I'm trying to convert an array of arrays into an array of nested objects in JavaScript. Let's assume each subarray in the array represents a file path. I want to create an array of objects where each object has 2 properties, the name of the current file and any files/children that come after the current file/parent.
So for example, if I have this array of arrays where each subarray represents a file path:
[['A', 'B', 'C'], ['A', 'B', 'D'], ['L', 'M', 'N']]
I want to get this as the result:
[
{
name :'A',
children: [
{
name: 'B',
children: [
{
name: 'C',
children: []
},
{
name: 'D',
children: []
}
]
}
]
},
{
name: 'L',
children: [
{
name: 'M',
children: [
{
name: 'N',
children: []
}
]
}
]
}
]
I tried mapping through the array of arrays and creating an object for the current file/parent if it hasn't been created yet. I think I may be on the right track but I can't seem to think of the best way to do so.
Something I could do in 5 minutes, it probably can be improved. This could also be written as a recursive function, I believe.
const result = [];
arr.forEach((subArr) => {
var ref = result;
subArr.forEach((name) => {
const obj = ref.find((obj) => obj.name == name);
if (obj) {
ref = obj.children;
} else {
ref.push({ name, children: [] });
ref = ref[ref.length - 1].children;
}
});
});
Here's mine:
// Function to convert path array ‘a’ from position ‘i’ into tree structure in ‘v’.
const tree = (a, i, v) => {
if (i < a.length) { tree(a, i+1, v[a[i]] ||= { }) }
}
// Function to convert a simple node into the desired record format.
const record = (v) => {
const a = [];
for (const [k, w] of Object.entries(v)) {
a.push({ name: k, children: record(w) });
}
return a;
}
const m = { }
for (const a of [['A', 'B', 'C'], ['A', 'B', 'D'], ['L', 'M', 'N']]) {
tree(a, 0, m);
}
const result = record(m);
While it might be overkill for this requirement, I have handy a variant of one of my utility functions, setPath, that is used for adding such a path to an existing array. This is a bit different from the other answers in that it does this in an immutable way, returning a new object that shares as much structure as possible with the original one. I always prefer to work with immutable data.
Using that, we can write a hydrate function to do this job as the one-liner, const hydrate = (paths) => paths .reduce (setPath, []).
This is quite likely overkill here, as there is probably no reason to build your output one immutable level after another. But it is a demonstration of the value of keeping utility functions handy.
The code looks like this:
const call = (fn, ...args) => fn (...args)
const setPath = (xs, [name, ...names]) => call (
(i = ((xs .findIndex (x => x .name == name) + 1) || xs .length + 1) - 1) =>
name == undefined
? [...xs]
: [
...xs .slice (0, i),
{name, children: setPath ((i == xs .length) ? [] : xs [i] .children, names)},
...xs .slice (i + 1)
]
)
const hydrate = (paths) => paths .reduce (setPath, [])
console .log (
hydrate ([['A', 'B', 'C'], ['A', 'B', 'D'], ['L', 'M', 'N']])
)
.as-console-wrapper {max-height: 100% !important; top: 0}
We have a trivial call helper function that I use here to avoid statements. As much as possible, I prefer to work with expressions, as these avoid temporal notions in code brought on by statements, and they compose much better into larger pieces. We could alternatively do this with default parameters, but they have other problems. A trivial call function handles this nicely.
In our main function, we destructure apart the first name in the input from the remaining ones. Then we pass to call a function that calculates the index of the element with our existing name in the input array. If it doesn't exist, the index will be the length of that array. This is perhaps over-tricky. findIndex returns -1 if no element matches. We add 1 to the result, and then, if it's 0, we choose one more than the length of the array. Finally we subtract 1 from the result, and now the index will be where we found our target or the length of the array if it wasn't found.
Now, if the path is empty, we return a copy of our array. (I prefer the copy just for consistency, but it would be legitimate to just return it directly.) If it's not empty, we use the index to tear apart our input array, keeping everything before it, building a new item for that index by recursively calling setPath with the remaining node names, and then keeping everything after that index.
And now, as noted, our hydrate function is a trivial fold of setPath starting with an empty array.
Say I have an object:
elmo = {
color: 'red',
annoying: true,
height: 'unknown',
meta: { one: '1', two: '2'}
};
I want to make a new object with a subset of its properties.
// pseudo code
subset = elmo.slice('color', 'height')
//=> { color: 'red', height: 'unknown' }
How may I achieve this?
Using Object Destructuring and Property Shorthand
const object = { a: 5, b: 6, c: 7 };
const picked = (({ a, c }) => ({ a, c }))(object);
console.log(picked); // { a: 5, c: 7 }
From Philipp Kewisch:
This is really just an anonymous function being called instantly. All of this can be found on the Destructuring Assignment page on MDN. Here is an expanded form
let unwrap = ({a, c}) => ({a, c});
let unwrap2 = function({a, c}) { return { a, c }; };
let picked = unwrap({ a: 5, b: 6, c: 7 });
let picked2 = unwrap2({a: 5, b: 6, c: 7})
console.log(picked)
console.log(picked2)
Two common approaches are destructuring and conventional Lodash-like pick/omit implementation. The major practical difference between them is that destructuring requires a list of keys to be static, can't omit them, includes non-existent picked keys, i.e. it's inclusive. This may or not be desirable and cannot be changed for destructuring syntax.
Given:
var obj = { 'foo-bar': 1, bar: 2, qux: 3 };
The expected result for regular picking of foo-bar, bar, baz keys:
{ 'foo-bar': 1, bar: 2 }
The expected result for inclusive picking:
{ 'foo-bar': 1, bar: 2, baz: undefined }
Destructuring
Destructuring syntax allows to destructure and recombine an object, with either function parameters or variables.
The limitation is that a list of keys is predefined, they cannot be listed as strings, as described in the question. Destructuring becomes more complicated if a key is non-alphanumeric, e.g. foo-bar.
The upside is that it's performant solution that is natural to ES6.
The downside is that a list of keys is duplicated, this results in verbose code in case a list is long. Since destructuring duplicates object literal syntax in this case, a list can be copied and pasted as is.
IIFE
const subset = (({ 'foo-bar': foo, bar, baz }) => ({ 'foo-bar': foo, bar, baz }))(obj);
Temporary variables
const { 'foo-bar': foo, bar, baz } = obj;
const subset = { 'foo-bar': foo, bar, baz };
A list of strings
Arbitrary list of picked keys consists of strings, as the question requires. This allows to not predefine them and use variables that contain key names, ['foo-bar', someKey, ...moreKeys].
ECMAScript 2017 has Object.entries and Array.prototype.includes, ECMAScript 2019 has Object.fromEntries, they can be polyfilled when needed.
One-liners
Considering that an object to pick contains extra keys, it's generally more efficient to iterate over keys from a list rather than object keys, and vice versa if keys need to be omitted.
Pick (ES5)
var subset = ['foo-bar', 'bar', 'baz']
.reduce(function (obj2, key) {
if (key in obj) // line can be removed to make it inclusive
obj2[key] = obj[key];
return obj2;
}, {});
Omit (ES5)
var subset = Object.keys(obj)
.filter(function (key) {
return ['baz', 'qux'].indexOf(key) < 0;
})
.reduce(function (obj2, key) {
obj2[key] = obj[key];
return obj2;
}, {});
Pick (ES6)
const subset = ['foo-bar', 'bar', 'baz']
.filter(key => key in obj) // line can be removed to make it inclusive
.reduce((obj2, key) => (obj2[key] = obj[key], obj2), {});
Omit (ES6)
const subset = Object.keys(obj)
.filter(key => ['baz', 'qux'].indexOf(key) < 0)
.reduce((obj2, key) => (obj2[key] = obj[key], obj2), {});
Pick (ES2019)
const subset = Object.fromEntries(
['foo-bar', 'bar', 'baz']
.filter(key => key in obj) // line can be removed to make it inclusive
.map(key => [key, obj[key]])
);
Omit (ES2019)
const subset = Object.fromEntries(
Object.entries(obj)
.filter(([key]) => !['baz', 'qux'].includes(key))
);
Reusable functions
One-liners can be represented as reusable helper functions similar to Lodash pick or omit, where a list of keys is passed through arguments, pick(obj, 'foo-bar', 'bar', 'baz').
JavaScript
const pick = (obj, ...keys) => Object.fromEntries(
keys
.filter(key => key in obj)
.map(key => [key, obj[key]])
);
const inclusivePick = (obj, ...keys) => Object.fromEntries(
keys.map(key => [key, obj[key]])
);
const omit = (obj, ...keys) => Object.fromEntries(
Object.entries(obj)
.filter(([key]) => !keys.includes(key))
);
TypeScript
Credit goes to #Claude.
const pick = <T extends {}, K extends keyof T>(obj: T, ...keys: K[]) => (
Object.fromEntries(
keys
.filter(key => key in obj)
.map(key => [key, obj[key]])
) as Pick<T, K>
);
const inclusivePick = <T extends {}, K extends (string | number | symbol)>(
obj: T, ...keys: K[]
) => (
Object.fromEntries(
keys
.map(key => [key, obj[key as unknown as keyof T]])
) as {[key in K]: key extends keyof T ? T[key] : undefined}
)
const omit = <T extends {}, K extends keyof T>(
obj: T, ...keys: K[]
) =>(
Object.fromEntries(
Object.entries(obj)
.filter(([key]) => !keys.includes(key as K))
) as Omit<T, K>
)
I suggest taking a look at Lodash; it has a lot of great utility functions.
For example pick() would be exactly what you seek:
var subset = _.pick(elmo, ['color', 'height']);
fiddle
If you are using ES6 there is a very concise way to do this using destructuring. Destructuring allows you to easily add on to objects using a spread, but it also allows you to make subset objects in the same way.
const object = {
a: 'a',
b: 'b',
c: 'c',
d: 'd',
}
// Remove "c" and "d" fields from original object:
const {c, d, ...partialObject} = object;
const subset = {c, d};
console.log(partialObject) // => { a: 'a', b: 'b'}
console.log(subset) // => { c: 'c', d: 'd'};
While it's a bit more verbose, you can accomplish what everyone else was recommending underscore/lodash for 2 years ago, by using Array.prototype.reduce.
var subset = ['color', 'height'].reduce(function(o, k) { o[k] = elmo[k]; return o; }, {});
This approach solves it from the other side: rather than take an object and pass property names to it to extract, take an array of property names and reduce them into a new object.
While it's more verbose in the simplest case, a callback here is pretty handy, since you can easily meet some common requirements, e.g. change the 'color' property to 'colour' on the new object, flatten arrays, etc. -- any of the things you need to do when receiving an object from one service/library and building a new object needed somewhere else. While underscore/lodash are excellent, well-implemented libs, this is my preferred approach for less vendor-reliance, and a simpler, more consistent approach when my subset-building logic gets more complex.
edit: es7 version of the same:
const subset = ['color', 'height'].reduce((a, e) => (a[e] = elmo[e], a), {});
edit: A nice example for currying, too! Have a 'pick' function return another function.
const pick = (...props) => o => props.reduce((a, e) => ({ ...a, [e]: o[e] }), {});
The above is pretty close to the other method, except it lets you build a 'picker' on the fly. e.g.
pick('color', 'height')(elmo);
What's especially neat about this approach, is you can easily pass in the chosen 'picks' into anything that takes a function, e.g. Array#map:
[elmo, grover, bigBird].map(pick('color', 'height'));
// [
// { color: 'red', height: 'short' },
// { color: 'blue', height: 'medium' },
// { color: 'yellow', height: 'tall' },
// ]
I am adding this answer because none of the answer used Comma operator.
It's very easy with destructuring assignment and , operator
const object = { a: 5, b: 6, c: 7 };
const picked = ({a,c} = object, {a,c})
console.log(picked);
One more solution:
var subset = {
color: elmo.color,
height: elmo.height
}
This looks far more readable to me than pretty much any answer so far, but maybe that's just me!
There is nothing like that built-in to the core library, but you can use object destructuring to do it...
const {color, height} = sourceObject;
const newObject = {color, height};
You could also write a utility function do it...
const cloneAndPluck = function(sourceObject, keys) {
const newObject = {};
keys.forEach((obj, key) => { newObject[key] = sourceObject[key]; });
return newObject;
};
const subset = cloneAndPluck(elmo, ["color", "height"]);
Libraries such as Lodash also have _.pick().
TypeScript solution:
function pick<T extends object, U extends keyof T>(
obj: T,
paths: Array<U>
): Pick<T, U> {
const ret = Object.create(null);
for (const k of paths) {
ret[k] = obj[k];
}
return ret;
}
The typing information even allows for auto-completion:
Credit to DefinitelyTyped for U extends keyof T trick!
TypeScript Playground
I want to mention that very good curation here:
pick-es2019.js
Object.fromEntries(
Object.entries(obj)
.filter(([key]) => ['whitelisted', 'keys'].includes(key))
);
pick-es2017.js
Object.entries(obj)
.filter(([key]) => ['whitelisted', 'keys'].includes(key))
.reduce((obj, [key, val]) => Object.assign(obj, { [key]: val }), {});
pick-es2015.js
Object.keys(obj)
.filter((key) => ['whitelisted', 'keys'].indexOf(key) >= 0)
.reduce((newObj, key) => Object.assign(newObj, { [key]: obj[key] }), {})
omit-es2019.js
Object.fromEntries(
Object.entries(obj)
.filter(([key]) => !['blacklisted', 'keys'].includes(key))
);
omit-es2017.js
Object.entries(obj)
.filter(([key]) => !['blacklisted', 'keys'].includes(key))
.reduce((obj, [key, val]) => Object.assign(obj, { [key]: val }), {});
omit-es2015.js
Object.keys(obj)
.filter((key) => ['blacklisted', 'keys'].indexOf(key) < 0)
.reduce((newObj, key) => Object.assign(newObj, { [key]: obj[key] }), {})
You can use Lodash also.
var subset = _.pick(elmo ,'color', 'height');
Complementing, let's say you have an array of "elmo"s :
elmos = [{
color: 'red',
annoying: true,
height: 'unknown',
meta: { one: '1', two: '2'}
},{
color: 'blue',
annoying: true,
height: 'known',
meta: { one: '1', two: '2'}
},{
color: 'yellow',
annoying: false,
height: 'unknown',
meta: { one: '1', two: '2'}
}
];
If you want the same behavior, using lodash, you would just:
var subsets = _.map(elmos, function(elm) { return _.pick(elm, 'color', 'height'); });
Destructuring into dynamically named variables is impossible in JavaScript as discussed in this question.
To set keys dynamically, you can use reduce function without mutating object as follows:
const getSubset = (obj, ...keys) => keys.reduce((a, c) => ({ ...a, [c]: obj[c] }), {});
const elmo = {
color: 'red',
annoying: true,
height: 'unknown',
meta: { one: '1', two: '2'}
}
const subset = getSubset(elmo, 'color', 'annoying')
console.log(subset)
Should note that you're creating a new object on every iteration though instead of updating a single clone. – mpen
below is a version using reduce with single clone (updating initial value passed in to reduce).
const getSubset = (obj, ...keys) => keys.reduce((acc, curr) => {
acc[curr] = obj[curr]
return acc
}, {})
const elmo = {
color: 'red',
annoying: true,
height: 'unknown',
meta: { one: '1', two: '2'}
}
const subset = getSubset(elmo, 'annoying', 'height', 'meta')
console.log(subset)
Dynamic solution
['color', 'height'].reduce((a,b) => (a[b]=elmo[b],a), {})
let subset= (obj,keys)=> keys.reduce((a,b)=> (a[b]=obj[b],a),{});
// TEST
let elmo = {
color: 'red',
annoying: true,
height: 'unknown',
meta: { one: '1', two: '2'}
};
console.log( subset(elmo, ['color', 'height']) );
Use pick method of lodash library if you are already using.
var obj = { 'a': 1, 'b': '2', 'c': 3 };
_.pick(object, ['a', 'c']);
// => { 'a': 1, 'c': 3 }
https://lodash.com/docs/4.17.10#pick
The easiest way I found, which doesn't create unnecessary variables, is a function you can call and works identically to lodash is the following:
pick(obj, keys){
return Object.assign({}, ...keys.map(key => ({ [key]: obj[key] })))
}
For example:
pick(obj, keys){
return Object.assign({}, ...keys.map(key => ({ [key]: obj[key] })))
}
const obj = {a:1, b:2, c:3, d:4}
const keys = ['a', 'c', 'f']
const picked = pick(obj,keys)
console.log(picked)
pick = (obj, keys) => {
return Object.assign({}, ...keys.map(key => ({
[key]: obj[key]
})))
}
const obj = {
a: 1,
b: 2,
c: 3,
d: 4
}
const keys = ['a', 'c', 'f']
const picked = pick(obj, keys)
console.log(picked)
An Array of Objects
const aListOfObjects = [{
prop1: 50,
prop2: "Nothing",
prop3: "hello",
prop4: "What's up",
},
{
prop1: 88,
prop2: "Whatever",
prop3: "world",
prop4: "You get it",
},
]
Making a subset of an object or objects can be achieved by destructuring the object this way.
const sections = aListOfObjects.map(({prop1, prop2}) => ({prop1, prop2}));
Using the "with" statement with shorthand object literal syntax
Nobody has demonstrated this method yet, probably because it's terrible and you shouldn't do it, but I feel like it has to be listed.
var o = {a:1,b:2,c:3,d:4,e:4,f:5}
with(o){
var output = {a,b,f}
}
console.log(output)
Pro: You don't have to type the property names twice.
Cons: The "with" statement is not recommended for many reasons.
Conclusion: It works great, but don't use it.
Just another way...
var elmo = {
color: 'red',
annoying: true,
height: 'unknown',
meta: { one: '1', two: '2'}
}
var subset = [elmo].map(x => ({
color: x.color,
height: x.height
}))[0]
You can use this function with an array of Objects =)
If you want to keep more properties than the ones you want to remove, you could use the rest parameter syntax:
const obj = {
a:1,
b:2,
c:3,
d:4
};
const { a, ...newObj } = obj;
console.log(newObj); // {b: 2, c: 3, d: 4}
To add another esoteric way, this works aswell:
var obj = {a: 1, b:2, c:3}
var newobj = {a,c}=obj && {a,c}
// {a: 1, c:3}
but you have to write the prop names twice.
How about:
function sliceObj(obj) {
var o = {}
, keys = [].slice.call(arguments, 1);
for (var i=0; i<keys.length; i++) {
if (keys[i] in obj) o[keys[i]] = obj[keys[i]];
}
return o;
}
var subset = sliceObj(elmo, 'color', 'height');
This works for me in Chrome console. Any problem with this?
var { color, height } = elmo
var subelmo = { color, height }
console.log(subelmo) // {color: "red", height: "unknown"}
convert arguments to array
use Array.forEach() to pick the property
Object.prototype.pick = function(...args) {
var obj = {};
args.forEach(k => obj[k] = this[k])
return obj
}
var a = {0:"a",1:"b",2:"c"}
var b = a.pick('1','2') //output will be {1: "b", 2: "c"}
Like several on this thread I agree with evert that the most obvious old school way of doing this is actually the best available, however for fun let me provide one other inadvisable way of doing it in certain circumstances, say when you already have your subset defined and you want to copy properties to it from another object that contains a superset or intersecting set of its properties.
let set = { a : 1, b : 2, c : 3 };
let subset = { a : null, b : null };
try {
Object.assign(Object.seal(subset), set);
} catch (e) {
console.log('its ok I meant to do that <(^.^)^');
}
console.log(subset);
I think this is your answer. (and everyone who is looking for it).
const object = { a: 5, b: 6, c: 7 };
const subset = (({ a, c }) => ({ a, c }))(object);
console.log(subset); // { a: 5, c: 7 }
Good-old Array.prototype.reduce:
const selectable = {a: null, b: null};
const v = {a: true, b: 'yes', c: 4};
const r = Object.keys(selectable).reduce((a, b) => {
return (a[b] = v[b]), a;
}, {});
console.log(r);
this answer uses the magical comma-operator, also:
https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Operators/Comma_Operator
if you want to get really fancy, this is more compact:
const r = Object.keys(selectable).reduce((a, b) => (a[b] = v[b], a), {});
Putting it all together into a reusable function:
const getSelectable = function (selectable, original) {
return Object.keys(selectable).reduce((a, b) => (a[b] = original[b], a), {})
};
const r = getSelectable(selectable, v);
console.log(r);
I've got the same problem and solved it easily by using the following libs:
object.pick
https://www.npmjs.com/package/object.pick
pick({a: 'a', b: 'b', c: 'c'}, ['a', 'b'])
//=> {a: 'a', b: 'b'}
object.omit
https://www.npmjs.com/package/object.omit
omit({a: 'a', b: 'b', c: 'c'}, ['a', 'c'])
//=> { b: 'b' }
I know it isn't the cleanest, but it's simple and easy to understand.
function obj_multi_select(obj, keys){
let return_obj = {};
for (let k = 0; k < keys.length; k++){
return_obj[keys[k]] = obj[keys[k]];
};
return return_obj;
};
function splice()
{
var ret = new Object();
for(i = 1; i < arguments.length; i++)
ret[arguments[i]] = arguments[0][arguments[i]];
return ret;
}
var answer = splice(elmo, "color", "height");
Destructuring assignment with dynamic properties
This solution not only applies to your specific example but is more generally applicable:
const subset2 = (x, y) => ({[x]:a, [y]:b}) => ({[x]:a, [y]:b});
const subset3 = (x, y, z) => ({[x]:a, [y]:b, [z]:c}) => ({[x]:a, [y]:b, [z]:c});
// const subset4...etc.
const o = {a:1, b:2, c:3, d:4, e:5};
const pickBD = subset2("b", "d");
const pickACE = subset3("a", "c", "e");
console.log(
pickBD(o), // {b:2, d:4}
pickACE(o) // {a:1, c:3, e:5}
);
You can easily define subset4 etc. to take more properties into account.