// @tag enterprise/** * @class Ext.data.amf.Packet * This class represents an Action Message Format (AMF) Packet. It contains all * the logic required to decode an AMF Packet from a byte array. * To decode a Packet, first construct a Packet: * * var packet = Ext.create('Ext.data.amf.Packet'); * * Then use the decode method to decode an AMF byte array: * * packet.decode(bytes); * * where "bytes" is a Uint8Array or an array of numbers representing the binary * AMF data. * * To access the decoded data use the #version, #headers, and #messages properties: * * console.log(packet.version, packet.headers, packet.messages); * * For more information on working with AMF data please refer to the * [AMF Guide](../guides/backend_connectors/amf.html). *//* global ActiveXObject */Ext.define('Ext.data.amf.Packet', function() { var twoPowN52 = Math.pow(2, -52), twoPow8 = Math.pow(2, 8), pos = 0, bytes, strings, objects, traits; return { /** * @property {Array} headers * @readonly * The decoded headers. Each header has the following properties: * * - `name`: String * The header name. Typically identifies a remote operation or method to * be invoked by this context header. * - `mustUnderstand`: Boolean * If `true` this flag instructs the endpoint to abort and generate an * error if the header is not understood. * - `byteLength`: Number * If the byte-length of a header is known it can be specified to optimize * memory allocation at the remote endpoint. * - `value`: Mixed * The header value */ /** * @property {Array} messages * @readonly * The decoded messages. Each message has the following properties: * * - `targetURI`: String * Describes which operation, function, or method is to be remotely * invoked. * - `responseURI`: String * A unique operation name * - `byteLength`: Number * Optional byte-length of the message body * - `body`: Mixed * The message body */ /** * @property {Number} version * @readonly * The AMF version number (0 or 3) */ /** * Mapping of AMF data types to the names of the methods responsible for * reading them. * @private */ typeMap: { // AMF0 mapping 0: { 0: 'readDouble', 1: 'readBoolean', 2: 'readAmf0String', 3: 'readAmf0Object', 5: 'readNull', 6: 'readUndefined', 7: 'readReference', 8: 'readEcmaArray', 10: 'readStrictArray', 11: 'readAmf0Date', 12: 'readLongString', 13: 'readUnsupported', 15: 'readAmf0Xml', 16: 'readTypedObject' }, // AMF3 mapping 3: { 0: 'readUndefined', 1: 'readNull', 2: 'readFalse', 3: 'readTrue', 4: 'readUInt29', 5: 'readDouble', 6: 'readAmf3String', 7: 'readAmf3Xml', 8: 'readAmf3Date', 9: 'readAmf3Array', 10: 'readAmf3Object', 11: 'readAmf3Xml', 12: 'readByteArray' } }, /** * Decodes an AMF btye array and sets the decoded data as the * Packet's #version, #headers, and #messages properties * @param {Array} byteArray A byte array containing the encoded AMF data. * @return {Ext.data.amf.Packet} this AMF Packet */ decode: function(byteArray) { var me = this, headers = me.headers = [], messages = me.messages = [], headerCount, messageCount; pos = 0; bytes = me.bytes = byteArray; // The strings array holds references to all of the deserialized // AMF3 strings for a given header value or message body so that // repeat instances of the same string can be deserialized by // reference strings = me.strings = []; // The objects array holds references to deserialized objects so // that repeat occurrences of the same object instance in the byte // array can be deserialized by reference. // If version is AMF0 this array holds anonymous objects, typed // objects, arrays, and ecma-arrays. // If version is AMF3 this array holds instances of Object, Array, XML, // XMLDocument, ByteArray, Date, and instances of user defined Classes objects = me.objects = []; // The traits array holds references to the "traits" (the // characteristics of objects that define a strong type such as the // class name and public member names) of deserialized AMF3 objects // so that if they are repeated they can be deserialized by reference. traits = me.traits = []; // The first two bytes of an AMF packet contain the AMF version // as an unsigned 16 bit integer. me.version = me.readUInt(2); // the next 2 bytes contain the header count for (headerCount = me.readUInt(2); headerCount--;) { headers.push({ name: me.readAmf0String(), mustUnderstand: me.readBoolean(), byteLength: me.readUInt(4), value: me.readValue() }); // reset references (reference indices are local to each header) strings = me.strings = []; objects = me.objects = []; traits = me.traits = []; } // The 2 bytes immediately after the header contain the message count. for (messageCount = me.readUInt(2); messageCount--;) { messages.push({ targetURI: me.readAmf0String(), responseURI: me.readAmf0String(), byteLength: me.readUInt(4), body: me.readValue() }); // reset references (reference indices are local to each message) strings = me.strings = []; objects = me.objects = []; traits = me.traits = []; } // reset the pointer pos = 0; // null the bytes array and reference arrays to free up memory. bytes = strings = objects = traits = me.bytes = me.strings = me.objects = me.traits = null; return me; }, /** * Decodes an AMF3 byte array and that has one value and returns it. * Note: Destroys previously stored data in this Packet. * @param {Array} byteArray A byte array containing the encoded AMF data. * @return {Object} the decoded object */ decodeValue: function(byteArray) { var me = this; bytes = me.bytes = byteArray; // reset the pointer pos = 0; // The first two bytes of an AMF packet contain the AMF version // as an unsigned 16 bit integer. me.version = 3; // The strings array holds references to all of the deserialized // AMF3 strings for a given header value or message body so that // repeat instances of the same string can be deserialized by // reference strings = me.strings = []; // The objects array holds references to deserialized objects so // that repeat occurrences of the same object instance in the byte // array can be deserialized by reference. // If version is AMF0 this array holds anonymous objects, typed // objects, arrays, and ecma-arrays. // If version is AMF3 this array holds instances of Object, Array, XML, // XMLDocument, ByteArray, Date, and instances of user defined Classes objects = me.objects = []; // The traits array holds references to the "traits" (the // characteristics of objects that define a strong type such as the // class name and public member names) of deserialized AMF3 objects // so that if they are repeated they can be deserialized by reference. traits = me.traits = []; // read one value and return it return me.readValue(); }, /** * Parses an xml string and returns an xml document * @private * @param {String} xml */ parseXml: function(xml) { var doc; if (window.DOMParser) { doc = (new DOMParser()).parseFromString(xml, "text/xml"); } else { doc = new ActiveXObject("Microsoft.XMLDOM"); doc.loadXML(xml); } return doc; }, /** * Reads an AMF0 date from the byte array * @private */ readAmf0Date: function() { var date = new Date(this.readDouble()); // An AMF0 date type ends with a 16 bit integer time-zone, but // according to the spec time-zone is "reserved, not supported, // should be set to 0x000". pos += 2; // discard the time zone return date; }, /** * Reads an AMF0 Object from the byte array * @private */ readAmf0Object: function(obj) { var me = this, key; obj = obj || {}; // add the object to the objects array so that the AMF0 reference // type decoder can refer to it by index if needed. objects.push(obj); // An AMF0 object consists of a series of string keys and variable- // type values. The end of the series is marked by an empty string // followed by the object-end marker (9). while ((key = me.readAmf0String()) || bytes[pos] !== 9) { obj[key] = me.readValue(); } // move the pointer past the object-end marker pos++; return obj; }, /** * Reads an AMF0 string from the byte array * @private */ readAmf0String: function() { // AMF0 strings begin with a 16 bit byte-length header. return this.readUtf8(this.readUInt(2)); }, readAmf0Xml: function() { return this.parseXml(this.readLongString()); }, readAmf3Array: function() { var me = this, header = me.readUInt29(), count, key, array, i; // AMF considers Arrays in two parts, the dense portion and the // associative portion. The binary representation of the associative // portion consists of name/value pairs (potentially none) terminated // by an empty string. The binary representation of the dense portion // is the size of the dense portion (potentially zero) followed by an // ordered list of values (potentially none). if (header & 1) { // If the first (low) bit is a 1 read an array instance. The // remaining 1-28 bits are used to encode the length of the // dense portion of the array. count = (header >> 1); // First read the associative portion of the array (if any). If // there is an associative portion, the array will be read as a // javascript object, otherwise it will be a javascript array. key = me.readAmf3String(); if (key) { // First key is not an empty string - this is an associative // array. Read keys and values from the byte array until // we get to an empty string key array = {}; objects.push(array); do { array[key] = me.readValue(); } while ((key = me.readAmf3String())); // The dense portion of the array is then read into the // associative object, keyed by ordinal index. for (i = 0; i < count; i++) { array[i] = me.readValue(); } } else { // First key is an empty string - this is an array with // ordinal indices. array = []; objects.push(array); for (i = 0; i < count; i++) { array.push(me.readValue()); } } } else { // If the first (low) bit is a 0 read an array reference. The // remaining 1-28 bits are used to encode the reference index array = objects[header >> 1]; } return array; }, /** * Reads an AMF3 date from the byte array * @private */ readAmf3Date: function() { var me = this, header = me.readUInt29(), date; if (header & 1) { // If the first (low) bit is a 1, this is a date instance. date = new Date(me.readDouble()); objects.push(date); } else { // If the first (low) bit is a 0, this is a date reference. // The remaining 1-28 bits encode the reference index date = objects[header >> 1]; } return date; }, /** * Reads an AMF3 object from the byte array * @private */ readAmf3Object: function() { var me = this, header = me.readUInt29(), members = [], headerLast3Bits, memberCount, className, dynamic, objectTraits, obj, key, klass, i; // There are 3 different types of object headers, distinguishable // by the 1-3 least significant bits. All object instances have // a 1 in the low bit position, while references have a 0: // // 0 : object reference // 011 : traits // 01 : traits-ref // 111 : traits-ext if (header & 1) { // first (low) bit of 1, denotes an encoded object instance // The next string is the class name. headerLast3Bits = (header & 0x07); if (headerLast3Bits === 3) { // If the 3 least significant bits of the header are "011" // then trait information follows. className = me.readAmf3String(); // A 1 in the header's 4th least significant byte position // indicates that dynamic members may follow the sealed // members. dynamic = !!(header & 0x08); // Shift off the 4 least significant bits, and the remaining // 1-25 bits encode the number of sealed member names. Read // as many strings from the byte array as member names. memberCount = (header >> 4); for (i = 0; i < memberCount; i++) { members.push(me.readAmf3String()); } objectTraits = { className: className, dynamic: dynamic, members: members }; // An objects traits are cached in the traits array enabling // the traits for a given class to only be encoded once for // a series of instances. traits.push(objectTraits); } else if ((header & 0x03) === 1) { // If the 2 least significant bits are "01", then a traits // reference follows. The remaining 1-27 bits are used // to encode the trait reference index. objectTraits = traits[header >> 2]; className = objectTraits.className; dynamic = objectTraits.dynamic; members = objectTraits.members; memberCount = members.length; } else if (headerLast3Bits === 7) { // if the 3 lease significant bits are "111" then // externalizable trait data follows // TODO: implement externalizable traits } if (className) { klass = Ext.ClassManager.getByAlias('amf.' + className); obj = klass ? new klass() : { $className: className }; } else { obj = {}; } objects.push(obj); // read the sealed member values for (i = 0; i < memberCount; i++) { obj[members[i]] = me.readValue(); } if (dynamic) { // If the dynamic flag is set, dynamic members may follow // the sealed members. Read key/value pairs until we // encounter an empty string key signifying the end of the // dynamic members. while ((key = me.readAmf3String())) { obj[key] = me.readValue(); } } // finally, check if we need to convert this class if ((!klass) && this.converters[className]) { obj = this.converters[className](obj); } } else { // If the first (low) bit of the header is a 0, this is an // object reference. The remaining 1-28 significant bits are // used to encode an object reference index. obj = objects[header >> 1]; } return obj; }, /** * Reads an AMF3 string from the byte array * @private */ readAmf3String: function() { var me = this, header = me.readUInt29(), value; if (header & 1) { // If the first (low) bit is a 1, this is a string literal. // Discard the low bit. The remaining 1-28 bits are used to // encode the string's byte-length. value = me.readUtf8(header >> 1); if (value) { // the emtpy string is never encoded by reference strings.push(value); } return value; } else { // If the first (low) bit is a 0, this is a string reference. // Discard the low bit, then look up and return the reference // from the strings array using the remaining 1-28 bits as the // index. return strings[header >> 1]; } }, /** * Reads an AMF3 XMLDocument type or XML type from the byte array * @private */ readAmf3Xml: function() { var me = this, header = me.readUInt29(), doc; if (header & 1) { // If the first (low) bit is a 1, this is an xml instance. The // remaining 1-28 bits encode the byte-length of the xml string. doc = me.parseXml(me.readUtf8(header >> 1)); objects.push(doc); } else { // if the first (low) bit is a 1, this is an xml reference. The // remaining 1-28 bits encode the reference index. doc = objects[header >> 1]; } return doc; }, /** * Reads an AMF0 boolean from the byte array * @private */ readBoolean: function() { return !!bytes[pos++]; }, /** * Reads an AMF3 ByteArray type from the byte array * @private */ readByteArray: function() { var header = this.readUInt29(), byteArray, end; if (header & 1) { // If the first (low) bit is a 1, this is a ByteArray instance. // The remaining 1-28 bits encode the ByteArray's byte-length. end = pos + (header >> 1); // Depending on the browser, "bytes" may be either a Uint8Array // or an Array. Uint8Arrays don't have Array methods, so // we have to use Array.prototype.slice to get the byteArray byteArray = Array.prototype.slice.call(bytes, pos, end); objects.push(byteArray); // move the pointer to the first byte after the byteArray that // was just read pos = end; } else { // if the first (low) bit is a 1, this is a ByteArray reference. // The remaining 1-28 bits encode the reference index. byteArray = objects[header >> 1]; } return byteArray; }, /** * Reads a IEEE 754 double-precision binary floating-point number * @private */ readDouble: function() { var byte1 = bytes[pos++], byte2 = bytes[pos++], // the first bit of byte1 is the sign (0 = positive, 1 = negative. // We read this bit by shifting the 7 least significant bits of // byte1 off to the right. sign = (byte1 >> 7) ? -1 : 1, // the exponent takes up the next 11 bits. exponent = // extract the 7 least significant bits from byte1 and then // shift them left by 4 bits to make room for the 4 remaining // bits from byte 2 (((byte1 & 0x7F) << 4) | // add the 4 most significant bits from byte 2 to complete // the exponent (byte2 >> 4)), // the remaining 52 bits make up the significand. read the 4 // least significant bytes of byte 2 to begin the significand significand = (byte2 & 0x0F), // The most significant bit of the significand is always 1 for // a normalized number, therefore it is not stored. This bit is // referred to as the "hidden bit". The true bit width of the // significand is 53 if you include the hidden bit. An exponent // of 0 indicates that this is a subnormal number, and subnormal // numbers always have a 0 hidden bit. hiddenBit = exponent ? 1 : 0, i = 6; // The operands of all bitwise operators in javascript are converted // to signed 32 bit integers. It is therefore impossible to construct // the 52 bit significand by repeatedly shifting its bits and then // bitwise OR-ing the result with the the next byte. To work around // this issue, we repeatedly multiply the significand by 2^8 which // produces the same result as (significand << 8), then we add the // next byte, which has the same result as a bitwise OR. while (i--) { significand = (significand * twoPow8) + bytes[pos++]; } if (!exponent) { if (!significand) { // if both exponent and significand are 0, the number is 0 return 0; } // If the exponent is 0, but the significand is not 0, this // is a subnormal number. Subnormal numbers are encoded with a // biased exponent of 0, but are interpreted with the value of // the smallest allowed exponent, which is one greater. exponent = 1; } // 0x7FF (2047) is a special exponent value that represents infinity // if the significand is 0, and NaN if the significand is not 0 if (exponent === 0x7FF) { return significand ? NaN : (Infinity * sign); } return sign * // The exponent is encoded using an offset binary // representation with the zero offset being 0x3FF (1023), // so we have to subtract 0x3FF to get the true exponent Math.pow(2, exponent - 0x3FF) * // convert the significand to its decimal value by multiplying // it by 2^52 and then add the hidden bit (hiddenBit + twoPowN52 * significand); }, /** * Reads an AMF0 ECMA Array from the byte array * @private */ readEcmaArray: function() { // An ecma array type is encoded exactly like an anonymous object // with the exception that it has a 32 bit "count" at the beginning. // We handle emca arrays by just throwing away the count and then // letting the object decoder handle the rest. pos += 4; return this.readAmf0Object(); }, /** * Returns false. Used for reading the false type * @private */ readFalse: function() { return false; }, /** * Reads a long string (longer than 65535 bytes) from the byte array * @private */ readLongString: function() { // long strings begin with a 32 bit byte-length header. return this.readUtf8(this.readUInt(4)); }, /** * Returns null. Used for reading the null type * @private */ readNull: function() { return null; }, /** * Reads a reference from the byte array. Reference types are used to * avoid duplicating data if the same instance of a complex object (which * is defined in AMF0 as an anonymous object, typed object, array, or * ecma-array) is included in the data more than once. * @private */ readReference: function() { // a reference type contains a single 16 bit integer that represents // the index of an already deserialized object in the objects array return objects[this.readUInt(2)]; }, /** * Reads an AMF0 strict array (an array with ordinal indices) * @private */ readStrictArray: function() { var me = this, len = me.readUInt(4), arr = []; objects.push(arr); while (len--) { arr.push(me.readValue()); } return arr; }, /** * Returns true. Used for reading the true type * @private */ readTrue: Ext.returnTrue, /** * Reads an AMF0 typed object from the byte array * @private */ readTypedObject: function() { var me = this, className = me.readAmf0String(), klass, instance, modified; klass = Ext.ClassManager.getByAlias('amf.' + className); // if there is no klass, mark the classname for easier parsing of returned results instance = klass ? new klass() : { $className: className }; modified = me.readAmf0Object(instance); // check if we need to convert this class if ((!klass) && this.converters[className]) { modified = this.converters[className](instance); } return modified; }, /** * Reads an unsigned integer from the byte array * @private * @param {Number} byteCount the number of bytes to read, e.g. 2 to read * a 16 bit integer, 4 to read a 32 bit integer, etc. * @return {Number} */ readUInt: function(byteCount) { var i = 1, result; // read the first byte result = bytes[pos++]; // if this is a multi-byte int, loop over the remaining bytes for (; i < byteCount; ++i) { // shift the result 8 bits to the left and add the next byte. result = (result << 8) | bytes[pos++]; } return result; }, /** * Reads an unsigned 29-bit integer from the byte array. * AMF 3 makes use of a special compact format for writing integers to * reduce the number of bytes required for encoding. As with a normal * 32-bit integer, up to 4 bytes are required to hold the value however * the high bit of the first 3 bytes are used as flags to determine * whether the next byte is part of the integer. With up to 3 bits of * the 32 bits being used as flags, only 29 significant bits remain for * encoding an integer. This means the largest unsigned integer value * that can be represented is 2^29-1. * * (hex) : (binary) * 0x00000000 - 0x0000007F : 0xxxxxxx * 0x00000080 - 0x00003FFF : 1xxxxxxx 0xxxxxxx * 0x00004000 - 0x001FFFFF : 1xxxxxxx 1xxxxxxx 0xxxxxxx * 0x00200000 - 0x3FFFFFFF : 1xxxxxxx 1xxxxxxx 1xxxxxxx xxxxxxxx * @private * @return {Number} */ readUInt29: function() { var value = bytes[pos++], nextByte; if (value & 0x80) { // if the high order bit of the first byte is a 1, the next byte // is also part of this integer. nextByte = bytes[pos++]; // remove the high order bit from both bytes before combining them value = ((value & 0x7F) << 7) | (nextByte & 0x7F); if (nextByte & 0x80) { // if the high order byte of the 2nd byte is a 1, then // there is a 3rd byte nextByte = bytes[pos++]; // remove the high order bit from the 3rd byte before // adding it to the value value = (value << 7) | (nextByte & 0x7F); if (nextByte & 0x80) { // 4th byte is also part of the integer nextByte = bytes[pos++]; // use all 8 bits of the 4th byte value = (value << 8) | nextByte; } } } return value; }, /** * @method * Returns undefined. Used for reading the undefined type * @private */ readUndefined: Ext.emptyFn, /** * @method * Returns undefined. Used for reading the unsupported type * @private */ readUnsupported: Ext.emptyFn, /** * Reads a UTF-8 string from the byte array * @private * @param {Number} byteLength The number of bytes to read * @return {String} */ readUtf8: function(byteLength) { var end = pos + byteLength, // the string's end position chars = [], charCount = 0, maxCharCount = 65535, charArrayCount = 1, result = [], i = 0, charArrays, byteCount, charCode; charArrays = [chars]; // UTF-8 characters may be encoded using 1-4 bytes. The number of // bytes that a character consumes is determined by reading the // leading byte. Values 0-127 in the leading byte indicate a single- // byte ASCII-compatible character. Values 192-223 (bytes with "110" // in the high-order position) indicate a 2-byte character, values // 224-239 (bytes with "1110" in the high-order position) indicate a // 3-byte character, and values 240-247 (bytes with "11110" in the // high-order position) indicate a 4-byte character. The remaining // bits of the leading byte hold bits of the encoded character, with // leading zeros if necessary. // // The continuation bytes all have "10" in the high-order position, // which means only the 6 least significant bits of continuation // bytes are available to hold the bits of the encoded character. // // The following table illustrates the binary format of UTF-8 // characters: // // Bits Byte 1 Byte 2 Byte 3 Byte 4 // ----------------------------------------------------- // 7 0xxxxxxx // 11 110xxxxx 10xxxxxx // 16 1110xxxx 10xxxxxx 10xxxxxx // 21 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx while (pos < end) { // read a byte from the byte array - if the byte's value is less // than 128 we are dealing with a single byte character charCode = bytes[pos++]; if (charCode > 127) { // if the byte's value is greater than 127 we are dealing // with a multi-byte character. if (charCode > 239) { // a leading-byte value greater than 239 means this is a // 4-byte character byteCount = 4; // Use only the 3 least-significant bits of the leading // byte of a 4-byte character. This is achieved by // applying the following bit mask: // (charCode & 0x07) // which is equivalent to: // 11110xxx (the byte) // AND 00000111 (the mask) charCode = (charCode & 0x07); } else if (charCode > 223) { // a leading-byte value greater than 223 but less than // 240 means this is a 3-byte character byteCount = 3; // Use only the 4 least-significant bits of the leading // byte of a 3-byte character. This is achieved by // applying the following bit mask: // (charCode & 0x0F) // which is equivalent to: // 1110xxxx (the byte) // AND 00001111 (the mask) charCode = (charCode & 0x0F); } else { // a leading-byte value less than 224 but (implicitly) // greater than 191 means this is a 2-byte character byteCount = 2; // Use only the 5 least-significant bits of the first // byte of a 2-byte character. This is achieved by // applying the following bit mask: // (charCode & 0x1F) // which is equivalent to: // 110xxxxx (the byte) // AND 00011111 (the mask) charCode = (charCode & 0x1F); } while (--byteCount) { // get one continuation byte. then strip off the leading // "10" by applying the following bit mask: // (b & 0x3F) // which is equialent to: // 10xxxxxx (the byte) // AND 00111111 (the mask) // That leaves 6 remaining bits on the continuation byte // which are concatenated onto the character's bits charCode = ((charCode << 6) | (bytes[pos++] & 0x3F)); } } chars.push(charCode); if (++charCount === maxCharCount) { charArrays.push(chars = []); charCount = 0; charArrayCount ++; } } // At this point we end up with an array of char arrays, each char // array being no longer than 65,535 characters, the fastest way to // turn these char arrays into strings is to pass them as the // arguments to fromCharCode (fortunately all currently supported // browsers can handle at least 65,535 function arguments). for (; i < charArrayCount; i++) { // create a result array containing the strings converted from // the individual character arrays. result.push(String.fromCharCode.apply(String, charArrays[i])); } return result.join(''); }, /** * Reads an AMF "value-type" from the byte array. Automatically detects * the data type by reading the "type marker" from the first byte after * the pointer. * @private */ readValue: function() { var me = this, marker = bytes[pos++]; // With the introduction of AMF3, a special type marker was added to // AMF0 to signal a switch to AMF3 serialization. This allows a packet // to start out in AMF 0 and switch to AMF 3 on the first complex type // to take advantage of the more the efficient encoding of AMF 3. if (marker === 17) { // change the version to AMF3 when we see a 17 marker me.version = 3; marker = bytes[pos++]; } return me[me.typeMap[me.version][marker]](); }, /** * Converters used in converting specific typed Flex classes to JavaScript usable form. * @private */ converters: { 'flex.messaging.io.ArrayCollection': function(obj) { // array collections have a source var that contains the actual data return obj.source || []; } } };});