What's the most efficient way to erase duplicates and sort a vector?

Question

I need to take a C++ vector with potentially a lot of elements, erase duplicates, and sort it.

I currently have the below code, but it doesn't work.

vec.erase(
      std::unique(vec.begin(), vec.end()),
      vec.end());
std::sort(vec.begin(), vec.end());

How can I correctly do this?

Additionally, is it faster to erase the duplicates first (similar to coded above) or perform the sort first? If I do perform the sort first, is it guaranteed to remain sorted after std::unique is executed?

Or is there another (perhaps more efficient) way to do all this?

Answer 1

According to my benchmark, a vector index sort based on this answer is fastest to implement deduplication.

• Fast: Beats common and opitimised hash sets (std::unordered_set, phmap:flat_hash_set).
• Minimal memory usage: 1.5 size_t per element (due to stable_sort).
• Stable (order-preserving). Non-stable availabe (10% faster, only 1 size_t per element).
• Few assumptions: No requirement that the type to be deduplicated can be converted to an integer. Uses normal binary comparison instead, so it works for high-dimensional data types.

When are hash sets faster?

• When there are few distinct elements (the hash set never grows larger than the CPU cache).

A minimal implementation from that answer is below.

I converted it to a more general, commented implementation here: nh2/cpp-dedup-benchmark

#include <algorithm>
#include <vector>

template<class It> It uniquify(It begin, It const end)
{
    std::vector<It> v;
    v.reserve(static_cast<size_t>(std::distance(begin, end)));
    for (It i = begin; i != end; ++i)
        v.push_back(i);
    std::stable_sort(v.begin(), v.end(), [](It const &a, It const &b){ return *a < *b; });
    v.erase(std::unique(v.begin(), v.end(), [](It const &a, It const &b) { return *a == *b; }), v.end());
    std::sort(v.begin(), v.end());
    size_t j = 0;
    for (It i = begin; i != end && j != v.size(); ++i)
    {
        if (i == v[j])
        {
            std::iter_swap(i, begin);
            ++j;
            ++begin;
        }
    }
    return begin;
}

Benchmark results from https://github.com/nh2/cpp-dedup-benchmark:

Memory usage

Measured:

what                      Bytes per element    Notes

Input data type (Point)      11                = 3 double + 3 char

stable_unique_iterators      12                = 1.5 size_t due to std::stable_sort (e.g. merge sort buffer)
unstable_unique_iterators     8                = 1 size_t
unordered_set                56
flat_hash_set                32

Answer 2

The following simple method of tracking iterator positions that has time complexity of O(nlogn) and space complexity of O(1) seems like has not been mentioned earlier:

#include <iostream>
#include <vector>
#include <algorithm>

using namespace std;

void removeDups(vector<int>& vect)
{
  int u = 0;
  sort(vect.begin(), vect.end());
  for (int i = 1; i < vect.size(); ) {
    if (vect[u] == vect[i]) {
      ++i;
    }
    else {
      ++u;
      swap(vect[u],vect[i]);
      ++i;
    }


  }
  vect.erase(vect.begin() + u + 1, vect.end() );
}


int main()
{
  vector<int> vect{ 3, 5, 7, 2, 2, 5, 7, 7, 9, 2, 3, 5, 7, 9, 3, 5, 7, 9 };
  removeDups(vect);
  for (int i = 0; i < vect.size(); i++)
    cout << vect[i] << " ";
  return 0;
}

Answer 3

That's the problem with naming in the std... std::unique should rather be called std::trim_consecutive_duplicates imho, that would make it clear that you need to sort the vector first to have elements with the same value adjacent to each other. In this case I doubt that anything related to a set is fasteor when arriving from vector, but if you have the opportunity to put everything into a set from the start you should definitely do that.

Here's a modern C++20 example (Demo):

#include <algorithm>
#include <iostream>
#include <ranges>
#include <vector>
#include <cstdint>

namespace rng = std::ranges;

int main() {
    std::vector<uint32_t> myvec = { 255, 1,3, 16, 5,6, 1, 3, 3, 255, 300 };

    rng::sort(myvec);
    const auto [first, last] = rng::unique(myvec);
    myvec.erase(first, last);

    // Print resulting vector
    std::cout << "my unique vector = {";
    rng::for_each(myvec, [](uint32_t val){ std::cout << val << ", "; });
    std::cout << "}" << std::endl;
}

Output:

my unique vector = {1, 3, 5, 6, 16, 255, 300, }

Answer 4

I agree with R. Pate and Todd Gardner; a std::set might be a good idea here. Even if you're stuck using vectors, if you have enough duplicates, you might be better off creating a set to do the dirty work.

Let's compare three approaches:

Just using vector, sort + unique

sort( vec.begin(), vec.end() );
vec.erase( unique( vec.begin(), vec.end() ), vec.end() );

Convert to set (manually)

set<int> s;
unsigned size = vec.size();
for( unsigned i = 0; i < size; ++i ) s.insert( vec[i] );
vec.assign( s.begin(), s.end() );

Convert to set (using a constructor)

set<int> s( vec.begin(), vec.end() );
vec.assign( s.begin(), s.end() );

Here's how these perform as the number of duplicates changes:

Summary: when the number of duplicates is large enough, it's actually faster to convert to a set and then dump the data back into a vector.

And for some reason, doing the set conversion manually seems to be faster than using the set constructor -- at least on the toy random data that I used.

Answer 5

Depends on use-case. If you expect less than 100 amount of positive integer unique values and if you have a cpu capable of avx512f instruction set, then you can insert elements at a rate of ~15 clock cycles per element or 300-500 million inserts per second, by using simple comparison against a small look-up-table.

Following implementation uses CPU registers to do value lookups for ~50 unique values and L1 cache for ~1000 unique values. For L1 cache version, it takes about 160 clock cycles per insertion which is equivalent to about ~25M insertions per second and becomes slower than using std::set. For only 4 unique values, it inserts at a rate of 5.8 cycles per element which is higher than 500M/s.

//g++  7.4.0
// time measurement taken from another answer
// valid C99 and C++

#include <stdint.h>  // <cstdint> is preferred in C++, but stdint.h works.

#ifdef _MSC_VER
# include <intrin.h>
#else
# include <x86intrin.h>
#endif

// optional wrapper if you don't want to just use __rdtsc() everywhere
inline
uint64_t readTSC() {
     _mm_lfence();  // optionally wait for earlier insns to retire before reading the clock
    uint64_t tsc = __rdtsc();
     _mm_lfence();  // optionally block later instructions until rdtsc retires
    return tsc;
}

// requires a Nehalem or newer CPU.  Not Core2 or earlier.  IDK when AMD added it.
inline
uint64_t readTSCp() {
    unsigned dummy;
    return __rdtscp(&dummy);  // waits for earlier insns to retire, but allows later to start
}



#include <iostream>

template<int n>
struct FastUnique
{
    public:
    FastUnique()
    {
         it=0;
         for(int i=0;i<n;i++)
             dict[i]=-1;
    }

    void insert(const int val)
    {
        if(!test(dict,val))
            dict[it++]=val;
    }

    const int get(const int index)
    {
        return dict[index];
    }

    const int size()
    {
        return it;
    }

    private:
    int dict[n];
    int it;
    bool test(const int * dict, const int val)
    {
        int c=0;
        for(int i=0;i<n;i++)
            c+=(dict[i]==val);
        return c>0;
    }
};

int main()
{
    std::cout << "Hello, world!\n";
    const int n=500000000;

    FastUnique<64> fastSet;

    auto t= readTSC();

    for(int i=0;i<n;i++)
        fastSet.insert(i&63);

    auto t2=readTSC();

    std::cout<<(t2-t)/(double)n<<"cycles per iteration"<<std::endl;
   
    for(int i=0;i<fastSet.size();i++)
        std::cout<<fastSet.get(i)<<std::endl;
    
    return 0;
}

Answer 6

Most of the answer seems to be using O(nlogn) but with the use of the unordered_set we can decrease it to O(n). I saw some of the solutions using sets, but I found this one and it seems more elegant to use set and iterators.

using Intvec = std::vector<int>;

void remove(Intvec &v) {
    // creating iterator starting with beginning of the vector 
    Intvec::iterator itr = v.begin();
    std::unordered_set<int> s;
    // loops from the beginning to the end of the list 
    for (auto curr = v.begin(); curr != v.end(); ++curr) {
        if (s.insert(*curr).second) { // if the 0 curr already exist in the set
            *itr++ = *curr; // adding a position to the iterator 
        }
    }
    // erasing repeating positions in the set 
    v.erase(itr, v.end());
}

Answer 7

If your class is easily convertible to an int, and you got some memory, unique can be done without sorting before, and it's much faster :

#include <vector>
#include <stdlib.h>
#include <algorithm>
int main (int argc, char* argv []) {
  //vector init
  std::vector<int> v (1000000, 0);
  std::for_each (v.begin (), v.end (), [] (int& s) {s = rand () %1000;});
  std::vector<int> v1 (v);
  int beg (0), end (0), duration (0);
  beg = clock ();
  {
    std::sort (v.begin (), v.end ());
    auto i (v.begin ());
    i = std::unique (v.begin (), v.end ());
    if (i != v.end ()) v.erase (i, v.end ());
  }
  end = clock ();
  duration = (int) (end - beg);
  std::cout << "\tduration sort + unique == " << duration << std::endl;

  int n (0);
  duration = 0;
  beg = clock ();
  std::for_each (v1.begin (), v1.end (), [&n] (const int& s) {if (s >= n) n = s+1;});
  std::vector<int> tab (n, 0);
  {
    auto i (v1.begin ());
    std::for_each (v1.begin (), v1.end (), [&i, &tab] (const int& s) {
      if (!tab [s]) {
        *i++ = s;
        ++tab [s];
      }
    });
    std::sort (v1.begin (), i);
    v1.erase (i, v1.end ());
  }
  end = clock ();
  duration = (int) (end - beg);
  std::cout << "\tduration unique + sort == " << duration << std::endl;
  if (v == v1) {
    std::cout << "and results are same" << std::endl;
  }
  else {
    std::cout << "but result differs" << std::endl;
  }  
}

Typical results : duration sort + unique == 38985 duration unique + sort == 2500 and results are same

Answer 8

With the Ranges v3 library, you can simply use

action::unique(vec);

Note that it actually removes the duplicate elements, not just move them.

Unfortunately, actions weren’t standardized in C++20 as other parts of the ranges library were you still have to use the original library even in C++20.

Answer 9

You can do this as follows:

std::sort(v.begin(), v.end());
v.erase(std::unique(v.begin(), v.end()), v.end());

Answer 10

void removeDuplicates(std::vector<int>& arr) {
    for (int i = 0; i < arr.size(); i++)
    {
        for (int j = i + 1; j < arr.size(); j++)
        {
            if (arr[i] > arr[j])
            {
                int temp = arr[i];
                arr[i] = arr[j];
                arr[j] = temp;
            }
        }
    }
    std::vector<int> y;
    int x = arr[0];
    int i = 0;
    while (i < arr.size())
    {
        if (x != arr[i])
        {
            y.push_back(x);
            x = arr[i];
        }
        i++;
        if (i == arr.size())
            y.push_back(arr[i - 1]);
    }
    arr = y;
}

Answer 11

Assuming that a is a vector, remove the contiguous duplicates using

a.erase(unique(a.begin(),a.end()),a.end()); runs in O(n) time.

Answer 12

More understandable code from: https://en.cppreference.com/w/cpp/algorithm/unique

#include <iostream>
#include <algorithm>
#include <vector>
#include <string>
#include <cctype>

int main() 
{
    // remove duplicate elements
    std::vector<int> v{1,2,3,1,2,3,3,4,5,4,5,6,7};
    std::sort(v.begin(), v.end()); // 1 1 2 2 3 3 3 4 4 5 5 6 7 
    auto last = std::unique(v.begin(), v.end());
    // v now holds {1 2 3 4 5 6 7 x x x x x x}, where 'x' is indeterminate
    v.erase(last, v.end()); 
    for (int i : v)
      std::cout << i << " ";
    std::cout << "\n";
}

ouput:

1 2 3 4 5 6 7

Answer 13

void EraseVectorRepeats(vector <int> & v){ 
TOP:for(int y=0; y<v.size();++y){
        for(int z=0; z<v.size();++z){
            if(y==z){ //This if statement makes sure the number that it is on is not erased-just skipped-in order to keep only one copy of a repeated number
                continue;}
            if(v[y]==v[z]){
                v.erase(v.begin()+z); //whenever a number is erased the function goes back to start of the first loop because the size of the vector changes
            goto TOP;}}}}

This is a function that I created that you can use to delete repeats. The header files needed are just <iostream> and <vector>.

Answer 14

If you are looking for performance and using std::vector, I recommend the one that this documentation link provides.

std::vector<int> myvector{10,20,20,20,30,30,20,20,10};             // 10 20 20 20 30 30 20 20 10
std::sort(myvector.begin(), myvector.end() );
const auto& it = std::unique (myvector.begin(), myvector.end());   // 10 20 30 ?  ?  ?  ?  ?  ?
                                                                   //          ^
myvector.resize( std::distance(myvector.begin(),it) ); // 10 20 30

Answer 15

I redid Nate Kohl's profiling and got different results. For my test case, directly sorting the vector is always more efficient than using a set. I added a new more efficient method, using an unordered_set.

Keep in mind that the unordered_set method only works if you have a good hash function for the type you need uniqued and sorted. For ints, this is easy! (The standard library provides a default hash which is simply the identity function.) Also, don't forget to sort at the end since unordered_set is, well, unordered :)

I did some digging inside the set and unordered_set implementation and discovered that the constructor actually construct a new node for every element, before checking its value to determine if it should actually be inserted (in Visual Studio implementation, at least).

Here are the 5 methods:

f1: Just using vector, sort + unique

sort( vec.begin(), vec.end() );
vec.erase( unique( vec.begin(), vec.end() ), vec.end() );

f2: Convert to set (using a constructor)

set<int> s( vec.begin(), vec.end() );
vec.assign( s.begin(), s.end() );

f3: Convert to set (manually)

set<int> s;
for (int i : vec)
    s.insert(i);
vec.assign( s.begin(), s.end() );

f4: Convert to unordered_set (using a constructor)

unordered_set<int> s( vec.begin(), vec.end() );
vec.assign( s.begin(), s.end() );
sort( vec.begin(), vec.end() );

f5: Convert to unordered_set (manually)

unordered_set<int> s;
for (int i : vec)
    s.insert(i);
vec.assign( s.begin(), s.end() );
sort( vec.begin(), vec.end() );

I did the test with a vector of 100,000,000 ints chosen randomly in ranges [1,10], [1,1000], and [1,100000]

The results (in seconds, smaller is better):

range         f1       f2       f3       f4      f5
[1,10]      1.6821   7.6804   2.8232   6.2634  0.7980
[1,1000]    5.0773  13.3658   8.2235   7.6884  1.9861
[1,100000]  8.7955  32.1148  26.5485  13.3278  3.9822

Answer 16

std::unique only removes duplicate elements if they're neighbours: you have to sort the vector first before it will work as you intend.

std::unique is defined to be stable, so the vector will still be sorted after running unique on it.

Answer 17

If you do not want to change the order of elements, then you can try this solution:

template <class T>
void RemoveDuplicatesInVector(std::vector<T> & vec)
{
    set<T> values;
    vec.erase(std::remove_if(vec.begin(), vec.end(), [&](const T & value) { return !values.insert(value).second; }), vec.end());
}

Answer 18

About alexK7 benchmarks. I tried them and got similar results, but when the range of values is 1 million the cases using std::sort (f1) and using std::unordered_set (f5) produce similar time. When the range of values is 10 million f1 is faster than f5.

If the range of values is limited and the values are unsigned int, it is possible to use std::vector, the size of which corresponds to the given range. Here is the code:

void DeleteDuplicates_vector_bool(std::vector<unsigned>& v, unsigned range_size)
{
    std::vector<bool> v1(range_size);
    for (auto& x: v)
    {
       v1[x] = true;    
    }
    v.clear();

    unsigned count = 0;
    for (auto& x: v1)
    {
        if (x)
        {
            v.push_back(count);
        }
        ++count;
    }
}

Answer 19

The standard approach suggested by Nate Kohl, just using vector, sort + unique:

sort( vec.begin(), vec.end() );
vec.erase( unique( vec.begin(), vec.end() ), vec.end() );

doesn't work for a vector of pointers.

Look carefully at this example on cplusplus.com.

In their example, the "so called duplicates" moved to the end are actually shown as ? (undefined values), because those "so called duplicates" are SOMETIMES "extra elements" and SOMETIMES there are "missing elements" that were in the original vector.

A problem occurs when using std::unique() on a vector of pointers to objects (memory leaks, bad read of data from HEAP, duplicate frees, which cause segmentation faults, etc).

Here's my solution to the problem: replace std::unique() with ptgi::unique().

See the file ptgi_unique.hpp below:

// ptgi::unique()
//
// Fix a problem in std::unique(), such that none of the original elts in the collection are lost or duplicate.
// ptgi::unique() has the same interface as std::unique()
//
// There is the 2 argument version which calls the default operator== to compare elements.
//
// There is the 3 argument version, which you can pass a user defined functor for specialized comparison.
//
// ptgi::unique() is an improved version of std::unique() which doesn't looose any of the original data
// in the collection, nor does it create duplicates.
//
// After ptgi::unique(), every old element in the original collection is still present in the re-ordered collection,
// except that duplicates have been moved to a contiguous range [dupPosition, last) at the end.
//
// Thus on output:
//  [begin, dupPosition) range are unique elements.
//  [dupPosition, last) range are duplicates which can be removed.
// where:
//  [] means inclusive, and
//  () means exclusive.
//
// In the original std::unique() non-duplicates at end are moved downward toward beginning.
// In the improved ptgi:unique(), non-duplicates at end are swapped with duplicates near beginning.
//
// In addition if you have a collection of ptrs to objects, the regular std::unique() will loose memory,
// and can possibly delete the same pointer multiple times (leading to SEGMENTATION VIOLATION on Linux machines)
// but ptgi::unique() won't.  Use valgrind(1) to find such memory leak problems!!!
//
// NOTE: IF you have a vector of pointers, that is, std::vector<Object*>, then upon return from ptgi::unique()
// you would normally do the following to get rid of the duplicate objects in the HEAP:
//
//  // delete objects from HEAP
//  std::vector<Object*> objects;
//  for (iter = dupPosition; iter != objects.end(); ++iter)
//  {
//      delete (*iter);
//  }
//
//  // shrink the vector. But Object * pointers are NOT followed for duplicate deletes, this shrinks the vector.size())
//  objects.erase(dupPosition, objects.end));
//
// NOTE: But if you have a vector of objects, that is: std::vector<Object>, then upon return from ptgi::unique(), it
// suffices to just call vector:erase(, as erase will automatically call delete on each object in the
// [dupPosition, end) range for you:
//
//  std::vector<Object> objects;
//  objects.erase(dupPosition, last);
//
//==========================================================================================================
// Example of differences between std::unique() vs ptgi::unique().
//
//  Given:
//      int data[] = {10, 11, 21};
//
//  Given this functor: ArrayOfIntegersEqualByTen:
//      A functor which compares two integers a[i] and a[j] in an int a[] array, after division by 10:
//  
//  // given an int data[] array, remove consecutive duplicates from it.
//  // functor used for std::unique (BUGGY) or ptgi::unique(IMPROVED)
//
//  // Two numbers equal if, when divided by 10 (integer division), the quotients are the same.
//  // Hence 50..59 are equal, 60..69 are equal, etc.
//  struct ArrayOfIntegersEqualByTen: public std::equal_to<int>
//  {
//      bool operator() (const int& arg1, const int& arg2) const
//      {
//          return ((arg1/10) == (arg2/10));
//      }
//  };
//  
//  Now, if we call (problematic) std::unique( data, data+3, ArrayOfIntegersEqualByTen() );
//  
//  TEST1: BEFORE UNIQ: 10,11,21
//  TEST1: AFTER UNIQ: 10,21,21
//  DUP_INX=2
//  
//      PROBLEM: 11 is lost, and extra 21 has been added.
//  
//  More complicated example:
//  
//  TEST2: BEFORE UNIQ: 10,20,21,22,30,31,23,24,11
//  TEST2: AFTER UNIQ: 10,20,30,23,11,31,23,24,11
//  DUP_INX=5
//  
//      Problem: 21 and 22 are deleted.
//      Problem: 11 and 23 are duplicated.
//  
//  
//  NOW if ptgi::unique is called instead of std::unique, both problems go away:
//  
//  DEBUG: TEST1: NEW_WAY=1
//  TEST1: BEFORE UNIQ: 10,11,21
//  TEST1: AFTER UNIQ: 10,21,11
//  DUP_INX=2
//  
//  DEBUG: TEST2: NEW_WAY=1
//  TEST2: BEFORE UNIQ: 10,20,21,22,30,31,23,24,11
//  TEST2: AFTER UNIQ: 10,20,30,23,11,31,22,24,21
//  DUP_INX=5
//
//  @SEE: look at the "case study" below to understand which the last "AFTER UNIQ" results with that order:
//  TEST2: AFTER UNIQ: 10,20,30,23,11,31,22,24,21
//
//==========================================================================================================
// Case Study: how ptgi::unique() works:
//  Remember we "remove adjacent duplicates".
//  In this example, the input is NOT fully sorted when ptgi:unique() is called.
//
//  I put | separatators, BEFORE UNIQ to illustrate this
//  10  | 20,21,22 |  30,31 |  23,24 | 11
//
//  In example above, 20, 21, 22 are "same" since dividing by 10 gives 2 quotient.
//  And 30,31 are "same", since /10 quotient is 3.
//  And 23, 24 are same, since /10 quotient is 2.
//  And 11 is "group of one" by itself.
//  So there are 5 groups, but the 4th group (23, 24) happens to be equal to group 2 (20, 21, 22)
//  So there are 5 groups, and the 5th group (11) is equal to group 1 (10)
//
//  R = result
//  F = first
//
//  10, 20, 21, 22, 30, 31, 23, 24, 11
//  R    F
//
//  10 is result, and first points to 20, and R != F (10 != 20) so bump R:
//       R
//       F
//
//  Now we hits the "optimized out swap logic".
//  (avoid swap because R == F)
//
//  // now bump F until R != F (integer division by 10)
//  10, 20, 21, 22, 30, 31, 23, 24, 11
//       R   F              // 20 == 21 in 10x
//       R       F              // 20 == 22 in 10x
//       R           F          // 20 != 30, so we do a swap of ++R and F
//  (Now first hits 21, 22, then finally 30, which is different than R, so we swap bump R to 21 and swap with  30)
//  10, 20, 30, 22, 21, 31, 23, 24, 11  // after R & F swap (21 and 30)
//           R       F 
//
//  10, 20, 30, 22, 21, 31, 23, 24, 11
//           R          F           // bump F to 31, but R and F are same (30 vs 31)
//           R               F      // bump F to 23, R != F, so swap ++R with F
//  10, 20, 30, 22, 21, 31, 23, 24, 11
//                  R           F       // bump R to 22
//  10, 20, 30, 23, 21, 31, 22, 24, 11  // after the R & F swap (22 & 23 swap)
//                  R            F      // will swap 22 and 23
//                  R                F      // bump F to 24, but R and F are same in 10x
//                  R                    F  // bump F, R != F, so swap ++R  with F
//                      R                F  // R and F are diff, so swap ++R  with F (21 and 11)
//  10, 20, 30, 23, 11, 31, 22, 24, 21
//                      R                F  // aftter swap of old 21 and 11
//                      R                  F    // F now at last(), so loop terminates
//                          R               F   // bump R by 1 to point to dupPostion (first duplicate in range)
//
//  return R which now points to 31
//==========================================================================================================
// NOTES:
// 1) the #ifdef IMPROVED_STD_UNIQUE_ALGORITHM documents how we have modified the original std::unique().
// 2) I've heavily unit tested this code, including using valgrind(1), and it is *believed* to be 100% defect-free.
//
//==========================================================================================================
// History:
//  130201  dpb dbednar@ptgi.com created
//==========================================================================================================

#ifndef PTGI_UNIQUE_HPP
#define PTGI_UNIQUE_HPP

// Created to solve memory leak problems when calling std::unique() on a vector<Route*>.
// Memory leaks discovered with valgrind and unitTesting.


#include <algorithm>        // std::swap

// instead of std::myUnique, call this instead, where arg3 is a function ptr
//
// like std::unique, it puts the dups at the end, but it uses swapping to preserve original
// vector contents, to avoid memory leaks and duplicate pointers in vector<Object*>.

#ifdef IMPROVED_STD_UNIQUE_ALGORITHM
#error the #ifdef for IMPROVED_STD_UNIQUE_ALGORITHM was defined previously.. Something is wrong.
#endif

#undef IMPROVED_STD_UNIQUE_ALGORITHM
#define IMPROVED_STD_UNIQUE_ALGORITHM

// similar to std::unique, except that this version swaps elements, to avoid
// memory leaks, when vector contains pointers.
//
// Normally the input is sorted.
// Normal std::unique:
// 10 20 20 20 30   30 20 20 10
// a  b  c  d  e    f  g  h  i
//
// 10 20 30 20 10 | 30 20 20 10
// a  b  e  g  i    f  g  h  i
//
// Now GONE: c, d.
// Now DUPS: g, i.
// This causes memory leaks and segmenation faults due to duplicate deletes of same pointer!


namespace ptgi {

// Return the position of the first in range of duplicates moved to end of vector.
//
// uses operator==  of class for comparison
//
// @param [first, last) is a range to find duplicates within.
//
// @return the dupPosition position, such that [dupPosition, end) are contiguous
// duplicate elements.
// IF all items are unique, then it would return last.
//
template <class ForwardIterator>
ForwardIterator unique( ForwardIterator first, ForwardIterator last)
{
    // compare iterators, not values
    if (first == last)
        return last;

    // remember the current item that we are looking at for uniqueness
    ForwardIterator result = first;

    // result is slow ptr where to store next unique item
    // first is  fast ptr which is looking at all elts

    // the first iterator moves over all elements [begin+1, end).
    // while the current item (result) is the same as all elts
    // to the right, (first) keeps going, until you find a different
    // element pointed to by *first.  At that time, we swap them.

    while (++first != last)
    {
        if (!(*result == *first))
        {
#ifdef IMPROVED_STD_UNIQUE_ALGORITHM
            // inc result, then swap *result and *first

//          THIS IS WHAT WE WANT TO DO.
//          BUT THIS COULD SWAP AN ELEMENT WITH ITSELF, UNCECESSARILY!!!
//          std::swap( *first, *(++result));

            // BUT avoid swapping with itself when both iterators are the same
            ++result;
            if (result != first)
                std::swap( *first, *result);
#else
            // original code found in std::unique()
            // copies unique down
            *(++result) = *first;
#endif
        }
    }

    return ++result;
}

template <class ForwardIterator, class BinaryPredicate>
ForwardIterator unique( ForwardIterator first, ForwardIterator last, BinaryPredicate pred)
{
    if (first == last)
        return last;

    // remember the current item that we are looking at for uniqueness
    ForwardIterator result = first;

    while (++first != last)
    {
        if (!pred(*result,*first))
        {
#ifdef IMPROVED_STD_UNIQUE_ALGORITHM
            // inc result, then swap *result and *first

//          THIS COULD SWAP WITH ITSELF UNCECESSARILY
//          std::swap( *first, *(++result));
//
            // BUT avoid swapping with itself when both iterators are the same
            ++result;
            if (result != first)
                std::swap( *first, *result);

#else
            // original code found in std::unique()
            // copies unique down
            // causes memory leaks, and duplicate ptrs
            // and uncessarily moves in place!
            *(++result) = *first;
#endif
        }
    }

    return ++result;
}

// from now on, the #define is no longer needed, so get rid of it
#undef IMPROVED_STD_UNIQUE_ALGORITHM

} // end ptgi:: namespace

#endif

And here is the UNIT Test program that I used to test it:

// QUESTION: in test2, I had trouble getting one line to compile,which was caused  by the declaration of operator()
// in the equal_to Predicate.  I'm not sure how to correctly resolve that issue.
// Look for //OUT lines
//
// Make sure that NOTES in ptgi_unique.hpp are correct, in how we should "cleanup" duplicates
// from both a vector<Integer> (test1()) and vector<Integer*> (test2).
// Run this with valgrind(1).
//
// In test2(), IF we use the call to std::unique(), we get this problem:
//
//  [dbednar@ipeng8 TestSortRoutes]$ ./Main7
//  TEST2: ORIG nums before UNIQUE: 10, 20, 21, 22, 30, 31, 23, 24, 11
//  TEST2: modified nums AFTER UNIQUE: 10, 20, 30, 23, 11, 31, 23, 24, 11
//  INFO: dupInx=5
//  TEST2: uniq = 10
//  TEST2: uniq = 20
//  TEST2: uniq = 30
//  TEST2: uniq = 33427744
//  TEST2: uniq = 33427808
//  Segmentation fault (core dumped)
//
// And if we run valgrind we seen various error about "read errors", "mismatched free", "definitely lost", etc.
//
//  valgrind --leak-check=full ./Main7
//  ==359== Memcheck, a memory error detector
//  ==359== Command: ./Main7
//  ==359== Invalid read of size 4
//  ==359== Invalid free() / delete / delete[]
//  ==359== HEAP SUMMARY:
//  ==359==     in use at exit: 8 bytes in 2 blocks
//  ==359== LEAK SUMMARY:
//  ==359==    definitely lost: 8 bytes in 2 blocks
// But once we replace the call in test2() to use ptgi::unique(), all valgrind() error messages disappear.
//
// 130212   dpb dbednar@ptgi.com created
// =========================================================================================================

#include <iostream> // std::cout, std::cerr
#include <string>
#include <vector>   // std::vector
#include <sstream>  // std::ostringstream
#include <algorithm>    // std::unique()
#include <functional>   // std::equal_to(), std::binary_function()
#include <cassert>  // assert() MACRO

#include "ptgi_unique.hpp"  // ptgi::unique()



// Integer is small "wrapper class" around a primitive int.
// There is no SETTER, so Integer's are IMMUTABLE, just like in JAVA.

class Integer
{
private:
    int num;
public:

    // default CTOR: "Integer zero;"
    // COMPRENSIVE CTOR:  "Integer five(5);"
    Integer( int num = 0 ) :
        num(num)
    {
    }

    // COPY CTOR
    Integer( const Integer& rhs) :
        num(rhs.num)
    {
    }

    // assignment, operator=, needs nothing special... since all data members are primitives

    // GETTER for 'num' data member
    // GETTER' are *always* const
    int getNum() const
    {
        return num;
    }   

    // NO SETTER, because IMMUTABLE (similar to Java's Integer class)

    // @return "num"
    // NB: toString() should *always* be a const method
    //
    // NOTE: it is probably more efficient to call getNum() intead
    // of toString() when printing a number:
    //
    // BETTER to do this:
    //  Integer five(5);
    //  std::cout << five.getNum() << "\n"
    // than this:
    //  std::cout << five.toString() << "\n"

    std::string toString() const
    {
        std::ostringstream oss;
        oss << num;
        return oss.str();
    }
};

// convenience typedef's for iterating over std::vector<Integer>
typedef std::vector<Integer>::iterator      IntegerVectorIterator;
typedef std::vector<Integer>::const_iterator    ConstIntegerVectorIterator;

// convenience typedef's for iterating over std::vector<Integer*>
typedef std::vector<Integer*>::iterator     IntegerStarVectorIterator;
typedef std::vector<Integer*>::const_iterator   ConstIntegerStarVectorIterator;

// functor used for std::unique or ptgi::unique() on a std::vector<Integer>
// Two numbers equal if, when divided by 10 (integer division), the quotients are the same.
// Hence 50..59 are equal, 60..69 are equal, etc.
struct IntegerEqualByTen: public std::equal_to<Integer>
{
    bool operator() (const Integer& arg1, const Integer& arg2) const
    {
        return ((arg1.getNum()/10) == (arg2.getNum()/10));
    }
};

// functor used for std::unique or ptgi::unique on a std::vector<Integer*>
// Two numbers equal if, when divided by 10 (integer division), the quotients are the same.
// Hence 50..59 are equal, 60..69 are equal, etc.
struct IntegerEqualByTenPointer: public std::equal_to<Integer*>
{
    // NB: the Integer*& looks funny to me!
    // TECHNICAL PROBLEM ELSEWHERE so had to remove the & from *&
//OUT   bool operator() (const Integer*& arg1, const Integer*& arg2) const
//
    bool operator() (const Integer* arg1, const Integer* arg2) const
    {
        return ((arg1->getNum()/10) == (arg2->getNum()/10));
    }
};

void test1();
void test2();
void printIntegerStarVector( const std::string& msg, const std::vector<Integer*>& nums );

int main()
{
    test1();
    test2();
    return 0;
}

// test1() uses a vector<Object> (namely vector<Integer>), so there is no problem with memory loss
void test1()
{
    int data[] = { 10, 20, 21, 22, 30, 31, 23, 24, 11};

    // turn C array into C++ vector
    std::vector<Integer> nums(data, data+9);

    // arg3 is a functor
    IntegerVectorIterator dupPosition = ptgi::unique( nums.begin(), nums.end(), IntegerEqualByTen() );

    nums.erase(dupPosition, nums.end());

    nums.erase(nums.begin(), dupPosition);
}

//==================================================================================
// test2() uses a vector<Integer*>, so after ptgi:unique(), we have to be careful in
// how we eliminate the duplicate Integer objects stored in the heap.
//==================================================================================
void test2()
{
    int data[] = { 10, 20, 21, 22, 30, 31, 23, 24, 11};

    // turn C array into C++ vector of Integer* pointers
    std::vector<Integer*> nums;

    // put data[] integers into equivalent Integer* objects in HEAP
    for (int inx = 0; inx < 9; ++inx)
    {
        nums.push_back( new Integer(data[inx]) );
    }

    // print the vector<Integer*> to stdout
    printIntegerStarVector( "TEST2: ORIG nums before UNIQUE", nums );

    // arg3 is a functor
#if 1
    // corrected version which fixes SEGMENTATION FAULT and all memory leaks reported by valgrind(1)
    // I THINK we want to use new C++11 cbegin() and cend(),since the equal_to predicate is passed "Integer *&"

//  DID NOT COMPILE
//OUT   IntegerStarVectorIterator dupPosition = ptgi::unique( const_cast<ConstIntegerStarVectorIterator>(nums.begin()), const_cast<ConstIntegerStarVectorIterator>(nums.end()), IntegerEqualByTenPointer() );

    // DID NOT COMPILE when equal_to predicate declared "Integer*& arg1, Integer*&  arg2"
//OUT   IntegerStarVectorIterator dupPosition = ptgi::unique( const_cast<nums::const_iterator>(nums.begin()), const_cast<nums::const_iterator>(nums.end()), IntegerEqualByTenPointer() );


    // okay when equal_to predicate declared "Integer* arg1, Integer*  arg2"
    IntegerStarVectorIterator dupPosition = ptgi::unique(nums.begin(), nums.end(), IntegerEqualByTenPointer() );
#else
    // BUGGY version that causes SEGMENTATION FAULT and valgrind(1) errors
    IntegerStarVectorIterator dupPosition = std::unique( nums.begin(), nums.end(), IntegerEqualByTenPointer() );
#endif

    printIntegerStarVector( "TEST2: modified nums AFTER UNIQUE", nums );
    int dupInx = dupPosition - nums.begin();
    std::cout << "INFO: dupInx=" << dupInx <<"\n";

    // delete the dup Integer* objects in the [dupPosition, end] range
    for (IntegerStarVectorIterator iter = dupPosition; iter != nums.end(); ++iter)
    {
        delete (*iter);
    }

    // shrink the vector
    // NB: the Integer* ptrs are NOT followed by vector::erase()
    nums.erase(dupPosition, nums.end());


    // print the uniques, by following the iter to the Integer* pointer
    for (IntegerStarVectorIterator iter = nums.begin(); iter != nums.end();  ++iter)
    {
        std::cout << "TEST2: uniq = " << (*iter)->getNum() << "\n";
    }

    // remove the unique objects from heap
    for (IntegerStarVectorIterator iter = nums.begin(); iter != nums.end();  ++iter)
    {
        delete (*iter);
    }

    // shrink the vector
    nums.erase(nums.begin(), nums.end());

    // the vector should now be completely empty
    assert( nums.size() == 0);
}

//@ print to stdout the string: "info_msg: num1, num2, .... numN\n"
void printIntegerStarVector( const std::string& msg, const std::vector<Integer*>& nums )
{
    std::cout << msg << ": ";
    int inx = 0;
    ConstIntegerStarVectorIterator  iter;

    // use const iterator and const range!
    // NB: cbegin() and cend() not supported until LATER (c++11)
    for (iter = nums.begin(), inx = 0; iter != nums.end(); ++iter, ++inx)
    {
        // output a comma seperator *AFTER* first
        if (inx > 0)
            std::cout << ", ";

        // call Integer::toString()
        std::cout << (*iter)->getNum();     // send int to stdout
//      std::cout << (*iter)->toString();   // also works, but is probably slower

    }

    // in conclusion, add newline
    std::cout << "\n";
}

Answer 20

std::set<int> s;
std::for_each(v.cbegin(), v.cend(), [&s](int val){s.insert(val);});
v.clear();
std::copy(s.cbegin(), s.cend(), v.cbegin());

Answer 21

Here's the example of the duplicate delete problem that occurs with std::unique(). On a LINUX machine, the program crashes. Read the comments for details.

// Main10.cpp
//
// Illustration of duplicate delete and memory leak in a vector<int*> after calling std::unique.
// On a LINUX machine, it crashes the progam because of the duplicate delete.
//
// INPUT : {1, 2, 2, 3}
// OUTPUT: {1, 2, 3, 3}
//
// The two 3's are actually pointers to the same 3 integer in the HEAP, which is BAD
// because if you delete both int* pointers, you are deleting the same memory
// location twice.
//
//
// Never mind the fact that we ignore the "dupPosition" returned by std::unique(),
// but in any sensible program that "cleans up after istelf" you want to call deletex
// on all int* poitners to avoid memory leaks.
//
//
// NOW IF you replace std::unique() with ptgi::unique(), all of the the problems disappear.
// Why? Because ptgi:unique merely reshuffles the data:
// OUTPUT: {1, 2, 3, 2}
// The ptgi:unique has swapped the last two elements, so all of the original elements in
// the INPUT are STILL in the OUTPUT.
//
// 130215   dbednar@ptgi.com
//============================================================================

#include <iostream>
#include <vector>
#include <algorithm>
#include <functional>

#include "ptgi_unique.hpp"

// functor used by std::unique to remove adjacent elts from vector<int*>
struct EqualToVectorOfIntegerStar: public std::equal_to<int *>
{
    bool operator() (const int* arg1, const int* arg2) const
    {
        return (*arg1 == *arg2);
    }
};

void printVector( const std::string& msg, const std::vector<int*>& vnums);

int main()
{
    int inums [] = { 1, 2, 2, 3 };
    std::vector<int*> vnums;

    // convert C array into vector of pointers to integers
    for (size_t inx = 0; inx < 4; ++ inx)
        vnums.push_back( new int(inums[inx]) );

    printVector("BEFORE UNIQ", vnums);

    // INPUT : 1, 2A, 2B, 3
    std::unique( vnums.begin(), vnums.end(), EqualToVectorOfIntegerStar() );
    // OUTPUT: 1, 2A, 3, 3 }
    printVector("AFTER  UNIQ", vnums);

    // now we delete 3 twice, and we have a memory leak because 2B is not deleted.
    for (size_t inx = 0; inx < vnums.size(); ++inx)
    {
        delete(vnums[inx]);
    }
}

// print a line of the form "msg: 1,2,3,..,5,6,7\n", where 1..7 are the numbers in vnums vector
// PS: you may pass "hello world" (const char *) because of implicit (automatic) conversion
// from "const char *" to std::string conversion.

void printVector( const std::string& msg, const std::vector<int*>& vnums)
{
    std::cout << msg << ": ";

    for (size_t inx = 0; inx < vnums.size(); ++inx)
    {
        // insert comma separator before current elt, but ONLY after first elt
        if (inx > 0)
            std::cout << ",";
        std::cout << *vnums[inx];

    }
    std::cout << "\n";
}

Answer 22

As already stated, unique requires a sorted container. Additionally, unique doesn't actually remove elements from the container. Instead, they are copied to the end, unique returns an iterator pointing to the first such duplicate element, and you are expected to call erase to actually remove the elements.

Answer 23

Here's a template to do it for you:

template<typename T>
void removeDuplicates(std::vector<T>& vec)
{
    std::sort(vec.begin(), vec.end());
    vec.erase(std::unique(vec.begin(), vec.end()), vec.end());
}

call it like:

removeDuplicates<int>(vectorname);

Answer 24

Efficiency is a complicated concept. There's time vs. space considerations, as well as general measurements (where you only get vague answers such as O(n)) vs. specific ones (e.g. bubble sort can be much faster than quicksort, depending on input characteristics).

If you have relatively few duplicates, then sort followed by unique and erase seems the way to go. If you had relatively many duplicates, creating a set from the vector and letting it do the heavy lifting could easily beat it.

Don't just concentrate on time efficiency either. Sort+unique+erase operates in O(1) space, while the set construction operates in O(n) space. And neither directly lends itself to a map-reduce parallelization (for really huge datasets).

Answer 25

I'm not sure what you are using this for, so I can't say this with 100% certainty, but normally when I think "sorted, unique" container, I think of a std::set. It might be a better fit for your usecase:

std::set<Foo> foos(vec.begin(), vec.end()); // both sorted & unique already

Otherwise, sorting prior to calling unique (as the other answers pointed out) is the way to go.

Answer 26

unique only removes consecutive duplicate elements (which is necessary for it to run in linear time), so you should perform the sort first. It will remain sorted after the call to unique.

Answer 27

You need to sort it before you call unique because unique only removes duplicates that are next to each other.

edit: 38 seconds...

Answer 28

std::unique only works on consecutive runs of duplicate elements, so you'd better sort first. However, it is stable, so your vector will remain sorted.