EDIT: Due to the fact that any recursive solution will have a limit to the size of the array you can create... I made another solution in my
PJs @ GitHub library. This one runs at pseudo-instant speed and can create and manage a multidimensional array of any size, any structure, with any dimensions at any branch. It also can simulate prefilling and/or use a node object of custom design. Check it out here: https://github.com/PimpTrizkit/PJs/wiki/14.-Complex-Multidimensional-Object--(pCMO.js)
Using a modified version of jfabrizio's solution:
function createNDimArray(dimensions) {
var t, i = 0, s = dimensions[0], arr = new Array(s);
if ( dimensions.length < 3 ) for ( t = dimensions[1] ; i < s ; ) arr[i++] = new Array(t);
else for ( t = dimensions.slice(1) ; i < s ; ) arr[i++] = createNDimArray(t);
return arr;
}
Usages:
var arr = createNDimArray([3, 2, 3]);
// arr = [[[,,],[,,]],[[,,],[,,]],[[,,],[,,]]]
console.log(arr[2][1]); // in FF: Array [ <3 empty slots> ]
console.log("Falsy = " + (arr[2][1][0]?true:false) ); // Falsy = false
I found this to be quite a bit faster. I might stretch to say that it could be the fastest way possible to generate a N Dimensional array in Javascript. This refactoring above had some good speed increases. But, the best speed increase came from not prefilling, of course. This version doesn't prefill the array. It only returns a fully created N dimensional array of Ns lengths where the last level is just an empty array. I would hope that arr[x][y][z]?arr[x][y][z]:null
is sufficient if you really need the null
value. It is for my uses. :)
If you need prefilling, use his original version.
And, if you don't really care about what I did; then stop reading.
Want more geek talk? A little something about recursion for those learning out there. Alright here are the tactics. When doing deep recursion, keep in mind the final level. Its where most of the work is done. In this case its the Nth dimension, literally. This is your "payload", the rest is logistics. In jfab's function, when dimensions.length
gets to 1
, its last dimension, its in the Nth dimension and performs the payload. Which is to create the array of nulls, or in my case, an empty array. Since the recursion gets so deep each dimension is a factor of the last one. By the time you get to the Nth dimension you will have a lot of function calls and logistics gets cumbersome to the computer. And at the Nth dimension you will call your base recursion function (createNDimArray
in our case) more times for the payload than you will for logistics. Now, as in jfab's original solution, putting the execution of the payload as the very first thing you do in recursion (if possible) is usually a good thing, especially if its simple. Here, by making the payload a building of the final 2D array (instead of just a 1D array by simply returning a new Array()
only). Then the excessive amount of function calls now don't have to happen at this level. Now, of course, if you want to prefill the array, then this shortcut doesn't always help. But more to the point, prefilling the array would be the appropriate payload. By not visiting every item on the Nth dimension we have effectively removed it. That way there is one less level of function calls and basically the Nth dimension's payload is actually done on the N-1 th Dimension. And we are never calling the recursive function again just to deliver the new Array()
. Unfortunately, the call to new Array(x)
(in general) doesn't see it that way. Its execution time does increase with a larger x
. Which is effectively still visiting every item in the Nth Dimension, but now we do it only once and with native code and wrapped in a tight and light loop. Now we require that createNDimArray
can only be called with N > 1, ie never used to create 1D arrays. Theoretically you could require a larger N, and unroll even more dimensions at the end. Basically, the line with if ( dimensions.length < 3 )
will read something like < 4
or < 5
and you would have to wrap that many more for
loops around the one thats there, and they would each all need their own set of var
s --- so I'm not sure how efficient it all might be, as you are trading excessive function call and stack space/manipulation with a similar idea but in embedded for
loops --- But I suppose it could speed up some environments if you know that N is always above a certain level or if its only for the final dimensions. Like here, I did it for the last two dimensions. But if you unroll too much, then your payload itself is a bear. Only testing will tell if thats worth it. It does seem that stack space is limited, and I think I remember having been able to make larger arrays with more unrolling. There is a limit to how big you can make an array. And recursion solutions that call themselves for each item at the Nth level had the lowest limit if I do.. recall.. correctly.... much lower.
The next part in revising his solution is just the logistics, its was just a simple refactor to get rid of excessive blocks and code. Join all the var
work together and thats it. Since you need a arr
to return, once the looping is over, might as well do all your var
work on one line first and luckily, three of the four var
s have the same initialization. Remember, Javascript can optimize code when joining with ,
if possible. This also makes for smaller code as well.
PT
how would you access a specific element in the already-created array?
– Piraeus