Option 1
Easiest option, works in almost all cases, except that null
!==undefined
but they both are converted to JSON representation null
and considered equal:
function arraysEqual(a1,a2) {
/* WARNING: arrays must not contain {objects} or behavior may be undefined */
return JSON.stringify(a1)==JSON.stringify(a2);
}
(This might not work if your array contains objects. Whether this still works with objects depends on whether the JSON implementation sorts keys. For example, the JSON of {1:2,3:4}
may or may not be equal to {3:4,1:2}
; this depends on the implementation, and the spec makes no guarantee whatsoever. [2017 update: Actually the ES6 specification now guarantees object keys will be iterated in order of 1) integer properties, 2) properties in the order they were defined, then 3) symbol properties in the order they were defined. Thus IF the JSON.stringify implementation follows this, equal objects (in the === sense but NOT NECESSARILY in the == sense) will stringify to equal values. More research needed. So I guess you could make an evil clone of an object with properties in the reverse order, but I cannot imagine it ever happening by accident...] At least on Chrome, the JSON.stringify function tends to return keys in the order they were defined (at least that I've noticed), but this behavior is very much subject to change at any point and should not be relied upon. If you choose not to use objects in your lists, this should work fine. If you do have objects in your list that all have a unique id, you can do a1.map(function(x)}{return {id:x.uniqueId}})
. If you have arbitrary objects in your list, you can read on for option #2.)
This works for nested arrays as well.
It is, however, slightly inefficient because of the overhead of creating these strings and garbage-collecting them.
Option 2
More "proper" option, which you can override to deal with special cases (like regular objects and null/undefined and custom objects, if you so desire):
// generally useful functions
function type(x) { // does not work in general, but works on JSONable objects we care about... modify as you see fit
// e.g. type(/asdf/g) --> "[object RegExp]"
return Object.prototype.toString.call(x);
}
function zip(arrays) {
// e.g. zip([[1,2,3],[4,5,6]]) --> [[1,4],[2,5],[3,6]]
return arrays[0].map(function(_,i){
return arrays.map(function(array){return array[i]})
});
}
// helper functions
function allCompareEqual(array) {
// e.g. allCompareEqual([2,2,2,2]) --> true
// does not work with nested arrays or objects
return array.every(function(x){return x==array[0]});
}
function isArray(x){ return type(x)==type([]) }
function getLength(x){ return x.length }
function allTrue(array){ return array.reduce(function(a,b){return a&&b},true) }
// e.g. allTrue([true,true,true,true]) --> true
// or just array.every(function(x){return x});
function allDeepEqual(things) {
// works with nested arrays
if( things.every(isArray) )
return allCompareEqual(things.map(getLength)) // all arrays of same length
&& allTrue(zip(things).map(allDeepEqual)); // elements recursively equal
//else if( this.every(isObject) )
// return {all have exactly same keys, and for
// each key k, allDeepEqual([o1[k],o2[k],...])}
// e.g. ... && allTrue(objectZip(objects).map(allDeepEqual))
//else if( ... )
// extend some more
else
return allCompareEqual(things);
}
Demo:
allDeepEqual([ [], [], [] ])
true
allDeepEqual([ [1], [1], [1] ])
true
allDeepEqual([ [1,2], [1,2] ])
true
allDeepEqual([ [[1,2],[3]], [[1,2],[3]] ])
true
allDeepEqual([ [1,2,3], [1,2,3,4] ])
false
allDeepEqual([ [[1,2],[3]], [[1,2],[],3] ])
false
allDeepEqual([ [[1,2],[3]], [[1],[2,3]] ])
false
allDeepEqual([ [[1,2],3], [1,[2,3]] ])
false
To use this like a regular function, do:
function allDeepEqual2() {
return allDeepEqual([].slice.call(arguments));
}
Demo:
allDeepEqual2([[1,2],3], [[1,2],3])
true
Options 3
edit: It's 2016 and my previous overcomplicated answer was bugging me. This recursive, imperative "recursive programming 101" implementation keeps the code really simple, and furthermore fails at the earliest possible point (giving us efficiency). It also doesn't generate superfluous ephemeral datastructures (not that there's anything wrong with functional programming in general, but just keeping it clean here).
If we wanted to apply this to a non-empty arrays of arrays, we could do seriesOfArrays.reduce(arraysEqual).
This is its own function, as opposed to using Object.defineProperties to attach to Array.prototype, since that would fail with a key error if we passed in an undefined value (that is however a fine design decision if you want to do so).
This only answers OPs original question.
function arraysEqual(a,b) {
/*
Array-aware equality checker:
Returns whether arguments a and b are == to each other;
however if they are equal-lengthed arrays, returns whether their
elements are pairwise == to each other recursively under this
definition.
*/
if (a instanceof Array && b instanceof Array) {
if (a.length!=b.length) // assert same length
return false;
for(var i=0; i
Examples:
arraysEqual([[1,2],3], [[1,2],3])
true
arraysEqual([1,2,3], [1,2,3,4])
false
arraysEqual([[1,2],[3]], [[1,2],[],3])
false
arraysEqual([[1,2],[3]], [[1],[2,3]])
false
arraysEqual([[1,2],3], undefined)
false
arraysEqual(undefined, undefined)
true
arraysEqual(1, 2)
false
arraysEqual(null, null)
true
arraysEqual(1, 1)
true
arraysEqual([], 1)
false
arraysEqual([], undefined)
false
arraysEqual([], [])
true
If you wanted to apply this to JSON-like data structures with js Objects, you could do so. Fortunately we're guaranteed that all objects keys are unique, so iterate over the objects OwnProperties and sort them by key, then assert that both the sorted key-array is equal and the value-array are equal, and just recurse. We can extend this to include Maps as well (where the keys are also unique). (However if we extend this to Sets, we run into the tree isomorphism problem http://logic.pdmi.ras.ru/~smal/files/smal_jass08_slides.pdf - fortunately it's not as hard as general graph isomorphism; there is in fact an O(#vertices) algorithm to solve it, but it can get very complicated to do it efficiently. The pathological case is if you have a set made up of lots of seemingly-indistinguishable objects, but upon further inspection some of those objects may differ as you delve deeper into them. You can also work around this by using hashing to reject almost all cases.)
Option 4:
(continuation of 2016 edit)
This should work with most objects:
function deepEquals(a,b) {
if (a instanceof Array && b instanceof Array)
return arraysEqual(a,b);
if (Object.getPrototypeOf(a)===Object.prototype && Object.getPrototypeOf(b)===Object.prototype)
return objectsEqual(a,b);
if (a instanceof Map && b instanceof Map)
return mapsEqual(a,b);
if (a instanceof Set && b instanceof Set)
throw "Error: set equality by hashing not implemented."
if ((a instanceof ArrayBuffer || ArrayBuffer.isView(a)) && (b instanceof ArrayBuffer || ArrayBuffer.isView(b)))
return typedArraysEqual(a,b);
return a==b; // see note[1] -- IMPORTANT
}
function arraysEqual(a,b) {
if (a.length!=b.length)
return false;
for(var i=0; ia[k]), aKeys.map(k=>b[k]));
}
function mapsEqual(a,b) {
if (a.size!=b.size)
return false;
var aPairs = Array.from(a);
var bPairs = Array.from(b);
aPairs.sort((x,y) => x[0] x[0]
Demo (not extensively tested):
var nineTen = new Float32Array(2);
nineTen[0]=9; nineTen[1]=10;
deepEquals(
[[1,[2,3]], 4, {a:5,b:6}, new Map([['c',7],['d',8]]), nineTen],
[[1,[2,3]], 4, {b:6,a:5}, new Map([['d',8],['c',7]]), nineTen]
)
(sidenote: Maps are es6 dictionaries. I can't tell if they have O(1) or O(log(N)) lookup performance, but in any case they are 'ordered' in the sense that they keep track of the order in which key-value pairs were inserted into them. However, the semantic of whether two Maps should be equal if elements were inserted in a different order into them is ambiguous. I give a sample implementation below of a deepEquals that considers two maps equal even if elements were inserted into them in a different order.)
(note [1]: IMPORTANT: NOTION OF EQUALITY: You may want to override the noted line with a custom notion of equality, which you'll also have to change in the other functions anywhere it appears. For example, do you or don't you want NaN==NaN? By default this is not the case. There are even more weird things like 0=='0'. Do you consider two objects to be the same if and only if they are the same object in memory? See https://stackoverflow.com/a/5447170/711085 . You should document the notion of equality you use. )
You should be able to extend the above to WeakMaps, WeakSets. Not sure if it makes sense to extend to DataViews. Should also be able to extend to RegExps probably, etc.
As you extend it, you realize you do lots of unnecessary comparisons. This is where the type
function that I defined way earlier (solution #2) can come in handy; then you can dispatch instantly. Whether that is worth the overhead of (possibly? not sure how it works under the hood) string representing the type is up to you. You can just then rewrite the dispatcher, i.e. the function deepEquals
, to be something like:
var dispatchTypeEquals = {
number: function(a,b) {...a==b...},
array: function(a,b) {...deepEquals(x,y)...},
...
}
function deepEquals(a,b) {
var typeA = extractType(a);
var typeB = extractType(a);
return typeA==typeB && dispatchTypeEquals[typeA](a,b);
}