NOTE
The question has been edited following the good advise from #Kaddath to highlight the fact that the ordering doesn't have to be alphabetical but depending on the position of items inside the arrays.
I have an array of arrays where each of the arrays are based on a given ordering but they can differ a bit.
For example, the base ordering is X -> D -> H -> B and here is my array of arrays:
const arrays = [
['X', 'D', 'H', 'B'],
['X', 'D', 'K', 'Z', 'H', 'B', 'A'],
['X', 'M', 'D', 'H', 'B'],
['X', 'H', 'T'],
['X', 'D', 'H', 'B']
]
I would like to merge all arrays into a single one and remove duplicates but by keeping the ordering. In my example the result would be ['X', 'M', 'D', 'K', 'Z', 'H', 'T', 'B', 'A'].
In the example we can see that M is between X and D inside the third array and it is so placed between X and D in the final output.
I know conflicts may arise but here are the following rules:
Every items should appear in the final output.
If an item is appearing in multiple arrays at different positions, the first appearance is the right one (skip others).
What I've done so far is merging all of these arrays into a single one by using
const merged = [].concat.apply([], arrays);
(cf. https://stackoverflow.com/a/10865042/3520621).
And then getting unique values by using this code snippet from https://stackoverflow.com/a/1584377/3520621 :
Array.prototype.unique = function() {
var a = this.concat();
for(var i=0; i<a.length; ++i) {
for(var j=i+1; j<a.length; ++j) {
if(a[i] === a[j])
a.splice(j--, 1);
}
}
return a;
};
const finalArray = merged.unique();
But my result is this:
[
"X",
"D",
"H",
"B",
"K",
"Z",
"A",
"M",
"T"
]
Any help is welcome!
Thanks.
const arrays = [
['X', 'D', 'H', 'B'],
['X', 'D', 'K', 'Z', 'H', 'B', 'A'],
['X', 'M', 'D', 'H', 'B'],
['X', 'H', 'T'],
['X', 'D', 'H', 'B']
];
const result = [];
arrays.forEach(array => {
array.forEach((item, idx) => {
// check if the item has already been added, if not, try to add
if(!~result.indexOf(item)) {
// if item is not first item, find position of his left sibling in result array
if(idx) {
const result_idx = result.indexOf(array[idx - 1]);
// add item after left sibling position
result.splice(result_idx + 1, 0, item);
return;
}
result.push(item);
}
});
});
console.log('expected result', ['X', 'M', 'D', 'K', 'Z', 'H', 'T', 'B', 'A'].join(','));
console.log(' current result',result.join(','));
Every array is in fact a set of rules that tells what is the relative order between the elements. Final list should return all elements while respecting relative order defined by all rules.
Some solutions have solved the initial request, some even didn't solve that one (all that suggest using sort kind of missed the point of the question). Nevertheless, none proposed a generic solution.
The problem
If we look at the problem asked in the OP, this is how the rules define what is the relative position between elements:
M K -> Z T
^ \ ^ \ ^
/ v/ v/
X -> D ------> H -> B -> A
So, it is easy to see that our array starts with X. Next element can be both D and M. But, D requires M to already be in array. That is why we will put M as our next element, and then D. Next, D points to both K and H. But since H has some other predecessor that are not collected until now, and K has none (actually it has D, but it is already collected in the list), we will put K and Z, and only then H.
H points to both T and B. It actually doesn't matter which one we put first. So, last three elements can be in any of the following three orders:
T, B, A
B, A, T
B, T, A
Let us also take into account a little bit more complex case. Here are the rules:
['10', '11', '12', '1', '2'],
['11', '12', '13', '2'],
['9', '13'],
['9', '10'],
If we draw the graph using those rules we would get following:
--------------> 13 ----
/ ^ \
/ / v
9 -> 10 -> 11 -> 12 > 1 -> 2
What is specific about this case? Two things:
Only in the last rule we "find out" that the number 9 is the beginning of the array
There are two non direct paths from 12 to 2 (one over the number 1, second over the number 13).
Solution
My idea is to create a node from each element. And then use that node to keep track of all immediate successors and immediate predecessors. After that we would find all elements that don't have predecessors and start "collecting" results from there. If we came to the node that has multiple predecessors, but some of them are not collected, we would stop recursion there. It can happen that some of the successors is already collected in some other path. We would skip that successor.
function mergeAndMaintainRelativeOrder(arrays/*: string[][]*/)/*: string[]*/ {
/*
interface NodeElement {
value: string;
predecessor: Set<NodeElement>;
successor: Set<NodeElement>;
collected: boolean;
}
*/
const elements/*: { [key: string]: NodeElement }*/ = {};
// For every element in all rules create NodeElement that will
// be used to keep track of immediate predecessors and successors
arrays.flat().forEach(
(value) =>
(elements[value] = {
value,
predecessor: new Set/*<NodeElement>*/(),
successor: new Set/*<NodeElement>*/(),
// Used when we form final array of results to indicate
// that this node has already be collected in final array
collected: false,
}),
);
arrays.forEach((list) => {
for (let i = 0; i < list.length - 1; i += 1) {
const node = elements[list[i]];
const nextNode = elements[list[i + 1]];
node.successor.add(nextNode);
nextNode.predecessor.add(node);
}
});
function addElementsInArray(head/*: NodeElement*/, array/*: string[]*/) {
let areAllPredecessorsCollected = true;
head.predecessor.forEach((element) => {
if (!element.collected) {
areAllPredecessorsCollected = false;
}
});
if (!areAllPredecessorsCollected) {
return;
}
array.push(head.value);
head.collected = true;
head.successor.forEach((element) => {
if (!element.collected) {
addElementsInArray(element, array);
}
});
}
const results/*: string[]*/ = [];
Object.values(elements)
.filter((element) => element.predecessor.size === 0)
.forEach((head) => {
addElementsInArray(head, results);
});
return results;
}
console.log(mergeAndMaintainRelativeOrder([
['X', 'D', 'H', 'B'],
['X', 'D', 'K', 'Z', 'H', 'B', 'A'],
['X', 'M', 'D', 'H', 'B'],
['X', 'H', 'T'],
['X', 'D', 'H', 'B'],
]));
console.log(mergeAndMaintainRelativeOrder([
['10', '11', '12', '1', '2'],
['11', '12', '13', '2'],
['9', '13'],
['9', '10'],
]));
Big O
If we say that n is the number of the rules, and m is number of elements in each rule, complexity of this algorithm is O(n*m). This takes into account that Set implementation for the JS is near O(1).
Flatten, remove duplicates and sort could be simpler:
const arrays = [
['A', 'B', 'C', 'D'],
['A', 'B', 'B-bis', 'B-ter', 'C', 'D', 'D-bis'],
['A', 'A-bis', 'B', 'C', 'D'],
['A', 'C', 'E'],
['A', 'B', 'C', 'D'],
];
console.log(
arrays
.flat()
.filter((u, i, all) => all.indexOf(u) === i)
.sort((a, b) => a.localeCompare(b)),
);
Or event simpler according to Mohammad Usman's now deleted post:
const arrays = [
['A', 'B', 'C', 'D'],
['A', 'B', 'B-bis', 'B-ter', 'C', 'D', 'D-bis'],
['A', 'A-bis', 'B', 'C', 'D'],
['A', 'C', 'E'],
['A', 'B', 'C', 'D'],
];
console.log(
[...new Set([].concat(...arrays))].sort((a, b) =>
a.localeCompare(b),
),
);
You can use .concat() with Set to get the resultant array of unique values:
const data = [
['A', 'B', 'C', 'D'],
['A', 'B', 'B-bis', 'B-ter', 'C', 'D', 'D-bis'],
['A', 'A-bis', 'B', 'C', 'D'],
['A', 'C', 'E'],
['A', 'B', 'C', 'D']
];
const result = [...new Set([].concat(...data))].sort((a, b) => a.localeCompare(b));
console.log(result);
.as-console-wrapper { max-height: 100% !important; top: 0; }
Create a single array using array#concat and then using Set get the unique values from this array then sort the array.
const arrays = [ ['A', 'B', 'C', 'D'], ['A', 'B', 'B-bis', 'B-ter', 'C', 'D', 'D-bis'], ['A', 'A-bis', 'B', 'C', 'D'], ['A', 'C', 'E'], ['A', 'B', 'C', 'D'] ],
result = [...new Set([].concat(...arrays))].sort();
console.log(result);
merge [].concat.apply([], arrays)
find uniq [...new Set(merged)]
sort .sort()
const arrays = [
['A', 'B', 'C', 'D'],
['A', 'B', 'B-bis', 'B-ter', 'C', 'D', 'D-bis'],
['A', 'A-bis', 'B', 'C', 'D'],
['A', 'C', 'E'],
['A', 'B', 'C', 'D']
];
let merged = [].concat.apply([], arrays); // merge array
let sort = [...new Set(merged)].sort(); // find uniq then sort
console.log(sort);
Fun problem to solve; I think I only partly succeeded.
I ignored the "underspecified" example of B -> A -> T vs T -> B -> A
It's very inefficient
Still posting cause I think it might help you get things right. Here's my approach:
Step 1: create a naive index
We're creating an object that, for each unique element in the nested arrays, tracks which it has succeeded or preceded:
{
"X": { prev: Set({}), next: Set({ "D", "H", "B", "K", "Z", "A", "M", "T" })
"M": { prev: Set({ "X" }), next: Set({ "D", "H", "B" })
// etc.
}
I named it "naive" because these Sets only contain information of one level deep.
I.e.: they only report relations between elements that were in the same array. They cannot see the M comes before the K because they were never in the same array.
Step 2: join the indexes recursively
This is where I ignored all big-O concerns one might have 😉. I merge the index recursively: The next of M is a join of the next of D, H, B. Recurse until you found an element that has no next, i.e. the T or A.
Step 3: create a sorter that respects the sort index:
const indexSorter = idx => (a, b) =>
idx[a].next.has(b) || idx[b].prev.has(a) ? -1 :
idx[a].prev.has(b) || idx[b].next.has(a) ? 1 :
0 ;
This function creates a sort method that uses the generated index to look up the sort order between any two elements.
Bringing it all together:
(function() {
const naiveSortIndex = xss => xss
.map(xs =>
// [ prev, cur, next ]
xs.map((x, i, xs) => [
xs.slice(0, i), x, xs.slice(i + 1)
])
)
// flatten
.reduce((xs, ys) => xs.concat(ys), [])
// add to index
.reduce(
(idx, [prev, cur, next]) => {
if (!idx[cur])
idx[cur] = {
prev: new Set(),
next: new Set()
};
prev.forEach(p => {
idx[cur].prev.add(p);
});
next.forEach(n => {
idx[cur].next.add(n);
});
return idx;
}, {}
);
const expensiveSortIndex = xss => {
const naive = naiveSortIndex(xss);
return Object
.keys(naive)
.reduce(
(idx, k) => Object.assign(idx, {
[k]: {
prev: mergeDir("prev", naive, k),
next: mergeDir("next", naive, k)
}
}), {}
)
}
const mergeDir = (dir, idx, k, s = new Set()) =>
idx[k][dir].size === 0
? s
: Array.from(idx[k][dir])
.reduce(
(s, k2) => mergeDir(dir, idx, k2, s),
new Set([...s, ...idx[k][dir]])
);
// Generate a recursive sort method based on an index of { key: { prev, next } }
const indexSorter = idx => (a, b) =>
idx[a].next.has(b) || idx[b].prev.has(a) ? -1 :
idx[a].prev.has(b) || idx[b].next.has(a) ? 1 :
0;
const uniques = xs => Array.from(new Set(xs));
// App:
const arrays = [
['X', 'D', 'H', 'B'],
['X', 'D', 'K', 'Z', 'H', 'B', 'A'],
['X', 'M', 'D', 'H', 'B'],
['X', 'H', 'T'],
['X', 'D', 'H', 'B']
];
const sortIndex = expensiveSortIndex(arrays);
const sorter = indexSorter(sortIndex);
console.log(JSON.stringify(
uniques(arrays.flat()).sort(sorter)
))
}())
Recommendations
I suppose the elegant solution to the problem might be able to skip all the merging of Sets by using a linked list / tree-like structure and injecting elements at the right indexes by traversing until an element of its prev/next is found.
I would just flatten the arrays, map them as keys to an object (thus removing the doubles), and then sort the final result
const arrays = [
['A', 'B', 'C', 'D'],
['A', 'B', 'B-bis', 'B-ter', 'C', 'D', 'D-bis'],
['A', 'A-bis', 'B', 'C', 'D'],
['A', 'C', 'E'],
['A', 'B', 'C', 'D']
];
const final = Object.keys( arrays.flat().reduce( (aggregate, entry) => {
aggregate[entry] = '';
return aggregate;
}, {} ) ).sort( (x1, x2) => x1.localeCompare(x2) );
console.log( final );
To your code, after the merge you need to remove the duplicates. So you will get the unique array.
Use the array.sort, to sort the array.
I hope this will solve the issue.
const arrays = [
['A', 'B', 'C', 'D'],
['A', 'B', 'B-bis', 'B-ter', 'C', 'D', 'D-bis'],
['A', 'A-bis', 'B', 'C', 'D'],
['A', 'C', 'E'],
['A', 'B', 'C', 'D']
]
const merged = [].concat.apply([], arrays);
const unique = Array.from(new Set(merged))
const sorted = unique.sort()
console.log("sorted Array", sorted)
// Single Line
const result = [...new Set([].concat(...arrays))].sort();
console.log("sorted Array single line", result)
Use a BST for this. Add in all elements to the bst and then traverse in-order.
function BST(){
this.key = null;
this.value = null;
this.left = null;
this.right = null;
this.add = function(key}{
const val = key;
key = someOrderFunction(key.replace(/\s/,''));
if(this.key == null) {
this.key = key;
this.val = val;
} else if(key < this.key) {
if(this.left){
this.left.add(val);
} else {
this.left = new BST();
this.left.key = key;
this.left.val = val;
}
} else if(key > this.key) {
if(this.right){
this.right.add(val);
} else {
this.right= new BST();
this.right.key = key;
this.right.val = val;
}
}
this.inOrder = function(){
const leftNodeOrder = this.left ? this.left.inOrder() : [],
rightNodeOrder = this.right? this.right.inOrder() : [];
return leftNodeOrder.concat(this.val).concat(this.rightNodeOrder);
}
}
// MergeArrays uses a BST to insert all elements of all arrays
// and then fetches them sorted in order
function mergeArrays(arrays) {
const bst = new BST();
arrays.forEach(array => array.forEach( e => bst.add(e)));
return bst.inOrder();
}
My solution focuses nothing on efficiency, so I wouldn't try this for large arrays. But it works fine for me.
The idea is to walk through all elements multiple times and only insert an element into the sorted array in one of three cases:
The current element is first in its array, and one of its successors is first in the sorted array.
The current element is last in its array, and one of its predecessors is last in the sorted array.
The preceding element is in the sorted array and one of the current elements successors are directly succeeding this preceding element in the sorted array.
For the current problem, as stated above, the order between T and B, A, isn't uniquely determined. To handle this I use a flag force which takes any legal option when no new inserts could be made during an iteration.
The following rule from the problem is not implemented in my solution. If an item is appearing in multiple arrays at different positions, the first appearance is the right one (skip others). There is no hierarchy between the arrays. It should however be easy to implement the desired check and continue if it's not satisfied.
let merge = (arrays) => {
let sorted = [...arrays[0]];
const unused_rules = arrays.slice(1);
let not_inserted = unused_rules.flat().filter((v) => !sorted.includes(v));
let last_length = -1;
let force = false;
// avoids lint warning
const sortedIndex = (sorted) => (v) => sorted.indexOf(v);
// loop until all elements are inserted, or until not even force works
while (not_inserted.length !== 0 && !force) {
force = not_inserted.length === last_length; //if last iteration didn't add elements, our arrays lack complete information and we must add something using what little we know
last_length = not_inserted.length;
for (let j = 0; j < unused_rules.length; j += 1) {
const array = unused_rules[j];
for (let i = 0; i < array.length; i += 1) {
// check if element is already inserted
if (sorted.indexOf(array[i]) === -1) {
if (i === 0) {
// if element is first in its array, check if it can be prepended to sorted array
const index = array.indexOf(sorted[0]);
if (index !== -1 || force) {
const insert = array.slice(0, force ? 1 : index);
sorted = [...insert, ...sorted];
not_inserted = not_inserted.filter((v) => !insert.includes(v));
force = false;
}
} else if (i === array.length - 1) {
// if element is last in its array, check if it can be appended to sorted array
const index = array.indexOf(sorted[sorted.length - 1]);
if (index !== -1 || force) {
const insert = array.slice(force ? array.length - 1 : index + 1);
sorted = [...sorted, ...insert];
not_inserted = not_inserted.filter((v) => !insert.includes(v));
force = false;
}
} else {
const indices = array.map(sortedIndex(sorted)); // map all elements to its index in sorted
const predecessorIndexSorted = indices[i - 1]; // index in the sorted array of the element preceding current element
let successorIndexArray;
if (force) {
successorIndexArray = i + 1;
} else {
successorIndexArray = indices.indexOf(predecessorIndexSorted + 1); // index in the current array of the element succeeding the current elements predecessor in the sorted array
}
if (predecessorIndexSorted !== -1 && successorIndexArray !== -1) {
// insert all elements between predecessor and successor
const insert = array.slice(i, successorIndexArray);
sorted.splice(i, 0, ...insert);
not_inserted = not_inserted.filter((v) => !insert.includes(v));
force = false;
}
}
}
}
}
}
return sorted;
};
In fact, the rule If an item is appearing in multiple arrays at different positions, the first appearance is the right one (skip others). is a bit vague. For example using the arrays below, is it okay to end up with arrays[3] as the sorted array, since it doesn't violate the first appearance of any element, or should arrays[2] take precedence?
const arrays = [['a', 'b', 'd'],
['a', 'c', 'd'],
['a', 'b', 'c', 'd']
['a', 'c', 'b', 'd']]
Is there a simple way in javascript to take a flat array and convert into an object with the even-indexed members of the array as properties and odd-indexed members as corresponding values (analgous to ruby's Hash[*array])?
For example, if I have this:
[ 'a', 'b', 'c', 'd', 'e', 'f' ]
Then I want this:
{ 'a': 'b', 'c': 'd', 'e': 'f' }
The best I've come up with so far seems more verbose than it has to be:
var arr = [ 'a', 'b', 'c', 'd', 'e', 'f' ];
var obj = {};
for (var i = 0, len = arr.length; i < len; i += 2) {
obj[arr[i]] = arr[i + 1];
}
// obj => { 'a': 'b', 'c': 'd', 'e': 'f' }
Is there a better, less verbose, or more elegant way to do this? (Or I have just been programming in ruby too much lately?)
I'm looking for an answer in vanilla javascript, but would also be interested if there is a better way to do this if using undercore.js or jQuery. Performance is not really a concern.
Pretty sure this will work and is shorter:
var arr = [ 'a', 'b', 'c', 'd', 'e', 'f' ];
var obj = {};
while (arr.length) {
obj[arr.shift()] = arr.shift();
}
See shift().
var arr = [ 'a', 'b', 'c', 'd', 'e', 'f' ];
var obj = arr.reduce( function( ret, value, i, values ) {
if( i % 2 === 0 ) ret[ value ] = values[ i + 1 ];
return ret;
}, { } );
If you need it multiple times you can also add a method to the Array.prototype:
Array.prototype.to_object = function () {
var obj = {};
for(var i = 0; i < this.length; i += 2) {
obj[this[i]] = this[i + 1];
}
return obj
};
var a = [ 'a', 'b', 'c', 'd', 'e', 'f' ];
a.to_object(); // => { 'a': 'b', 'c': 'd', 'e': 'f' }
You could first chunk your array into groups of two:
[['a', 'b'], ['c', 'd'], ['e', 'f']]
so that is is in a valid format to be used by Object.fromEntries(), which will build your object for you:
const chunk = (arr, size) => arr.length ? [arr.slice(0, size), ...chunk(arr.slice(size), size)] : [];
const arr = ['a', 'b', 'c', 'd', 'e', 'f'];
const res = Object.fromEntries(chunk(arr, 2));
console.log(res); // {a: "b", c: "d", e: "f"}
With underscore.js and lodash, you don't need to implement the chunk() method yourself, and can instead use _.chunk(), a method built into both libraries. The full lodash equivalent of the above would be:
// lodash
> _.fromPairs(_.chunk(arr, 2));
> {a: "b", c: "d", e: "f"}
Using _.fromPairs provides better browser support, so if using lodash, it is preferred over Object.fromEntries()
Similarly, we can use _.object() if you're using underscore.js to build the object:
// underscore.js
> _.object(_.chunk(arr, 2));
> {a: "b", c: "d", e: "f"}