pdf.js/PDFFont.js
2011-06-10 06:12:59 +02:00

1110 lines
33 KiB
JavaScript
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

/**
* This dictionary holds decoded fonts data.
*/
var Fonts = new Dict();
/**
* This simple object keep a trace of the fonts that have already been decoded
* by storing a map between the name given by the PDF and the name gather from
* the font (aka the PostScript code of the font itself for Type1 font).
*/
var _Fonts = {};
var Stack = function() {
var innerStack = [];
this.push = function(aOperand) {
innerStack.push(aOperand);
};
this.pop = function() {
if (!this.count())
throw new Error("stackunderflow");
return innerStack.pop();
};
this.peek = function() {
if (!this.count())
return null;
return innerStack[innerStack.length - 1];
};
this.get = function(aIndex) {
return innerStack[aIndex];
};
this.clear = function() {
innerStack = [];
};
this.count = function() {
return innerStack.length;
};
this.dump = function() {
for (var i = 0; i < this.length; i++)
log(innerStack[i]);
};
this.clone = function() {
return innerStack.slice();
};
};
var Base64Encoder = {
encode: function(aData) {
var str = [];
var count = aData.length;
for (var i = 0; i < count; i++)
str.push(aData.getChar ? aData.getChar : String.fromCharCode(aData[i]));
return window.btoa(str.join(""));
}
};
var TrueTypeFont = function(aFontName, aFontFile) {
if (_Fonts[aFontName])
return;
_Fonts[aFontName] = true;
//log("Loading a TrueType font: " + aFontName);
var fontData = Base64Encoder.encode(aFontFile);
Fonts.set(aFontName, fontData);
// Add the css rule
var url = "url(data:font/ttf;base64," + fontData + ");";
document.styleSheets[0].insertRule("@font-face { font-family: '" + aFontName + "'; src: " + url + " }", 0);
};
var Type1Parser = function(aAsciiStream, aBinaryStream) {
var lexer = aAsciiStream ? new Lexer(aAsciiStream) : null;
// Turn on this flag for additional debugging logs
var debug = false;
var dump = function(aData) {
if (debug)
log(aData);
};
// Hold the fontName as declared inside the /FontName postscript directive
// XXX This is a hack but at the moment I need it to map the name declared
// in the PDF and the name in the PS code.
var fontName = "";
/*
* Parse a whole Type1 font stream (from the first segment to the last)
* assuming the 'eexec' block is binary data and fill up the 'Fonts'
* dictionary with the font informations.
*/
var self = this;
this.parse = function() {
if (!debug) {
while (!processNextToken()) {};
return fontName;
} else {
// debug mode is used to debug postcript processing
setTimeout(function() {
if (!processNextToken())
self.parse();
}, 0);
}
};
/*
* Decrypt a Sequence of Ciphertext Bytes to Produce the Original Sequence
* of Plaintext Bytes. The function took a key as a parameter which can be
* for decrypting the eexec block of for decoding charStrings.
*/
var kEexecEncryptionKey = 55665;
var kCharStringsEncryptionKey = 4330;
function decrypt(aStream, aKey, aDiscardNumber) {
var start = Date.now();
var r = aKey, c1 = 52845, c2 = 22719;
var decryptedString = [];
var value = "";
var count = aStream.length;
for (var i = 0; i < count; i++) {
value = aStream.getByte();
decryptedString[i] = String.fromCharCode(value ^ (r >> 8));
r = ((value + r) * c1 + c2) & ((1 << 16) - 1);
}
var end = Date.now();
dump("Time to decrypt string of length " + count + " is " + (end - start));
return decryptedString.slice(aDiscardNumber);
};
/*
* CharStrings are encoded following the the CharString Encoding sequence
* describe in Chapter 6 of the "Adobe Type1 Font Format" specification.
* The value in a byte indicates a command, a number, or subsequent bytes
* that are to be interpreted in a special way.
*
* CharString Number Encoding:
* A CharString byte containing the values from 32 through 255 inclusive
* indicate an integer. These values are decoded in four ranges.
*
* 1. A CharString byte containing a value, v, between 32 and 246 inclusive,
* indicate the integer v - 139. Thus, the integer values from -107 through
* 107 inclusive may be encoded in single byte.
*
* 2. A CharString byte containing a value, v, between 247 and 250 inclusive,
* indicates an integer involving the next byte, w, according to the formula:
* [(v - 247) x 256] + w + 108
*
* 3. A CharString byte containing a value, v, between 251 and 254 inclusive,
* indicates an integer involving the next byte, w, according to the formula:
* -[(v - 251) * 256] - w - 108
*
* 4. A CharString containing the value 255 indicates that the next 4 bytes
* are a two complement signed integer. The first of these bytes contains the
* highest order bits, the second byte contains the next higher order bits
* and the fourth byte contain the lowest order bits.
*
*
* CharString Command Encoding:
* CharStrings commands are encoded in 1 or 2 bytes.
*
* Single byte commands are encoded in 1 byte that contains a value between
* 0 and 31 inclusive.
* If a command byte contains the value 12, then the value in the next byte
* indicates a command. This "escape" mechanism allows many extra commands
* to be encoded and this encoding technique helps to minimize the length of
* the charStrings.
*/
var charStringDictionary = {
"1": "hstem",
"3": "vstem",
"4": "vmoveto",
"5": "rlineto",
"6": "hlineto",
"7": "vlineto",
"8": "rrcurveto",
"9": "closepath",
"10": "callsubr",
"11": "return",
"12": {
"0": "dotsection",
"1": "vstem3",
"3": "hstem3",
"6": "seac",
"7": "sbw",
"12": "div",
"16": "callothersubr",
"17": "pop",
"33": "setcurrentpoint"
},
"13": "hsbw",
"14": "endchar",
"21": "rmoveto",
"22": "hmoveto",
"30": "vhcurveto",
"31": "hcurveto"
};
function decodeCharString(aStream) {
var start = Date.now();
var charString = [];
var value = "";
var count = aStream.length;
for (var i = 0; i < count; i++) {
value = aStream.getByte();
if (value < 32) {
if (value == 12) {
value = charStringDictionary["12"][aStream.getByte()];
i++;
} else {
value = charStringDictionary[value];
}
} else if (value <= 246) {
value = parseInt(value) - 139;
} else if (value <= 250) {
value = ((value - 247) * 256) + parseInt(aStream.getByte()) + 108;
i++;
} else if (value <= 254) {
value = -((value - 251) * 256) - parseInt(aStream.getByte()) - 108;
i++;
} else {
var byte = aStream.getByte();
var high = (byte >> 1);
value = (byte - high) << 24 | aStream.getByte() << 16 |
aStream.getByte() << 8 | aStream.getByte();
i += 4;
}
charString.push(value);
}
var end = Date.now();
dump("Time to decode charString of length " + count + " is " + (end - start));
return charString;
}
/*
* The operand stack holds arbitrary PostScript objects that are the operands
* and results of PostScript operators being executed. The interpreter pushes
* objects on the operand stack when it encounters them as literal data in a
* program being executed. When an operator requires one or more operands, it
* obtains them by popping them off the top of the operand stack. When an
* operator returns one or more results, it does so by pushing them on the
* operand stack.
*/
var operandStack = new Stack();
// Flag indicating if the topmost operand of the operandStack is an array
var operandIsArray = 0;
/*
* The dictionary stack holds only dictionary objects. The current set of
* dictionaries on the dictionary stack defines the environment for all
* implicit name searches, such as those that occur when the interpreter
* encounters an executable name. The role of the dictionary stack is
* introduced in Section 3.3, “Data Types and Objects,” and is further
* explained in Section 3.5, “Execution.” of the PostScript Language
* Reference.
*/
var systemDict = new Dict(),
globalDict = new Dict(),
userDict = new Dict();
var dictionaryStack = new Stack();
dictionaryStack.push(systemDict);
dictionaryStack.push(globalDict);
dictionaryStack.push(userDict);
/*
* The execution stack holds executable objects (mainly procedures and files)
* that are in intermediate stages of execution. At any point in the
* execution of a PostScript program, this stack represents the programs
* call stack. Whenever the interpreter suspends execution of an object to
* execute some other object, it pushes the new object on the execution
* stack. When the interpreter finishes executing an object, it pops that
* object off the execution stack and resumes executing the suspended object
* beneath it.
*/
var executionStack = new Stack();
/*
* Return the next token in the execution stack
*/
function nextInStack() {
var currentProcedure = executionStack.peek();
if (currentProcedure) {
var command = currentProcedure.shift();
if (!currentProcedure.length)
executionStack.pop();
return command;
}
return lexer.getObj();
};
/*
* Get the next token from the executionStack and process it.
* Actually the function does not process the third segment of a Type1 font
* and end on 'closefile'.
*
* The method thrown an error if it encounters an unknown token.
*/
function processNextToken() {
var obj = nextInStack();
if (operandIsArray && !IsCmd(obj, "{") && !IsCmd(obj, "[") &&
!IsCmd(obj, "]") && !IsCmd(obj, "}")) {
dump("Adding an object: " + obj +" to array " + operandIsArray);
var currentArray = operandStack.peek();
for (var i = 1; i < operandIsArray; i++)
currentArray = currentArray[currentArray.length - 1];
currentArray.push(obj);
} else if (IsBool(obj) || IsInt(obj) || IsNum(obj) || IsString(obj)) {
dump("Value: " + obj);
operandStack.push(obj);
} else if (IsName(obj)) {
dump("Name: " + obj.name);
operandStack.push(obj.name);
} else if (IsCmd(obj)) {
var command = obj.cmd;
dump(command);
switch (command) {
case "[":
case "{":
dump("Start" + (command == "{" ? " Executable " : " ") + "Array");
operandIsArray++;
var currentArray = operandStack;
for (var i = 1; i < operandIsArray; i++)
if (currentArray.peek)
currentArray = currentArray.peek();
else
currentArray = currentArray[currentArray.length - 1];
currentArray.push([]);
break;
case "]":
case "}":
var currentArray = operandStack.peek();
for (var i = 1; i < operandIsArray; i++)
currentArray = currentArray[currentArray.length - 1];
dump("End" + (command == "}" ? " Executable " : " ") + "Array: " + currentArray.join(" "));
operandIsArray--;
break;
case "if":
var procedure = operandStack.pop();
var bool = operandStack.pop();
if (!IsBool(bool)) {
dump("if: " + bool);
// we need to execute things, let be dirty
executionStack.push(bool);
} else {
dump("if ( " + bool + " ) { " + procedure + " }");
if (bool)
executionStack.push(procedure);
}
break;
case "ifelse":
var procedure1 = operandStack.pop();
var procedure2 = operandStack.pop();
var bool = !!operandStack.pop();
dump("if ( " + bool + " ) { " + procedure2 + " } else { " + procedure1 + " }");
executionStack.push(bool ? procedure2 : procedure1);
break;
case "for":
var procedure = operandStack.pop();
var limit = operandStack.pop();
var increment = operandStack.pop();
var initial = operandStack.pop();
for (var i = 0; i < limit; i += increment) {
operandStack.push(i);
executionStack.push(procedure.slice());
}
break;
case "dup":
dump("duplicate: " + operandStack.peek());
operandStack.push(operandStack.peek());
break;
case "mark":
operandStack.push("mark");
break;
case "cleartomark":
var command = "";
do {
command = operandStack.pop();
} while (command != "mark");
break;
case "put":
var data = operandStack.pop();
var indexOrKey = operandStack.pop();
var object = operandStack.pop();
dump("put " + data + " in " + object + "[" + indexOrKey + "]");
object.set ? object.set(indexOrKey, data)
: object[indexOrKey] = data;
break;
case "pop":
operandStack.pop();
break;
case "exch":
var operand1 = operandStack.pop();
var operand2 = operandStack.pop();
operandStack.push(operand1);
operandStack.push(operand2);
break;
case "get":
var indexOrKey = operandStack.pop();
var object = operandStack.pop();
var data = object.get ? object.get(indexOrKey) : object[indexOrKey];
dump("get " + object + "[" + indexOrKey + "]: " + data);
operandStack.push(data);
break;
case "currentdict":
var dict = dictionaryStack.peek();
operandStack.push(dict);
break;
case "systemdict":
operandStack.push(systemDict);
break;
case "readonly":
case "executeonly":
case "noaccess":
// Do nothing for the moment
break;
case "currentfile":
operandStack.push("currentfile");
break;
case "array":
var size = operandStack.pop();
var array = new Array(size);
operandStack.push(array);
break;
case "dict":
var size = operandStack.pop();
var dict = new Dict(size);
operandStack.push(dict);
break;
case "begin":
dictionaryStack.push(operandStack.pop());
break;
case "end":
dictionaryStack.pop();
break;
case "def":
var value = operandStack.pop();
var key = operandStack.pop();
// XXX we don't want to do that here but for some reasons the names
// are different between what is declared and the FontName directive
if (key == "FontName" && Fonts.get(value)) {
// The font has already be decoded, stop!
return true;
}
dump("def: " + key + " = " + value);
dictionaryStack.peek().set(key, value);
break;
case "definefont":
var font = operandStack.pop();
var key = operandStack.pop();
dump("definefont " + font + " with key: " + key);
// The key will be the identifier to recognize this font
fontName = key;
Fonts.set(key, font);
operandStack.push(font);
break;
case "known":
var name = operandStack.pop();
var dict = operandStack.pop();
var data = !!dict.get(name);
dump("known: " + data + " :: " + name + " in dict: " + dict);
operandStack.push(data);
break;
case "exec":
executionStack.push(operandStack.pop());
break;
case "eexec":
// All the first segment data has been read, decrypt the second segment
// and start interpreting it in order to decode it
var file = operandStack.pop();
var eexecString = decrypt(aBinaryStream, kEexecEncryptionKey, 4).join("");
dump(eexecString);
lexer = new Lexer(new StringStream(eexecString));
break;
case "LenIV":
error("LenIV: argh! we need to modify the length of discard characters for charStrings");
break;
case "closefile":
var file = operandStack.pop();
return true;
break;
case "index":
var operands = [];
var size = operandStack.pop();
for (var i = 0; i < size; i++)
operands.push(operandStack.pop());
var newOperand = operandStack.peek();
while (operands.length)
operandStack.push(operands.pop());
operandStack.push(newOperand);
break;
case "string":
var size = operandStack.pop();
var str = (new Array(size + 1)).join(" ");
operandStack.push(str);
break;
case "readstring":
var str = operandStack.pop();
var size = str.length;
var file = operandStack.pop();
// Add '1' because of the space separator, this is dirty
var stream = lexer.stream.makeSubStream(lexer.stream.pos + 1, size);
lexer.stream.skip(size + 1);
var charString = decrypt(stream, kCharStringsEncryptionKey, 4).join("");
var charStream = new StringStream(charString);
var decodedCharString = decodeCharString(charStream);
dump("decodedCharString: " + decodedCharString);
operandStack.push(decodedCharString);
// boolean indicating if the operation is a success or not
operandStack.push(true);
break;
case "StandardEncoding":
// For some reason the value is considered as a command, maybe it is
// because of the uppercase 'S'
operandStack.push(obj.cmd);
break;
default:
var command = null;
if (IsCmd(obj)) {
for (var i = 0; i < dictionaryStack.count(); i++) {
if (command = dictionaryStack.get(i).get(obj.cmd)) {
dump("found in dictionnary for " + obj.cmd + " command: " + command);
executionStack.push(command.slice());
break;
}
}
}
if (!command) {
log("operandStack: " + operandStack);
log("dictionaryStack: " + dictionaryStack);
log(obj);
error("Unknow command while parsing font");
}
break;
}
} else if (obj) {
dump("unknow: " + obj);
operandStack.push(obj);
} else { // The End!
operandStack.dump();
return true;
}
return false;
}
function aggregateCommand(aCommand) {
var command = aCommand;
switch (command) {
case "hstem":
case "vstem":
break;
case "rrcurveto":
var stack = [operandStack.pop(), operandStack.pop(),
operandStack.pop(), operandStack.pop(),
operandStack.pop(), operandStack.pop()];
var next = true;
while (next) {
var op = operandStack.peek();
if (op == "rrcurveto") {
operandStack.pop();
stack.push(operandStack.pop());
stack.push(operandStack.pop());
stack.push(operandStack.pop());
stack.push(operandStack.pop());
stack.push(operandStack.pop());
stack.push(operandStack.pop());
} else {
next = false;
}
}
break;
case "hlineto":
case "vlineto":
var last = command;
var stack = [operandStack.pop()];
var next = true;
while (next) {
var op = operandStack.peek();
if (op == "vlineto" && last == "hlineto") {
operandStack.pop();
stack.push(operandStack.pop());
} else if (op == "hlineto" && last == "vlineto") {
operandStack.pop();
stack.push(operandStack.pop());
} else if (op == "rlineto" && command == "hlineto") {
operandStack.pop();
var x = stack.pop();
operandStack.push(0);
operandStack.push(x);
command = "rlineto";
} else if (op == "rlineto" && command == "vlineto") {
operandStack.pop();
operandStack.push(0);
command = "rlineto";
} else {
next = false;
}
last = op;
}
break;
case "rlineto":
var stack = [operandStack.pop(), operandStack.pop()];
var next = true;
while (next) {
var op = operandStack.peek();
if (op == "rlineto") {
operandStack.pop();
stack.push(operandStack.pop());
stack.push(operandStack.pop());
} else if (op == "hlineto") {
operandStack.pop();
stack.push(0);
stack.push(operandStack.pop());
} else if (op == "vlineto") {
operandStack.pop();
stack.push(operandStack.pop());
stack.push(0);
} else {
next= false;
}
}
break;
}
while (stack.length)
operandStack.push(stack.pop());
operandStack.push(command);
};
/*
* Flatten the commands by interpreting the postscript code and replacing
* every 'callsubr', 'callothersubr' by the real commands.
* At the moment OtherSubrs are not fully supported and only otherSubrs 0-4
* as descrived in 'Using Subroutines' of 'Adobe Type 1 Font Format',
* chapter 8.
*/
this.flattenCharstring = function(aCharstring, aDefaultWidth, aNominalWidth, aSubrs) {
operandStack.clear();
executionStack.clear();
executionStack.push(aCharstring);
var leftSidebearing = 0;
var lastPoint = 0;
while (true) {
var obj = nextInStack();
if (IsBool(obj) || IsInt(obj) || IsNum(obj)) {
dump("Value: " + obj);
operandStack.push(obj);
} else if (IsString(obj)) {
dump("String: " + obj);
switch (obj) {
case "hsbw":
var charWidthVector = operandStack.pop();
leftSidebearing = operandStack.pop();
if (charWidthVector != aDefaultWidth)
operandStack.push(charWidthVector - aNominalWidth);
break;
case "setcurrentpoint":
case "dotsection":
case "seac":
case "sbw":
error(obj + " parsing is not implemented (yet)");
break;
case "vstem3":
operandStack.push("vstem");
break;
case "vstem":
//log(obj + " is not converted (yet?)");
operandStack.push("vstem");
break;
case "closepath":
case "return":
break;
case "hlineto":
case "vlineto":
case "rlineto":
case "rrcurveto":
aggregateCommand(obj);
break;
case "rmoveto":
var dy = operandStack.pop();
var dx = operandStack.pop();
if (leftSidebearing) {
dx += leftSidebearing;
leftSidebearing = 0;
}
operandStack.push(dx);
operandStack.push(dy);
operandStack.push("rmoveto");
break;
case "hstem":
case "hstem3":
var dy = operandStack.pop();
var y = operandStack.pop();
if (operandStack.peek() == "hstem" ||
operandStack.peek() == "hstem3")
operandStack.pop();
operandStack.push(y - lastPoint);
lastPoint = y + dy;
operandStack.push(dy);
operandStack.push("hstem");
break;
case "callsubr":
var index = operandStack.pop();
executionStack.push(aSubrs[index].slice());
break;
case "callothersubr":
// XXX need to be improved
var index = operandStack.pop();
var count = operandStack.pop();
var data = operandStack.pop();
if (index != 3)
log("callothersubr for index: " + index);
operandStack.push(3);
operandStack.push("callothersubr");
break;
case "endchar":
operandStack.push("endchar");
return operandStack.clone();
case "pop":
operandStack.pop();
break;
default:
operandStack.push(obj);
break;
}
}
}
}
};
var type1hack = false;
var Type1Font = function(aFontName, aFontFile) {
if (_Fonts[aFontName])
return;
_Fonts[aFontName] = true;
// All Type1 font program should begin with the comment %!
if (aFontFile.getByte() != 0x25 || aFontFile.getByte() != 0x21)
error("Invalid file header");
if (!type1hack) {
type1hack = true;
var start = Date.now();
var ASCIIStream = aFontFile.makeSubStream(0, aFontFile.dict.get("Length1"), aFontFile.dict);
var binaryStream = aFontFile.makeSubStream(aFontFile.dict.get("Length1"), aFontFile.dict.get("Length2"), aFontFile.dict);
this.parser = new Type1Parser(ASCIIStream, binaryStream);
var fontName = this.parser.parse();
this.convertToOTF(fontName);
var end = Date.now();
log("Time to parse font is:" + (end - start));
}
};
Type1Font.prototype = {
getDefaultWidths: function(aCharstrings) {
var defaultWidth = 0;
var defaultUsedCount = 0;
var widths = {};
for (var glyph in aCharstrings.map) {
var width = aCharstrings.get(glyph)[1];
var usedCount = (widths[width] || 0) + 1;
if (usedCount > defaultUsedCount) {
defaultUsedCount = usedCount;
defaultWidth = width;
}
widths[width] = usedCount;
}
defaultWidth = parseInt(defaultWidth);
var maxNegDistance = 0, maxPosDistance = 0;
for (var width in widths) {
var diff = width - defaultWidth;
if (diff < 0 && diff < maxNegDistance) {
maxNegDistance = diff;
} else if (diff > 0 && diff > maxPosDistance) {
maxPosDistance = diff;
}
}
return {
default: defaultWidth,
nominal: defaultWidth + (maxPosDistance + maxNegDistance) / 2
};
},
convertToOTF: function(aFontName) {
var font = Fonts.get(aFontName);
var charstrings = font.get("CharStrings")
var defaultWidths = this.getDefaultWidths(charstrings);
var defaultWidth = defaultWidths.default;
var nominalWidth = defaultWidths.nominal;
log("defaultWidth to used: " + defaultWidth);
log("nominalWidth to used: " + nominalWidth);
log("Hack nonimal:" + (nominalWidth = 615));
var glyphs = {};
var subrs = font.get("Private").get("Subrs");
var parser = new Type1Parser();
for (var glyph in charstrings.map) {
var charstring = charstrings.get(glyph);
glyphs[glyph] = parser.flattenCharstring(charstring, defaultWidth, nominalWidth, subrs);
//log("=================================== " + glyph + " ==============================");
//log(charstrings.get(glyph));
//log(flattenedCharstring);
//log(validationData[glyph]);
}
// Create a CFF font data
var cff = new Uint8Array(20000);
var currentOffset = 0;
// Font header (major version, minor version, header size, offset size)
var header = [0x01, 0x00, 0x04, 0x04];
currentOffset += header.length;
cff.set(header);
// Names Index
var nameIndex = this.createCFFIndexHeader([aFontName]);
cff.set(nameIndex, currentOffset);
currentOffset += nameIndex.length;
//Top Dict Index
var topDictIndex = [
0x00, 0x01, 0x01, 0x01, 0x29,
248, 27, 0, // version
248, 28, 1, // Notice
248, 29, 2, // FullName
248, 30, 3, // FamilyName
248, 20, 4, // Weigth
82, 251, 98, 250, 105, 249, 72, 5, // FontBBox
248, 136, 15, // charset (offset: 500)
28, 0, 0, 16, // Encoding (offset: 600)
248, 236, 17, // CharStrings
28, 0, 55, 28, 15, 160, 18 // Private (offset: 4000)
];
cff.set(topDictIndex, currentOffset);
currentOffset += topDictIndex.length;
// Strings Index
var stringsIndex = [
0x00, 0x04, 0x01,
0x01, 0x05, 0x06, 0x07, 0x08,
0x31, 0x2E, 0x030, 0x35, // 1.05
0x2B, // +
0x28, // (
0x29 // )
];
cff.set(stringsIndex, currentOffset);
currentOffset += stringsIndex.length;
// Global Subrs Index
var globalSubrsIndex = [
0x00, 0x00, 0x00
];
cff.set(globalSubrsIndex, currentOffset);
currentOffset += globalSubrsIndex.length;
// Fill the space between this and the charset by '1'
var empty = new Array(500 - currentOffset);
for (var i = 0; i < empty.length; i++)
empty[i] = 0x01;
cff.set(empty, currentOffset);
currentOffset += empty.length;
//Declare the letter 'C'
var charset = [
0x00, 0x00, 0x42
];
cff.set(charset, currentOffset);
currentOffset += charset.length;
// Fill the space between this and the charstrings data by '1'
var empty = new Array(600 - currentOffset);
for (var i = 0; i < empty.length; i++)
empty[i] = 0x01;
cff.set(empty, currentOffset);
currentOffset += empty.length;
// Encode the glyph and add it to the FUX
var charStringsIndex = [
0x00, 0x02, 0x01, 0x01, 0x03, 0x05,
0x40, 0x0E,
0xAF, 0x0E
];
cff.set(charStringsIndex, currentOffset);
currentOffset += charStringsIndex.length;
// Fill the space between this and the private dict data by '1'
var empty = new Array(4000 - currentOffset);
for (var i = 0; i < empty.length; i++)
empty[i] = 0x01;
cff.set(empty, currentOffset);
currentOffset += empty.length;
// Private Data
var privateData = [
248, 136, 20,
248, 251, 21,
119, 159, 248, 97, 159, 247, 87, 159, 6,
30, 10, 3, 150, 37, 255, 12, 9,
139, 12, 10,
172, 10,
172, 150, 143, 146, 150, 146, 12, 12,
247, 32, 11,
247, 10, 161, 147, 154, 150, 143, 12, 13,
139, 12, 14,
28, 0, 55, 19
];
cff.set(privateData, currentOffset);
currentOffset += privateData.length;
// Dump shit at the end of the file
var shit = [
0x00, 0x01, 0x01, 0x01,
0x13, 0x5D, 0x65, 0x64,
0x5E, 0x5B, 0xAF, 0x66,
0xBA, 0xBB, 0xB1, 0xB0,
0xB9, 0xBA, 0x65, 0xB2,
0x5C, 0x1F, 0x0B
];
cff.set(shit, currentOffset);
currentOffset += shit.length;
var file = new Uint8Array(cff, 0, currentOffset);
var parser = new Type2Parser();
parser.parse(new Stream(file));
var file64 = Base64Encoder.encode(file);
console.log(file64);
var data = [];
for (var i = 0; i < currentOffset; i++)
data.push(cff[i]);
log(data);
},
createCFFIndexHeader: function(aObjects, aIsByte) {
var data = [];
// First 2 bytes contains the number of objects contained into this index
var count = aObjects.length;
var bytes = integerToBytes(count, 2);
for (var i = 0; i < bytes.length; i++)
data.push(bytes[i]);
// Next byte contains the offset size use to reference object in the file
// Actually we're using 0x04 to be sure to be able to store everything
// without thinking of it while coding.
data.push(0x04);
// Add another offset after this one because we need a new offset
var relativeOffset = 1;
for (var i = 0; i < count + 1; i++) {
var bytes = integerToBytes(relativeOffset, 4);
for (var j = 0; j < bytes.length; j++)
data.push(bytes[j]);
if (aObjects[i])
relativeOffset += aObjects[i].length;
}
for (var i =0; i < count; i++) {
for (var j = 0; j < aObjects[i].length; j++)
data.push(aIsByte ? aObjects[i][j] : aObjects[i][j].charCodeAt(0));
}
return data;
}
};
function integerToBytes(aValue, aBytesCount) {
var bytes = [];
do {
bytes[--aBytesCount] = (aValue & 0xFF);
aValue = aValue >> 8;
} while (aBytesCount && aValue > 0);
return bytes;
};
function encodeNumber(aValue) {
var x = 0;
if (aValue >= -107 && aValue <= 107) {
return [aValue + 139];
} else if (aValue >= 108 && aValue <= 1131) {
x = aValue - 108;
return [
integerToBytes(x / 256 + 247, 1),
x % 256
];
} else if (aValue >= -1131 && aValue <= -108) {
x = Math.abs(aValue) - 108;
return [
integerToBytes(x / 256 + 251, 1),
x % 256
];
} else if (aValue >= -32768 && aValue <= 32767) {
return [
28,
integerToBytes(aValue >> 8, 1),
integerToBytes(aValue, 1)
];
} else if (aValue >= (-2147483647-1) && aValue <= 2147483647) {
return [
0xFF,
integerToBytes(aValue >> 24, 1),
integerToBytes(aValue >> 16, 1),
integerToBytes(aValue >> 8, 1),
integerToBytes(aValue, 1)
];
} else {
error("Value: " + aValue + " is not allowed");
}
}