Algorithm: efficient way to remove duplicate integers from an array

I got this problem from an interview with Microsoft.

Given an array of random integers, write an algorithm in C that removes duplicated numbers and return the unique numbers in the original array.

E.g Input: {4, 8, 4, 1, 1, 2, 9} Output: {4, 8, 1, 2, 9, ?, ?}

One caveat is that the expected algorithm should not required the array to be sorted first. And when an element has been removed, the following elements must be shifted forward as well. Anyway, value of elements at the tail of the array where elements were shifted forward are negligible.

Update: The result must be returned in the original array and helper data structure (e.g. hashtable) should not be used. However, I guess order preservation is not necessary.

Update2: For those who wonder why these impractical constraints, this was an interview question and all these constraints are discussed during the thinking process to see how I can come up with different ideas.

How to efficiently eliminate duplicate elements in double array (in C)?

Although I have read through the interesting thread Algorithm: efficient way to remove duplicate integers from an array , I have not been able to find a satisfying answer to my question: I have a one-

Find running median from a stream of integers [duplicate]

Possible Duplicate: Rolling median algorithm in C Given that integers are read from a data stream. Find median of elements read so for in efficient way. Solution I have read: We can use a max hea

Most efficient way to remove duplicate elements from a list

I have an unsorted list, from which I want to remove duplicates. What is the most efficient to perform this in C. In my case, the list is a linked list of arrays. In other words several arrays linked

What is the most efficient way to remove all the elements of one array from another array?

I have two arrays, array A contains a long list with some elements I want to remove. Array B is the full list of those elements that I wish to remove from array A. What is the most efficient way to ac

Algorithm: efficient way to search an integer in a two dimensional integer array? [duplicate]

Possible Duplicates: Given a 2d array sorted in increasing order from left to right and top to bottom, what is the best way to search for a target number? Search a sorted 2D matrix A time efficient

Is there a more efficient way to sort integers using pointers WITHOUT using an array? [duplicate]

This question already has an answer here: Simpler way of sorting three numbers 5 answers Is there a more efficient way to do what I have done here? Maybe without having to use so many if statem

Most efficient way to remove an item from array?

I have two arrays. If a user adds a product, we put it in the ProductArray. If they remove that product, we add it to the ProductArrayRemove array as well as remove it from the product array. (We need

Fastest way to get a sorted array out of an Iterable of integers [duplicate]

Possible Duplicate: Fastest strategy to form and sort an array of positive integers What would be the fastest way to get a sorted array from an unsorted iterable of integers ? Currently I do it by i

efficient way of removing None’s from numpy array

Is there an efficient way to remove Nones from numpy arrays and resize the array to its new size? For example, how would you remove the None from this frame without iterating through it in python. I c

Remove duplicate from KOObservable array

Want to remove duplicate values from KoObservableArray. I have following JS file. I want to bind the name column in the UI as dropdown as well as to remove the duplicate values. Suggest the best way t


If you are looking for the superior O-notation, then sorting the array with an O(n log n) sort then doing a O(n) traversal may be the best route. Without sorting, you are looking at O(n^2).

Edit: if you are just doing integers, then you can also do radix sort to get O(n).

An array should obviously be “traversed” right-to-left to avoid unneccessary copying of values back and forth.

If you have unlimited memory, you can allocate a bit array for sizeof(type-of-element-in-array) / 8 bytes to have each bit signify whether you’ve already encountered corresponding value or not.

If you don’t, I can’t think of anything better than traversing an array and comparing each value with values that follow it and then if duplicate is found, remove these values altogether. This is somewhere near O(n^2) (or O((n^2-n)/2)).

IBM has an article on kinda close subject.

Well, it’s basic implementation is quite simple. Go through all elements, check whether there are duplicates in the remaining ones and shift the rest over them.

It’s terrible inefficient and you could speed it up by a helper-array for the output or sorting/binary trees, but this doesn’t seem to be allowed.

You could do this in a single traversal, if you are willing to sacrifice memory. You can simply tally whether you have seen an integer or not in a hash/associative array. If you have already seen a number, remove it as you go, or better yet, move numbers you have not seen into a new array, avoiding any shifting in the original array.

In Perl:

foreach $i (@myary) {
    if(!defined $seen{$i}) {
        $seen{$i} = 1;
        push @newary, $i;

Let’s see:

  • O(N) pass to find min/max allocate
  • bit-array for found
  • O(N) pass swapping duplicates to end.

If you are allowed to use C++, a call to std::sort followed by a call to std::unique will give you the answer. The time complexity is O(N log N) for the sort and O(N) for the unique traversal.

And if C++ is off the table there isn’t anything that keeps these same algorithms from being written in C.

It’d be cool if you had a good DataStructure that could quickly tell if it contains an integer. Perhaps a tree of some sort.

DataStructure elementsSeen = new DataStructure();
int elementsRemoved = 0;
for(int i=0;i<array.Length;i++){
    array[i-elementsRemoved] = array[i];
array.Length = array.Length - elementsRemoved;

This can be done in a single pass, in O(N) time in the number of integers in the input list, and O(N) storage in the number of unique integers.

Walk through the list from front to back, with two pointers “dst” and “src” initialized to the first item. Start with an empty hash table of “integers seen”. If the integer at src is not present in the hash, write it to the slot at dst and increment dst. Add the integer at src to the hash, then increment src. Repeat until src passes the end of the input list.

You can sort the array using radix sort which you can do in O(n) and then simple O(n) traversal. So the final time complexity is O(n).

Update: Size of int in C is limited, therefore the length of value representation is constant.

How about:

void rmdup(int *array, int length)
    int *current , *end = array + length - 1;

    for ( current = array + 1; array < end; array++, current = array + 1 )
        while ( current <= end )
            if ( *current == *array )
                *current = *end--;

Should be O(n^2) or less.

A solution suggested by my girlfriend is a variation of merge sort. The only modification is that during the merge step, just disregard duplicated values. This solution would be as well O(n log n). In this approach, the sorting/duplication removal are combined together. However, I’m not sure if that makes any difference, though.

One more efficient implementation

int i, j;

/* new length of modified array */
int NewLength = 1;

for(i=1; i< Length; i++){

   for(j=0; j< NewLength ; j++)

      if(array[i] == array[j])

   /* if none of the values in index[0..j] of array is not same as array[i],
      then copy the current value to corresponding new position in array */

  if (j==NewLength )
      array[NewLength++] = array[i];

In this implementation there is no need for sorting the array. Also if a duplicate element is found, there is no need for shifting all elements after this by one position.

The output of this code is array[] with size NewLength

Here we are starting from the 2nd elemt in array and comparing it with all the elements in array up to this array. We are holding an extra index variable ‘NewLength’ for modifying the input array. NewLength variabel is initialized to 0.

Element in array[1] will be compared with array[0]. If they are different, then value in array[NewLength] will be modified with array[1] and increment NewLength. If they are same, NewLength will not be modified.

So if we have an array [1 2 1 3 1], then

In First pass of ‘j’ loop, array[1] (2) will be compared with array0, then 2 will be written to array[NewLength] = array[1] so array will be [1 2] since NewLength = 2

In second pass of ‘j’ loop, array[2] (1) will be compared with array0 and array1. Here since array[2] (1) and array0 are same loop will break here. so array will be [1 2] since NewLength = 2

and so on

Some of the answers that are written here are pretty trivial (O(n^2) or sorting and traversing in O(NlogN)) and I’m assuming that is not what was expected in an interview for Microsoft. Obviously any answer above O(n) wasn’t what they were looking for. The update states that there shouldn’t be any helper data structures so any answer that has one (a hash table, tree, bit array or whatever) shouldn’t be a valid solution.

If you can allocate additional memory then Jeff B’s answer is probably easiest way to do it. I have a good answer for questions like these but the MAXINT needs to be bounded by the size of the array. (Example: An array of size 100 may contain any number between 1 and 100. Remove the dups as the original question)

The answer to this in O(n) time and O(1) memory is:

int maxNumInArray = findMaxNumInArray(arr);
int dup = findMinNumInArray(arr) - 1;
for (int i=0; i < arrLength; ++i) {
    int seekIndex = arr[i] % (maxNumInArray+1);
    if (arr[seekIndex] > maxNumInArray)
        arr[i] = dup; // invalidate index
        arr[seekIndex] = arr[seekIndex] + maxNumInArray;

int i = 0;
int j = arrLength(arr)-1;
while (i<j) {
    while (arr[i] != dup)
    while (arr[j] == dup)
    swap(arr[i], arr[j]);

If you don’t know the bounds my answer isn’t useful but u can try and play with it. Oh, and this specific variation wont work with negative numbers but its not a problem to fix it.

In Java I would solve it like this. Don’t know how to write this in C.

   int length = array.length;
   for (int i = 0; i < length; i++) 
      for (int j = i + 1; j < length; j++) 
         if (array[i] == array[j]) 
            int k, j;
            for (k = j + 1, l = j; k < length; k++, l++) 
               if (array[k] != array[i]) 
                  array[l] = array[k];
            length = l;

Insert all the elements in a binary tree the disregards duplicates – O(nlog(n)). Then extract all of them back in the array by doing a traversal – O(n). I am assuming that you don’t need order preservation.

I’ve posted this once before on SO, but I’ll reproduce it here because it’s pretty cool. It uses hashing, building something like a hash set in place. It’s guaranteed to be O(1) in axillary space (the recursion is a tail call), and is typically O(N) time complexity. The algorithm is as follows:

  1. Take the first element of the array, this will be the sentinel.
  2. Reorder the rest of the array, as much as possible, such that each element is in the position corresponding to its hash. As this step is completed, duplicates will be discovered. Set them equal to sentinel.
  3. Move all elements for which the index is equal to the hash to the beginning of the array.
  4. Move all elements that are equal to sentinel, except the first element of the array, to the end of the array.
  5. What’s left between the properly hashed elements and the duplicate elements will be the elements that couldn’t be placed in the index corresponding to their hash because of a collision. Recurse to deal with these elements.

This can be shown to be O(N) provided no pathological scenario in the hashing: Even if there are no duplicates, approximately 2/3 of the elements will be eliminated at each recursion. Each level of recursion is O(n) where small n is the amount of elements left. The only problem is that, in practice, it’s slower than a quick sort when there are few duplicates, i.e. lots of collisions. However, when there are huge amounts of duplicates, it’s amazingly fast.

Edit: In current implementations of D, hash_t is 32 bits. Everything about this algorithm assumes that there will be very few, if any, hash collisions in full 32-bit space. Collisions may, however, occur frequently in the modulus space. However, this assumption will in all likelihood be true for any reasonably sized data set. If the key is less than or equal to 32 bits, it can be its own hash, meaning that a collision in full 32-bit space is impossible. If it is larger, you simply can’t fit enough of them into 32-bit memory address space for it to be a problem. I assume hash_t will be increased to 64 bits in 64-bit implementations of D, where datasets can be larger. Furthermore, if this ever did prove to be a problem, one could change the hash function at each level of recursion.

Here’s an implementation in the D programming language:

void uniqueInPlace(T)(ref T[] dataIn) {
    uniqueInPlaceImpl(dataIn, 0);

void uniqueInPlaceImpl(T)(ref T[] dataIn, size_t start) {
    if(dataIn.length - start < 2)

    invariant T sentinel = dataIn[start];
    T[] data = dataIn[start + 1..$];

    static hash_t getHash(T elem) {
        static if(is(T == uint) || is(T == int)) {
            return cast(hash_t) elem;
        } else static if(__traits(compiles, elem.toHash)) {
            return elem.toHash;
        } else {
            static auto ti = typeid(typeof(elem));
            return ti.getHash(&elem);

    for(size_t index = 0; index < data.length;) {
        if(data[index] == sentinel) {

        auto hash = getHash(data[index]) % data.length;
        if(index == hash) {

        if(data[index] == data[hash]) {
            data[index] = sentinel;

        if(data[hash] == sentinel) {
            swap(data[hash], data[index]);

        auto hashHash = getHash(data[hash]) % data.length;
        if(hashHash != hash) {
            swap(data[index], data[hash]);
            if(hash < index)
        } else {

    size_t swapPos = 0;
    foreach(i; {
        if(data[i] != sentinel && i == getHash(data[i]) % data.length) {
            swap(data[i], data[swapPos++]);

    size_t sentinelPos = data.length;
    for(size_t i = swapPos; i < sentinelPos;) {
        if(data[i] == sentinel) {
            swap(data[i], data[--sentinelPos]);
        } else {

    dataIn = dataIn[0..sentinelPos + start + 1];
    uniqueInPlaceImpl(dataIn, start + swapPos + 1);
void f(vector<int>& data) {
    data.erase(unique(data.begin(), data.end());

This can be done in one pass with an O(N log N) algorithm and no extra storage.

Proceed from element a[1] to a[N]. At each stage i, all of the elements to the left of a[i] comprise a sorted heap of elements a[0] through a[j]. Meanwhile, a second index j, initially 0, keeps track of the size of the heap.

Examine a[i] and insert it into the heap, which now occupies elements a[0] to a[j+1]. As the element is inserted, if a duplicate element a[k] is encountered having the same value, do not insert a[i] into the heap (i.e., discard it); otherwise insert it into the heap, which now grows by one element and now comprises a[0] to a[j+1], and increment j.

Continue in this manner, incrementing i until all of the array elements have been examined and inserted into the heap, which ends up occupying a[0] to a[j]. j is the index of the last element of the heap, and the heap contains only unique element values.

int algorithm(int[] a, int n)
    int   i, j;  

    for (j = 0, i = 1;  i < n;  i++)
        // Insert a[i] into the heap a[0...j]
        if (heapInsert(a, j, a[i]))
    return j;

bool heapInsert(a[], int n, int val)
    // Insert val into heap a[0...n]
    ...code omitted for brevity...
    if (duplicate element a[k] == val)
        return false;
    a[k] = val;
    return true;

Looking at the example, this is not exactly what was asked for since the resulting array preserves the original element order. But if this requirement is relaxed, the algorithm above should do the trick.

1. Using O(1) extra space, in O(n log n) time

This is possible, for instance:

  • first do an in-place O(n log n) sort
  • then walk through the list once, writing the first instance of every back to the beginning of the list

I believe ejel’s partner is correct that the best way to do this would be an in-place merge sort with a simplified merge step, and that that is probably the intent of the question, if you were eg. writing a new library function to do this as efficiently as possible with no ability to improve the inputs, and there would be cases it would be useful to do so without a hash-table, depending on the sorts of inputs. But I haven’t actually checked this.

2. Using O(lots) extra space, in O(n) time

  • declare a zero’d array big enough to hold all integers
  • walk through the array once
  • set the corresponding array element to 1 for each integer.
  • If it was already 1, skip that integer.

This only works if several questionable assumptions hold:

  • it’s possible to zero memory cheaply, or the size of the ints are small compared to the number of them
  • you’re happy to ask your OS for 256^sizepof(int) memory
  • and it will cache it for you really really efficiently if it’s gigantic

It’s a bad answer, but if you have LOTS of input elements, but they’re all 8-bit integers (or maybe even 16-bit integers) it could be the best way.

3. O(little)-ish extra space, O(n)-ish time

As #2, but use a hash table.

4. The clear way

If the number of elements is small, writing an appropriate algorithm is not useful if other code is quicker to write and quicker to read.

Eg. Walk through the array for each unique elements (ie. the first element, the second element (duplicates of the first having been removed) etc) removing all identical elements. O(1) extra space, O(n^2) time.

Eg. Use library functions which do this. efficiency depends which you have easily available.

How about the following?

int* temp = malloc(sizeof(int)*len);
int count = 0;
int x =0;
int y =0;
        *(temp+count) = *(array+x);
memcpy(array, temp, sizeof(int)*len);

I try to declare a temp array and put the elements into that before copying everything back to the original array.

After review the problem, here is my delphi way, that may help

A: Array of Integer;
I,J,C,K, P: Integer;
A[0]:=1; A[1]:=4; A[2]:=2; A[3]:=6; A[4]:=3; A[5]:=4;
A[6]:=3; A[7]:=4; A[8]:=2; A[9]:=5;

for I := 0 to C-1 do
  for J := I+1 to C-1 do
    if A[I]=A[J] then
      for K := C-1 Downto J do
        if A[J]<>A[k] then

//tructate array

Use bloom filter for hashing. This will reduce the memory overhead very significantly.

Here is a Java Version.

int[] removeDuplicate(int[] input){

        int arrayLen = input.length;
        for(int i=0;i<arrayLen;i++){
            for(int j = i+1; j< arrayLen ; j++){
                if(((input[i]^input[j]) == 0)){
                    input[j] = 0;
                if((input[j]==0) && j<arrayLen-1){
                        input[j] = input[j+1];
                        input[j+1] = 0;
        return input;       

Create a BinarySearchTree which has O(n) complexity.

The following example should solve your problem:

def check_dump(x):
   if not x in t:
      return True


output = filter(check_dump, input)



    Integer[] arrayInteger = {1,2,3,4,3,2,4,6,7,8,9,9,10};

    String value ="";

    for(Integer i:arrayInteger)
            value +=Integer.toString(i)+",";


    String[] arraySplitToString = value.split(",");
    Integer[] arrayIntResult = new Integer[arraySplitToString.length];
    for(int i = 0 ; i < arraySplitToString.length ; i++){
        arrayIntResult[i] = Integer.parseInt(arraySplitToString[i]);

output: { 1, 2, 3, 4, 6, 7, 8, 9, 10}

hope this will help

Integer[] arrayInteger = {1,2,3,4,3,2,4,6,7,8,9,9,10}; 

Set set = new HashSet();
for(Integer i:arrayInteger)


Here is my solution.

///// find duplicates in an array and remove them

void unique(int* input, int n)
     merge_sort(input, 0, n) ;

     int prev = 0  ;

     for(int i = 1 ; i < n ; i++)
          if(input[i] != input[prev])
               if(prev < i-1)
                   input[prev++] = input[i] ;                         

The return value of the function should be the number of unique elements and they are all stored at the front of the array. Without this additional information, you won’t even know if there were any duplicates.

Each iteration of the outer loop processes one element of the array. If it is unique, it stays in the front of the array and if it is a duplicate, it is overwritten by the last unprocessed element in the array. This solution runs in O(n^2) time.

#include <stdio.h>
#include <stdlib.h>

size_t rmdup(int *arr, size_t len)
  size_t prev = 0;
  size_t curr = 1;
  size_t last = len - 1;
  while (curr <= last) {
    for (prev = 0; prev < curr && arr[curr] != arr[prev]; ++prev);
    if (prev == curr) {
    } else {
      arr[curr] = arr[last];
  return curr;

void print_array(int *arr, size_t len)
  size_t curr = 0;
  for (curr = 0; curr < len; ++curr) {
    if (curr > 0) printf(", ");
    printf("%d", arr[curr]);

int main()
  int arr[] = {4, 8, 4, 1, 1, 2, 9};
  printf("Before: ");
  size_t len = sizeof (arr) / sizeof (arr[0]);
  print_array(arr, len);
  len = rmdup(arr, len);
  printf("/nAfter: ");
  print_array(arr, len);
  return 0;
import java.util.ArrayList;

public class C {

    public static void main(String[] args) {

        int arr[] = {2,5,5,5,9,11,11,23,34,34,34,45,45};

        ArrayList<Integer> arr1 = new ArrayList<Integer>();

        for(int i=0;i<arr.length-1;i++){

            if(arr[i] == arr[i+1]){
                arr[i] = 99999;

        for(int i=0;i<arr.length;i++){
            if(arr[i] != 99999){



This is the naive (N*(N-1)/2) solution. It uses constant additional space and maintains the original order. It is similar to the solution by @Byju, but uses no if(){} blocks. It also avoids copying an element onto itself.

#include <stdio.h>
#include <stdlib.h>

int numbers[] = {4, 8, 4, 1, 1, 2, 9};
#define COUNT (sizeof numbers / sizeof numbers[0])

size_t undup_it(int array[], size_t len)
size_t src,dst;

  /* an array of size=1 cannot contain duplicate values */
if (len <2) return len; 
  /* an array of size>1 will cannot at least one unique value */
for (src=dst=1; src < len; src++) {
        size_t cur;
        for (cur=0; cur < dst; cur++ ) {
                if (array[cur] == array[src]) break;
        if (cur != dst) continue; /* found a duplicate */

                /* array[src] must be new: add it to the list of non-duplicates */
        if (dst < src) array[dst] = array[src]; /* avoid copy-to-self */
return dst; /* number of valid alements in new array */

void print_it(int array[], size_t len)
size_t idx;

for (idx=0; idx < len; idx++)  {
        printf("%c %d", (idx) ? ',' :'{' , array[idx] );
printf("}/n" );

int main(void) {    
    size_t cnt = COUNT;

    printf("Before undup:" );    
    print_it(numbers, cnt);    

    cnt = undup_it(numbers,cnt);

    printf("After undup:" );    
    print_it(numbers, cnt);

    return 0;

For someone who want to have simple solution in C++:

int* rmdup(int path[], int start, int end, int& newEnd) {
    int ret[100];
newEnd = end;
int j = start;

for (int i = start; i < end; i++) {
    if (path[i] == path[i+1]) {
    ret[j++] = path[i];

ret[j++] = path[end];

for(int i = start; i <= newEnd; i++)
     path[i] = ret[i];

First, you should create an array check[n] where n is the number of elements of the array you want to make duplicate-free and set the value of every element(of the check array) equal to 1. Using a for loop traverse the array with the duplicates, say its name is arr, and in the for-loop write this :

    if (check[arr[i]] != 1) {
        arr[i] = 0;
    else {
        check[arr[i]] = 0;

With that, you set every duplicate equal to zero. So the only thing is left to do is to traverse the arr array and print everything it’s not equal to zero. The order stays and it takes linear time (3*n).

Simply take a variable x=arr[0] and do xor operation by traversing the rest of the elements . If an element has repeated then the x will become zero.

This way we know that the element has repeated previously . This also will just take o(n) to scan through all of the elements in the original array.

Given an array of n elements, write an algorithm to remove all duplicates from the array in time O(nlogn)

Algorithm delete_duplicates (a[1....n])
//Remove duplicates from the given array 
//input parameters :a[1:n], an array of n elements.


temp[1:n]; //an array of n elements. 

temp[i]=a[i];for i=1 to n



 //based on 'value' sort the array temp.

//based on 'value' delete duplicate elements from temp.

//based on 'key' sort the array temp.//construct an array p using temp.


  return p.

In other of elements is maintained in the output array using the ‘key’. Consider the key is of length O(n), the time taken for performing sorting on the key and value is O(nlogn). So the time taken to delete all duplicates from the array is O(nlogn).

this is what i’ve got, though it misplaces the order we can sort in ascending or descending to fix it up.

#include <stdio.h>
int main(void){
int x,n,myvar=0;
printf("Enter a number: /t");
int arr[n],changedarr[n];

    printf("Enter a number for array[%d]: ",x);
printf("/nOriginal Number in an array/n");

int i=0,j=0;
// printf("i/tj/tarr/tchanged/n");

for (int i = 0; i < n; i++)
    // printf("%d/t%d/t%d/t%d/n",i,j,arr[i],changedarr[i] );
    for (int j = 0; j <n; j++)
        if (i==j)

        else if(arr[i]==arr[j]){


    // printf("%d/t%d/t%d/t%d/n",i,j,arr[i],changedarr[i] );
// printf("/n/nmyvar=%d/n",myvar);
int count=0;
printf("/nThe unique items:/n");
for (int i = 0; i < myvar; i++)