst.bShiftByteCnt <= 62) {
+			// We need a header before the first character, and before
+			// character 31 when the total byte code is <= 62
+			bits.AddBits(31, 5) // BINARY_SHIFT
+			if bst.bShiftByteCnt > 62 {
+				bits.AddBits(bst.bShiftByteCnt-31, 16)
+			} else if i == 0 {
+				// 1 <= binaryShiftByteCode <= 62
+				if bst.bShiftByteCnt < 31 {
+					bits.AddBits(bst.bShiftByteCnt, 5)
+				} else {
+					bits.AddBits(31, 5)
+				}
+			} else {
+				// 32 <= binaryShiftCount <= 62 and i == 31
+				bits.AddBits(bst.bShiftByteCnt-31, 5)
+			}
+		}
+		bits.AddByte(text[bst.bShiftStart+i])
+	}
+}
+
+func (bst *binaryShiftToken) String() string {
+	return fmt.Sprintf("<%d::%d>", bst.bShiftStart, (bst.bShiftStart + bst.bShiftByteCnt - 1))
+}