# # osc-utilities.coffee # ## Intro # This file contains some lower-level utilities for OSC handling. # My guess is client code won't need this. If you do need this, you must # require coffee first, then write: # # require("coffee-script/register"); # osc-utils = require("osc/lib/osc-utilities"); # # See the comments in osc.coffee for more information about the structure of # the objects we're dealing with here. # # ## Dependencies # require the minimal binary packing utilities binpack = require "binpack" # ## Exported Functions # Utility for working with buffers. takes an array of buffers, # output one buffer with all of the array concatenated # # This is really only exported for TDD, but maybe it'll be useful # to someone else too. exports.concat = (buffers) -> if not IsArray buffers throw new Error "concat must take an array of buffers" for buffer in buffers if not Buffer.isBuffer(buffer) throw new Error "concat must take an array of buffers" sumLength = 0 sumLength += buffer.length for buffer in buffers destBuffer = new Buffer(sumLength) copyTo = 0 for buffer in buffers buffer.copy destBuffer, copyTo copyTo += buffer.length destBuffer # # Convert a javascript string into a node.js Buffer containing an OSC-String. # # str must not contain any \u0000 characters. # # `strict` is an optional boolean paramter that fails if the string is invalid # (i.e. contains a \u0000 character) exports.toOscString = (str, strict) -> if not (typeof str == "string") throw new Error "can't pack a non-string into an osc-string" # strip off any \u0000 characters. nullIndex = str.indexOf("\u0000") # if we're being strict, we can't allow strings with null characters if (nullIndex != -1 and strict) throw StrictError "Can't pack an osc-string that contains NULL characters" str = str[0...nullIndex] if nullIndex != -1 # osc-strings must have length divisible by 4 and end with at least one zero. for i in [0...(padding str)] str += "\u0000" # create a new buffer from the string. new Buffer(str) # # Try to split a buffer into a leading osc-string and the rest of the buffer, # with the following layout: # { string : "blah" rest : }. # # `strict`, as above, is an optional boolean parameter that defaults to false - # if it is true, then an invalid buffer will always return null. # exports.splitOscString = (buffer, strict) -> if not Buffer.isBuffer buffer throw StrictError "Can't split something that isn't a buffer" # extract the string rawStr = buffer.toString "utf8" nullIndex = rawStr.indexOf "\u0000" # the rest of the code doesn't apply if there's no null character. if nullIndex == -1 throw new Error "All osc-strings must contain a null character" if strict return {string:rawStr, rest:(new Buffer 0)} # extract the string. str = rawStr[0...nullIndex] # find the length of the string's buffer splitPoint = Buffer.byteLength(str) + padding(str) # in strict mode, don't succeed if there's not enough padding. if strict and splitPoint > buffer.length throw StrictError "Not enough padding for osc-string" # if we're in strict mode, check that all the padding is null if strict for i in [Buffer.byteLength(str)...splitPoint] if buffer[i] != 0 throw StrictError "Not enough or incorrect padding for osc-string" # return a split rest = buffer[splitPoint...(buffer.length)] {string: str, rest: rest} # This has similar semantics to splitOscString but works with integers instead. # bytes is the number of bytes in the integer, defaults to 4. exports.splitInteger = (buffer, type) -> type = "Int32" if not type? bytes = (binpack["pack" + type] 0).length if buffer.length < bytes throw new Error "buffer is not big enough for integer type" num = 0 # integers are stored in big endian format. value = binpack["unpack" + type] buffer[0...bytes], "big" rest = buffer[bytes...(buffer.length)] return {integer : value, rest : rest} exports.toIntegerBuffer = (number, type) -> type = "Int32" if not type? if typeof number isnt "number" throw new Error "cannot pack a non-number into an integer buffer" binpack["pack" + type] number, "big" # This mapping contains three fields for each type: # - representation : the javascript string representation of this type. # - split : a function to split a buffer into a decoded value and # the rest of the buffer. # - toArg : a function that takes the representation of the type and # outputs a buffer. oscTypeCodes = s : { representation : "string" split : (buffer, strict) -> # just pass it through to splitOscString split = exports.splitOscString buffer, strict {value : split.string, rest : split.rest} toArg : (value, strict) -> throw new Error "expected string" if typeof value isnt "string" exports.toOscString value, strict } i : { representation : "integer" split : (buffer, strict) -> split = exports.splitInteger buffer {value : split.integer, rest : split.rest} toArg : (value, strict) -> throw new Error "expected number" if typeof value isnt "number" exports.toIntegerBuffer value } t : { representation : "timetag" split : (buffer, strict) -> split = exports.splitInteger buffer, "UInt64" {value : split.integer, rest : split.rest} toArg : (value, strict) -> throw new Error "expected number" if typeof value isnt "number" exports.toIntegerBuffer value, "UInt64" } f : { representation : "float" split : (buffer, strict) -> value : (binpack.unpackFloat32 buffer[0...4], "big") rest : buffer[4...(buffer.length)] toArg : (value, strict) -> throw new Error "expected number" if typeof value isnt "number" binpack.packFloat32 value, "big" } d : { representation : "double" split : (buffer, strict) -> value : (binpack.unpackFloat64 buffer[0...8], "big") rest : buffer[8...(buffer.length)] toArg : (value, strict) -> throw new Error "expected number" if typeof value isnt "number" binpack.packFloat64 value, "big" } b : { representation : "blob" split : (buffer, strict) -> # not much to do here, first grab an 4 byte int from the buffer {integer : length, rest : buffer} = exports.splitInteger buffer {value : buffer[0...length], rest : buffer[length...(buffer.length)]} toArg : (value, strict) -> throw new Error "expected node.js Buffer" if not Buffer.isBuffer value size = exports.toIntegerBuffer value.length exports.concat [size, value] } T : { representation : "true" split : (buffer, strict) -> rest : buffer value : true toArg : (value, strict) -> throw new Error "true must be true" if not value and strict new Buffer 0 } F : { representation : "false" split : (buffer, strict) -> rest : buffer value : false toArg : (value, strict) -> throw new Error "false must be false" if value and strict new Buffer 0 } N : { representation : "null" split : (buffer, strict) -> rest : buffer value : null toArg : (value, strict) -> throw new Error "null must be false" if value and strict new Buffer 0 } I : { representation : "bang" split : (buffer, strict) -> rest : buffer value : "bang" toArg : (value, strict) -> new Buffer 0 } # simple function that converts a type code into it's javascript # string representation. exports.oscTypeCodeToTypeString = (code) -> oscTypeCodes[code]?.representation # simple function that converts a javascript string representation # into its OSC type code. exports.typeStringToOscTypeCode = (rep) -> for own code, {representation : str} of oscTypeCodes return code if str is rep return null exports.argToTypeCode = (arg, strict) -> # if there's an explicit type annotation, back-translate that. if arg?.type? and (typeof arg.type is 'string') and (code = exports.typeStringToOscTypeCode arg.type)? return code value = if arg?.value? then arg.value else arg # now, we try to guess the type. throw new Error 'Argument has no value' if strict and not value? # if it's a string, use 's' if typeof value is 'string' return 's' # if it's a number, use 'f' by default. if typeof value is 'number' return 'f' # if it's a buffer, use 'b' if Buffer.isBuffer(value) return 'b' #### These are 1.1 specific types. # if it's a boolean, use 'T' or 'F' if typeof value is 'boolean' if value then return 'T' else return 'F' # if it's null, use 'N' if value is null return 'N' throw new Error "I don't know what type this is supposed to be." # Splits out an argument from buffer. Same thing as splitOscString but # works for all argument types. exports.splitOscArgument = (buffer, type, strict) -> osctype = exports.typeStringToOscTypeCode type if osctype? oscTypeCodes[osctype].split buffer, strict else throw new Error "I don't understand how I'm supposed to unpack #{type}" # Create a buffer with the given javascript type exports.toOscArgument = (value, type, strict) -> osctype = exports.typeStringToOscTypeCode type if osctype? oscTypeCodes[osctype].toArg value, strict else throw new Error "I don't know how to pack #{type}" # # translates an OSC message into a javascript representation. # exports.fromOscMessage = (buffer, strict) -> # break off the address { string : address, rest : buffer} = exports.splitOscString buffer, strict # technically, addresses have to start with '/'. if strict and address[0] isnt '/' throw StrictError 'addresses must start with /' # if there's no type string, this is technically illegal, but # the specification says we should accept this until all # implementations that send message without a type string are fixed. # this will never happen, so we should accept this, even in # strict mode. return {address : address, args : []} if not buffer.length # if there's more data but no type string, we can't parse the arguments. {string : types, rest : buffer} = exports.splitOscString buffer, strict # if the first letter isn't a ',' this isn't a valid type so we can't # parse the arguments. if types[0] isnt ',' throw StrictError 'Argument lists must begin with ,' if strict return {address : address, args : []} # we don't need the comma anymore types = types[1..(types.length)] args = [] # we use this to build up array arguments. # arrayStack[-1] is always the currently contructing # array. arrayStack = [args] # grab each argument. for type in types # special case: we're beginning construction of an array. if type is '[' arrayStack.push([]) continue # special case: we've just finished constructing an array. if type is ']' if arrayStack.length <= 1 throw new StrictError "Mismatched ']' character." if strict else built = arrayStack.pop() arrayStack[arrayStack.length-1].push( type: 'array' value: built ) continue # by the standard, we have to ignore the whole message # if we don't understand an argument typeString = exports.oscTypeCodeToTypeString type if not typeString? throw new Error "I don't understand the argument code #{type}" arg = exports.splitOscArgument buffer, typeString, strict # consume the argument from the buffer buffer = arg.rest if arg? # add it to the list. arrayStack[arrayStack.length-1].push( type : typeString value : arg?.value ) if arrayStack.length isnt 1 and strict throw new StrictError "Mismatched '[' character" {address : address, args : args, oscType : "message"} # # Try to parse an OSC bundle into a javascript object. # exports.fromOscBundle = (buffer, strict) -> # break off the bundletag { string : bundleTag, rest : buffer} = exports.splitOscString buffer, strict # bundles have to start with "#bundle". if bundleTag isnt "\#bundle" throw new Error "osc-bundles must begin with \#bundle" # grab the 8 - bit timetag { integer : timetag, rest : buffer} = exports.splitInteger buffer, "UInt64" # convert each element. convertedElems = mapBundleList buffer, (buffer) -> exports.fromOscPacket buffer, strict return {timetag : timetag, elements : convertedElems, oscType : "bundle"} # # convert the buffer into a bundle or a message, depending on the first string # exports.fromOscPacket = (buffer, strict) -> if isOscBundleBuffer buffer, strict exports.fromOscBundle buffer, strict else exports.fromOscMessage buffer, strict # helper - is it an argument that represents an array? getArrayArg = (arg) -> if IsArray arg arg else if (arg?.type is "array") and (IsArray arg?.value) arg.value else if arg? and (not arg.type?) and (IsArray arg.value) arg.value else null # helper - converts an argument list into a pair of a type string and a # data buffer # argList must be an array!!! toOscTypeAndArgs = (argList, strict) -> osctype = "" oscargs = [] for arg in argList if (getArrayArg arg)? [thisType, thisArgs] = toOscTypeAndArgs (getArrayArg arg), strict osctype += "[" + thisType + "]" oscargs = oscargs.concat thisArgs continue typeCode = exports.argToTypeCode arg, strict if typeCode? value = arg?.value if value is undefined value = arg buff = exports.toOscArgument value, (exports.oscTypeCodeToTypeString typeCode), strict if buff? oscargs.push buff osctype += typeCode [osctype, oscargs] # # convert a javascript format message into an osc buffer # exports.toOscMessage = (message, strict) -> # the message must have addresses and arguments. address = if message?.address? then message.address else message if typeof address isnt "string" throw new Error "message must contain an address" args = message?.args if args is undefined args = [] # pack single args if not IsArray args old_arg = args args = [] args[0] = old_arg oscaddr = exports.toOscString address, strict [osctype, oscargs] = toOscTypeAndArgs args, strict osctype = "," + osctype # bundle everything together. allArgs = exports.concat oscargs # convert the type tag into an oscString. osctype = exports.toOscString osctype exports.concat [oscaddr, osctype, allArgs] # # convert a javascript format bundle into an osc buffer # exports.toOscBundle = (bundle, strict) -> # the bundle must have timetag and elements. if strict and not bundle?.timetag? throw StrictError "bundles must have timetags." timetag = bundle?.timetag ? 0 elements = bundle?.elements ? [] if not IsArray elements elemstr = elements elements = [] elements.push elemstr oscBundleTag = exports.toOscString "\#bundle" oscTimeTag = exports.toIntegerBuffer timetag, "UInt64" oscElems = [] for elem in elements try # try to convert this sub-element into a buffer buff = exports.toOscPacket elem, strict # okay, pack in the size. size = exports.toIntegerBuffer buff.length oscElems.push exports.concat [size, buff] catch e null allElems = exports.concat oscElems exports.concat [oscBundleTag, oscTimeTag, allElems] # convert a javascript format bundle or message into a buffer exports.toOscPacket = (bundleOrMessage, strict) -> # first, determine whether or not this is a bundle. if bundleOrMessage?.oscType? if bundleOrMessage.oscType is "bundle" return exports.toOscBundle bundleOrMessage, strict return exports.toOscMessage bundleOrMessage, strict # bundles have "timetags" and "elements" if bundleOrMessage?.timetag? or bundleOrMessage?.elements? return exports.toOscBundle bundleOrMessage, strict exports.toOscMessage bundleOrMessage, strict # # Helper function for transforming all messages in a bundle with a given message # transform. # exports.applyMessageTranformerToBundle = (transform) -> (buffer) -> # parse out the bundle-id and the tag, we don't want to change these { string, rest : buffer} = exports.splitOscString buffer # bundles have to start with "#bundle". if string isnt "\#bundle" throw new Error "osc-bundles must begin with \#bundle" bundleTagBuffer = exports.toOscString string # we know that the timetag is 8 bytes, we don't want to mess with it, # so grab it as a buffer. There is some subtle loss of precision with # the round trip from int64 to float64. timetagBuffer = buffer[0...8] buffer = buffer[8...buffer.length] # convert each element. elems = mapBundleList buffer, (buffer) -> exports.applyTransform( buffer, transform, exports.applyMessageTranformerToBundle transform ) totalLength = bundleTagBuffer.length + timetagBuffer.length totalLength += 4 + elem.length for elem in elems # okay, now we have to reconcatenate everything. outBuffer = new Buffer totalLength bundleTagBuffer.copy outBuffer, 0 timetagBuffer.copy outBuffer, bundleTagBuffer.length copyIndex = bundleTagBuffer.length + timetagBuffer.length for elem in elems lengthBuff = exports.toIntegerBuffer elem.length lengthBuff.copy outBuffer, copyIndex copyIndex += 4 elem.copy outBuffer, copyIndex copyIndex += elem.length outBuffer # # Applies a transformation function (that is, a function from buffers # to buffers) to each element of given osc-bundle or message. # # `buffer` is the buffer to transform, which must be a buffer of a full packet. # `messageTransform` is function from message buffers to message buffers # `bundleTransform` is an optional parameter for functions from bundle buffers # to bundle buffers. # if `bundleTransform` is not set, it defaults to just applying the # `messageTransform` to each message in the bundle. # exports.applyTransform = (buffer, mTransform, bundleTransform) -> if not bundleTransform? bundleTransform = exports.applyMessageTranformerToBundle mTransform if isOscBundleBuffer buffer bundleTransform buffer else mTransform buffer # Converts a javascript function from string to string to a function # from message buffer to message buffer, applying the function to the # parsed strings. # # We pre-curry this because we expect to use this with `applyMessageTransform` # above # exports.addressTransform = (transform) -> (buffer) -> # parse out the address {string, rest} = exports.splitOscString buffer # apply the function string = transform string # re-concatenate exports.concat [ exports.toOscString string rest ] # # Take a function that transform a javascript _OSC Message_ and # convert it to a function that transforms osc-buffers. # exports.messageTransform = (transform) -> (buffer) -> message = exports.fromOscMessage buffer exports.toOscMessage transform message ## Private utilities # # is it an array? # IsArray = Array.isArray # # An error that only throws when we're in strict mode. # StrictError = (str) -> new Error "Strict Error: " + str # this private utility finds the amount of padding for a given string. padding = (str) -> bufflength = Buffer.byteLength(str) 4 - (bufflength % 4) # # Internal function to check if this is a message or bundle. # isOscBundleBuffer = (buffer, strict) -> # both formats begin with strings, so we should just grab the front but not # consume it. {string} = exports.splitOscString buffer, strict return string is "\#bundle" # # Does something for each element in an array of osc-message-or-bundles, # each prefixed by a length (such as appears in osc-messages), then # return the result as an array. # # This is not exported because it doesn't validate the format and it's # not really a generally useful function. # # If a function throws on an element, we discard that element in the map # but we don't give up completely. # mapBundleList = (buffer, func) -> elems = while buffer.length # the length of the element is stored in an integer {integer : size, rest : buffer} = exports.splitInteger buffer # if the size is bigger than the packet, something's messed up, so give up. if size > buffer.length throw new Error( "Invalid bundle list: size of element is bigger than buffer") thisElemBuffer = buffer[0...size] # move the buffer to after the element we're just parsing. buffer = buffer[size...buffer.length] # record this element try func thisElemBuffer catch e null # remove all null from elements nonNullElems = [] for elem in elems (nonNullElems.push elem) if elem? nonNullElems