It seems notable that while map is mentioned there is no performance measurement shown for map. To me The proof of the pudding would be in map performance. I

It seems notable that while map is mentioned there is no performance measurement shown for map. To me The proof of the pudding would be in map performance. In my spare time I have been building various versions of the nbody benchmark to try out different fundamental data methods, arrays, objects, immutable or not etc. Every alternative version is, of course, slower than the original mutating object version. In principle however, many forms should be optimisable by a 'sufficiently smart JIT' to be at least as fast. It would be nice at least for a JavaScript engine to be able to do in-place maps on arrays defined from literals or single creation points like map/filter etc. , and have only one copy. For instance. { let x = [1,5,9]; let q = x.map(a=>a+1) } The JIT would have to determine that . x held the only reference to the array. . x.map is the last reference to x . the array in x was not modified between creation and last reference. . the used version of a=>a+1 returns the same type as it is passed To place the result of the map into the same memory as the original array and assign that to q. Of course in this ultra simple example an optimizer could just figure out that q is [2,6,10] and never used anyway so throw it away. but in real world code you can frequently have arrays like this that are only ever used as a source for a map, especially if you do q.map(something).map(somethingelse).map(anotherthing).