Comparison

The comparators (operators for comparison) in JavaScript are deeply broken, and it is not possible to teach them new tricks.

MochiKit exposes an extensible comparison facility as a simple compare(a, b) function, which should be used in lieu of JavaScript's operators whenever comparing objects other than numbers or strings (though you can certainly use compare for those, too!).

The compare function has the same signature and return value as a sort function for Array.prototype.sort, and is often used in that context.

Defining new comparators for the compare function to use is done by adding an entry to its AdapterRegistry with the registerComparator function.

Programmer Representation

JavaScript's default representation mechanism, toString, is notorious for having terrible default behavior. It's also very unwise to change that default, as other JavaScript code you may be using may depend on it.

It's also useful to separate the concept of a "string representation" and a "string representation for programmers", much like Python does with its str and repr protocols.

repr provides this programmer representation for JavaScript, in a way that doesn't require object prototype hacking: using an AdapterRegistry.

Objects that implement the repr protocol can either implement a .repr() or .__repr__() method, or they can simply have an adapter setup to generate programmer representations. By default, the registry provides nice representations for null, undefined, Array, and objects or functions with a NAME attribute that use the default toString. For objects that repr doesn't already understand, it simply defaults to toString, so it will integrate seamlessly with code that implements the idiomatic JavaScript toString method!

To define a programmer representation for your own objects, simply add a .repr() or .__repr__() method that returns a string. For objects that you didn't create (e.g., from a script you didn't write, or a built-in object), it is instead recommended that you create an adapter with registerRepr.

JSON Serialization

JSON [1], JavaScript Object Notation, is a widely used serialization format in the context of web development. It's extremely simple, lightweight, and fast. In its essence, JSON is a restricted subset of JavaScript syntax suitable for sending over the wire that can be unserialized with a simple eval. It's often used as an alternative to XML in "AJAX" contexts because it is compact, fast, and much simpler to use for most purposes.

To create a JSON serialization of any object, simply call serializeJSON() with that object. To unserialize a JSON string, simply call evalJSON() with the serialization.

In order of precedence, serializeJSON coerces the given argument into a JSON serialization:

1. Primitive types are returned as their JSON representation: undefined, string, number, boolean, null.
2. If the object has a __json__ or json method, then it is called with no arguments. If the result of this method is not the object itself, then the new object goes through rule processing again (e.g. it may return a string, which is then serialized in JSON format).
3. If the object is array-like (has a length property that is a number, and is not a function), then it is serialized as a JSON array. Each element will be processed according to these rules in order. Elements that can not be serialized (e.g. functions) will be replaced with undefined.
4. The jsonRegistry AdapterRegistry is consulted for an adapter for this object. JSON adapters take one argument (the object), and are expected to behave like a __json__ or json method (return another object to be serialized, or itself).
5. If no adapter is available, the object is enumerated and serialized as a JSON object (name:value pairs). All names are expected to be strings. Each value is serialized according to these rules, and if it can not be serialized (e.g. methods), then that name:value pair will be skipped.

Adapter Registries

MochiKit makes extensive use of adapter registries, which enable you to implement object-specific behaviors for objects that you do not necessarily want to modify, such as built-in objects. This is especially useful because JavaScript does not provide a method for hiding user-defined properties from for propName in obj enumeration.

AdapterRegistry is simply an encapsulation for an ordered list of "check" and "wrap" function pairs. Each AdapterRegistry instance should perform one function, but may have multiple ways to achieve that function based upon the arguments. One way to think of it is as a poor man's generic function, or multiple dispatch (on arbitrary functions, not just type!).

Check functions take one or more arguments, and return true if the argument list is suitable for the wrap function. Check functions should perform "cheap" checks of an object's type or contents, before the "expensive" wrap function is called.

Wrap functions take the same arguments as check functions and do some operation, such as creating a programmer representation or comparing both arguments.

Convenience Functions

Much of MochiKit.Base is there to simply remove the grunt work of doing generic JavaScript programming.

Need to take every property from one object and set them on another? No problem, just call update(dest, src)! What if you just wanted to update keys that weren't already set? Look no further than setdefault(dest, src[, ...]).

Want to return a mutable object, but don't want to suffer the consequences if the user mutates it? Just clone(it) and you'll get a copy-on-write clone. Cheaper than a copy!

Need to extend an Array with another array? Or even an array-like object such as a NodeList or the special arguments object? Even if you need to skip the first few elements of the source array-like object, it's no problem with extend(dstArray, srcArrayLike[, skip])!

Wouldn't it be convenient to have all of the JavaScript operators were available as functions somewhere? That's what the operators table is for, and it even comes with additional operators based on the compare function.

Need to walk some tree of objects and manipulate or find something in it? A DOM element tree perhaps? Use nodeWalk(node, visitor)!

There's plenty more, so check out the API Reference below.

Functional Programming

Functional programming constructs such as map and filter can save you a lot of time, because JavaScript iteration is error-prone and arduous. Writing less code is the best way to prevent bugs, and functional programming can help you do that.

MochiKit.Base ships with a few simple Array-based functional programming constructs, namely map and filter, and their "extended" brethren, xmap and xfilter.

map(func, arrayLike[, ...]) takes a function and an array-like object, and creates a new Array. The new Array is the result of func(element) for every element of arrayLike, much like the Array.prototype.map extension in Mozilla. However, MochiKit.Base takes that a step further and gives you the full blown Python version of map, which will take several array-like objects, and calls the function with one argument per given array-like, like this:

var arrayOne = [1, 2, 3, 4, 5];
var arrayTwo = [1, 5, 2, 4, 3];
var arrayThree = [5, 2, 1, 3, 4];
var biggestElements = map(objMax, arrayOne, arrayTwo, arrayThree);
assert( objEqual(biggestElements, [5, 5, 3, 4, 5]) );

filter(func, arrayLike[, self]) takes a function and an array-like object, and returns a new Array. This is basically identical to the Array.prototype.filter extension in Mozilla. self, if given, will be used as this in the context of func when called.

xmap and xfilter are just special forms of map and filter that accept a function as the first argument, and use the extra arguments as the array-like. Not terribly interesting, but a definite time-saver in some cases.

If you appreciate the functional programming facilities here, you should definitely check out MochiKit.Iter, which provides full blown iterators, MochiKit.Iter.range, MochiKit.Iter.reduce, and a near-complete port of Python's itertools [2] module, with some extra stuff thrown in for good measure!

Bound and Partial Functions

JavaScript's method-calling special form and lack of bound functions (functions that know what this should be) are one of the first stumbling blocks that programmers new to JavaScript face. The bind(func, self) method fixes that right up by returning a new function that calls func with the right this.

In order to take real advantage of all this fancy functional programming stuff, you're probably going to want partial application. This allows you to create a new function from an existing function that remembers some of the arguments. For example, if you wanted to compare a given object to a slew of other objects, you could do something like this:

compareWithOne = partial(compare, 1);
results = map(compareWithOne, [0, 1, 2, 3]);
assert( objEqual(results, [-1, 0, 1, 1]) );

One of the better uses of partial functions is in MochiKit.DOM, which is certainly a must-see for those of you creating lots of DOM elements with JavaScript!

Errors

NotFound:

A singleton error raised when no suitable adapter is found

Constructors

AdapterRegistry:

A registry to facilitate adaptation.

All check/wrap functions in this registry should be of the same arity.

AdapterRegistry.prototype.register(name, check, wrap[, override]):

The check function should return true if the given arguments are appropriate for the wrap function.

If override is given and true, the check function will be given highest priority. Otherwise, it will be the lowest priority adapter.

AdapterRegistry.prototype.match(obj[, ...]):

Find an adapter for the given arguments.

If no suitable adapter is found, throws NotFound.

AdapterRegistry.prototype.unregister(name):

Remove a named adapter from the registry

NamedError:

Convenience constructor for creating new errors (e.g. NotFound)

Functions

arrayEqual(self, arr):

Compare two arrays for equality, with a fast-path for length differences.

bind(func, self[, arg, ...]):

Return a copy of func bound to self. This means whenever and however the returned function is called, this will always reference the given self. func may be either a function object, or a string. If it is a string, then self[func] will be used, making these two statements equivalent:

bind("method", self);
bind(self.method, self);

Calling bind(func, self) on an already bound function will return a new function that is bound to the new self! If self is undefined, then the previous self is used. If self is null, then the this object is used (which may or may not be the global object). To force binding to the global object, you should pass it explicitly.

Additional arguments, if given, will be partially applied to the function. These three expressions are equivalent and return equally efficient functions (bind and partial share the same code path):

• bind(oldfunc, self, arg1, arg2)
• bind(partial(oldfunc, arg1, arg2), self)
• partial(bind(oldfunc, self), arg1, arg2)

bindMethods(self):

Bind all methods of self present on self to self, which gives you a semi-Pythonic sort of instance.

clone(obj):

Return a new object using obj as its prototype. Use this if you want to return a mutable object (e.g. instance state), but don't want the user to mutate it. If they do, it won't have any effect on the original obj.

Note that this is a shallow clone, so mutable properties will have to be cloned separately if you want to "protect" them.

compare(a, b):

Compare two objects in a sensible manner. Currently this is:

1. undefined and null compare equal to each other
2. undefined and null are less than anything else
3. If JavaScript says a == b, then we trust it
4. comparators registered with registerComparator are used to find a good comparator. Built-in comparators are currently available for Array-like and Date-like objects.
5. Otherwise hope that the built-in comparison operators do something useful, which should work for numbers and strings.
6. If neither a < b or a > b, then throw a TypeError

Returns what one would expect from a comparison function:

Value	Condition
0	a == b
1	a > b
-1	a < b

concat(lst[, ...]):

Concatenates all given array-like arguments and returns a new Array:

var lst = concat(["1","3","5"], ["2","4","6"]);
assert( lst.toString() == "1,3,5,2,4,6" );

counter(n=1):

Returns a function that will return a number one greater than the previous returned value, starting at n. For example:

nextId = counter()
assert( nextId() == 1 )
assert( nextId() == 2 )

For an iterator with this behavior, see MochiKit.Iter.count.

extend(self, obj[, skip]):

Mutate an array by extending it with an array-like obj, starting with the "skip" index of obj. If null is given as the initial array, a new one will be created.

This mutates and returns the given array, be warned.

evalJSON(aJSONString):

Unserialize a JSON [1] represenation of an object.

Note that this uses the eval function of the interpreter, and therefore trusts the contents of aJSONString to be safe. This is acceptable when the JSON and JavaScript application originate from the same server, but in other scenarios it may not be the appropriate security model. Currently, a validating JSON parser is beyond the scope of MochiKit, but there is one available from json.org [1].

filter(fn, lst):

Returns a new Array composed of all elements from lst where fn(lst[i]) returns a true value.

If fn is null, operator.truth will be used.

find(lst, value, start=0, end=lst.length):

Finds the index of value in the Array-like object lst using compare. The search starts at the index start, and ends at the index end - 1. If value is not found in lst, it will return -1.

For example:

assert( find([1, 2, 3, 2, 1], 2) == 1 )
assert( find([1, 2, 3, 2, 1], 2, 2) == 3 )

findIdentical(lst, value, start=0, end=lst.length):

Finds the index of value in the Array-like object lst using the === operator. The search starts at the index start, and ends at the index end - 1. If value is not found in lst, it will return -1.

You should use this function instead of find if lst may be comprised of objects for which no comparator is defined and all you care about is finding an identical object (e.g. the same instance), or if lst is comprised of just numbers or strings and performance is important.

For example:

assert( findIdentical([1, 2, 3, 2, 1], 2) == 1 )
assert( findIdentical([1, 2, 3, 2, 1], 2, 2) == 3 )

forward(name):

Returns a function that forwards a method call to this.name(...)

isArrayLike(obj[, ...]):

Returns true if all given arguments are Array-like (have a .length property and typeof(obj) == 'object')

isDateLike(obj[, ...]):

Returns true if all given arguments are Date-like (have a .getTime() method)

isNotEmpty(obj[, ...]):

Returns true if all the given Array-like or string arguments are not empty (obj.length > 0)

isNull(obj[, ...]):

Returns true if all arguments are null.

isUndefinedOrNull(obj[, ...]):

Returns true if all arguments are undefined or null

itemgetter(name):

Returns a function(obj) that returns obj[name]

items(obj):

Return an Array of [propertyName, propertyValue] pairs for the given obj (in the order determined by for propName in obj).

keyComparator(key[, ...]):

A comparator factory that compares a[key] with b[key]. e.g.:

var lst = ["a", "bbb", "cc"];
lst.sort(keyComparator("length"));
assert( lst.toString() == "a,cc,bbb" );

keys(obj):

Return an Array of the property names of an object (in the order determined by for propName in obj).

listMax(lst):

Return the largest element of an Array-like object, as determined by compare. This is a special form of listMinMax, specifically partial(listMinMax, 1).

listMin(lst):

Return the smallest element of an Array-like object, as determined by compare. This is a special form of listMinMax, specifically partial(listMinMax, -1).

listMinMax(which, lst):

If which == -1 then it will return the smallest element of the Array-like lst. This is also available as listMin(lst).

If which == 1 then it will return the largest element of the array-like lst. This is also available as listMax(list).

map(fn, lst[, ...]):

Return a new array composed of the results of fn(x) for every x in lst.

If fn is null, and only one sequence argument is given the identity function is used.

map(null, lst) -> lst.slice();

If fn is null, and more than one sequence is given as arguments, then the Array function is used, making it equivalent to zip.

map(null, p, q, ...)

-> zip(p, q, ...) -> [[p0, q0, ...], [p1, q1, ...], ...];

merge(obj[, ...]):

Create a new instance of Object that contains every property from all given objects. If a property is defined on more than one of the objects, the last property is used.

This is a special form of update(self, obj[, ...]), specifically, it is defined as partial(update, null).

nameFunctions(namespace):

Given a namespace with a NAME property, find all functions in it and give them nice NAME properties too (for use with repr). e.g.:

namespace = {
    NAME: "Awesome",
    Dude: function () {}
}
nameFunctions(namespace);
assert( namespace.Dude.NAME == 'Awesome.Dude' );

objEqual(a, b):

Compare the equality of two objects.

nodeWalk(node, visitor):

Non-recursive generic node walking function (e.g. for a DOM).

The walk order for nodeWalk is breadth first, meaning that all siblings will be visited before any children.

node:

The initial node to be searched.

visitor:

The visitor function, will be called as visitor(node), and should return an Array-like of nodes to be searched next (e.g. node.childNodes).

objMax(obj[, ...]):

Return the maximum object out of the given arguments. This is similar to listMax, except is uses the arguments instead of a given Array-like.

objMin(obj[, ...]):

Return the minimum object out of the given arguments. This is similar to listMin, except it uses the arguments instead of a given Array-like.

operator:

A table of JavaScript's operators for usage with map, filter, etc.

Unary Logic Operators:

Operator	Implementation	Description
truth(a)	!!a	Logical truth
lognot(a)	!a	Logical not
identity(a)	a	Logical identity

Unary Math Operators:

Operator	Implementation	Description
not(a)	~a	Bitwise not
neg(a)	-a	Negation

Binary Operators:

Operator	Implementation	Description
add(a, b)	a + b	Addition
sub(a, b)	a - b	Subtraction
div(a, b)	a / b	Division
mod(a, b)	a % b	Modulus
mul(a, b)	a * b	Multiplication
and(a, b)	a & b	Bitwise and
or(a, b)	a | b	Bitwise or
xor(a, b)	a ^ b	Bitwise exclusive or
lshift(a, b)	a << b	Bitwise left shift
rshift(a, b)	a >> b	Bitwise signed right shift
zrshfit(a, b)	a >>> b	Bitwise unsigned right shift

Built-in Comparators:

Operator	Implementation	Description
eq(a, b)	a == b	Equals
ne(a, b)	a != b	Not equals
gt(a, b)	a > b	Greater than
ge(a, b)	a >= b	Greater than or equal to
lt(a, b)	a < b	Less than
le(a, b)	a <= b	Less than or equal to

Extended Comparators (uses compare):

Operator	Implementation	Description
ceq(a, b)	compare(a, b) == 0	Equals
cne(a, b)	compare(a, b) != 0	Not equals
cgt(a, b)	compare(a, b) == 1	Greater than
cge(a, b)	compare(a, b) != -1	Greater than or equal to
clt(a, b)	compare(a, b) == -1	Less than
cle(a, b)	compare(a, b) != 1	Less than or equal to

Binary Logical Operators:

Operator	Implementation	Description
logand(a, b)	a && b	Logical and
logor(a, b)	a || b	Logical or
contains(a, b)	b in a	Has property (note order)

parseQueryString(encodedString[, useArrays=false]):

Parse a name=value pair URL query string into an object with a property for each pair. e.g.:

var args = parseQueryString("foo=value%20one&bar=two");
assert( args.foo == "value one" && args.bar == "two" );

If you expect that the query string will reuse the same name, then give true as a second argument, which will use arrays to store the values. e.g.:

var args = parseQueryString("foo=one&foo=two", true);
assert( args.foo[0] == "one" && args.foo[1] == "two" );

partial(func, arg[, ...]):

Return a partially applied function, e.g.:

addNumbers = function (a, b) {
    return a + b;
}

addOne = partial(addNumbers, 1);

assert(addOne(2) == 3);

partial is a special form of bind that does not alter the bound self (if any). It is equivalent to calling:

bind(func, undefined, arg[, ...]);

See the documentation for bind for more details about this facility.

queryString(names, values):

Creates a URL query string from a pair of array-like objects representing names and values. Each name=value pair will be URL encoded by urlEncode. name=value pairs with a value of undefined or null will be skipped. e.g.:

var keys = ["foo", "bar"];
var values = ["value one", "two"];
assert( queryString(keys, values) == "foo=value%20one&bar=two" );

Alternate form 1:

queryString(domElement)

If MochiKit.DOM is loaded, one argument is given, and that argument is either a string or has a nodeType property greater than zero, then names and values will be the result of MochiKit.DOM.formContents(domElement).

Alternate form 2:

queryString({name: value, ...})

Note that when using the alternate form, the order of the name=value pairs in the resultant query string is dependent on how the particular JavaScript implementation handles for (..in..) property enumeration.

When using the second alternate form, name=value pairs with typeof(value) == "function" are ignored. This is a workaround for the case where a poorly designed library has modified Object.prototype and inserted "convenience functions".

registerComparator(name, check, comparator[, override]):

Register a comparator for use with compare.

name should be a unique identifier describing the comparator.

check is a function(a, b) that returns true if a and b can be compared with comparator.

comparator is a function(a, b) that returns:

Value	Condition
0	a == b
1	a > b
-1	a < b

comparator is guaranteed to only be called if check(a, b) returns a true value.

If override is true, then it will be made the highest precedence comparator. Otherwise, the lowest.

registerJSON(name, check, simplifier[, override]):

Register a simplifier function for use with serializeJSON.

name should be a unique identifier describing the serialization.

check is a function(obj) that returns true if obj can can be simplified for serialization by simplifier.

simplifier is a function(obj) that returns a simpler object that can be further serialized by serializeJSON. For example, you could simplify Date-like objects to ISO 8601 timestamp strings with the following simplifier:

var simplifyDateAsISO = function (obj) {
    return toISOTimestamp(obj, true);
};
registerJSON("DateLike", isDateLike, simplifyDateAsISO);

simplifier is guaranteed to only be called if check(obj) returns a true value.

If override is true, then it will be made the highest precedence comparator. Otherwise, the lowest.

registerRepr(name, check, wrap[, override]):

Register a programmer representation function. repr functions should take one argument and return a string representation of it suitable for developers, primarily used when debugging.

If override is given, it is used as the highest priority repr, otherwise it will be used as the lowest.

repr(o):

Return a programmer representation for an object. See the Programmer Representation overview for more information about this function.

reverseKeyComparator(key):

A comparator factory that compares a[key] with b[key] in reverse. e.g.:

var lst = ["a", "bbb", "cc"];
lst.sort(reverseKeyComparator("length"));
assert(lst.toString() == "bbb,cc,aa");

serializeJSON(anObject):

Serialize any object in the JSON [1] format, see JSON Serialization for the coercion rules. For unserializable objects (functions that do not have an adapter, __json__ method, or json method), this will return undefined.

For those familiar with Python, JSON is similar in scope to pickle, but it can not handle recursive object graphs.

setdefault(self, obj[, ...]):

Mutate an object by adding all properties from other object(s) that it does not already have set.

If self is null, a new Object instance will be created and returned.

This mutates and returns the given self, be warned.

typeMatcher(typ[, ...]):

Given a set of types (as string arguments), returns a function(obj[, ...]) that will return true if the types of the given arguments are all members of that set.

update(self, obj[, ...]):

Mutate an object by replacing its key:value pairs with those from other object(s). Key:value pairs from later objects will overwrite those from earlier objects.

If null is given as the initial object, a new one will be created.

This mutates and returns the given object, be warned.

A version of this function that creates a new object is available as merge(a, b[, ...])

updatetree(self, obj[, ...]):

Mutate an object by replacing its key:value pairs with those from other object(s). If a given key has an object value in both self and obj, then this function will be called recursively, updating instead of replacing that object.

If null is given as the initial object, a new one will be created.

This mutates and returns the given object, be warned.

urlEncode(unencoded):

Converts a string into a URL-encoded string. Note that, in this implementation, spaces are converted to %20 instead of "+". e.g.:

assert( URLencode("1+2=2") == "1%2B2%3D2");

xfilter(fn, obj[, ...]):

Returns a new Array composed of the arguments where fn(obj) returns a true value.

If fn is null, operator.truth will be used.

xmap(fn, obj[, ...):

Return a new Array composed of fn(obj) for every obj given as an argument.

If fn is null, operator.identity is used.