Code Review Stack Exchange is a question and answer site for peer programmer code reviews. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I did two methods to determine if an unordered_set is a subset of another unordered_set, using hashing, and the built-in includes().

I want to see if using a hash function will improve the run time, but I've noticed that there is no really big different in their time, so I start wondering, if what I did is the right idea of hashing!

Could you please look at my code to see if I did the hashing correctly?

#include <iostream> 
#include <unordered_set>
#include <unordered_map>
#include <iomanip>
#include <algorithm>
#include <functional>

struct Sstruct {
    int Iv;
    float Fv1;
    float Fv2;

bool operator==(Sstruct a, Sstruct b) { return a.Iv == b.Iv; }

struct MyHash {
    size_t operator()(const Sstruct& x) const { return std::hash<int>()(x.Iv); }

bool IsIn(const std::unordered_set<int>& S1, const std::unordered_map <int, std::unordered_set<int>>::const_iterator& iter)
    for (std::unordered_set<int>::const_iterator iter2 = iter->second.begin(); iter2 != iter->second.end(); ++iter2)

        if (S1.find(*iter2) == S1.end())
            return false;

    return true;


int main()

    std::unordered_set<Sstruct, MyHash> S1{ { 1, 0.9, 0.1 }, { 2, 0.8, 0.2 }, { 3, 0.9, 0.01 }, { 4, 0.7, 0.5 }, { 7, 0.7, 0.2 }, { 8, 0.1, 0.5 } };

    std::unordered_map <int, std::unordered_set<int>> S2;

    S2[0].insert({ { 1 }, { 2 }, { 3} }); //  S2 is a subset of S1
    S2[1].insert({ { 3}, { 4 }, { 5} }); // S2 is a subset of S1
    S2[3].insert({ { 2 }, { 4 } }); // S2 is  a subset of S1
    S2[4].insert({ { 3 }, { 4 }}); // S2 is  a subset of S1
    S2[5].insert({ { 3 },  { 5 } }); // S2 is a subset of S1
    S2[6].insert({ { 1 }, { 2 }, { 8 } }); //  S2 is a subset of S1
    S2[7].insert({ { 3 }, { 4 }, { 6 } }); // S2 is not a subset of S1
    S2[8].insert({ { 2 }, {7 } }); // S2 is  a subset of S1
    S2[9].insert({ { 7 }, { 8 } }); // S2 is  a subset of S1
    S2[10].insert({ { 5 }, { 6 } }); // S2 is not a subset of S1

    std::unordered_set<int > SetOne;

    for (std::unordered_set<Sstruct, MyHash> ::iterator iter = S1.begin(); iter != S1.end(); ++iter)

    float time1, time2;

    time1 = (float)clock() / CLOCKS_PER_SEC;

    // method 1: using hashing
    /*for (std::unordered_map <int, std::unordered_set<int>>::const_iterator SetTwo = S2.begin(); SetTwo != S2.end(); ++SetTwo)
        if ((IsIn(SetOne, SetTwo)) == true)
            std::cout << "S2 is a subset of S1";

            std::cout << "S2 is not a subset of S1";

        std::cout << "\n";

    // method 2 using STL includes()
    std::unordered_map <int, std::unordered_set<int>>::const_iterator SetTwo = S2.begin();
    while (SetTwo != S2.end())
        if (std::includes(SetOne.begin(), SetOne.end(), SetTwo->second.begin(), SetTwo->second.end()))

            std::cout << "S2 is a subset of S1";

            std::cout << "S2 is not a subset of S1";
        std::cout << "\n";

    time2 = (float)clock() / CLOCKS_PER_SEC;

    std::cout << "\n  Finishing Time " << time2 - time1 << " secs";
    std::cout << "\n\n";

share|improve this question
up vote 7 down vote accepted

There are several things that are problematic:

Broken code

The version using std::includes is actually broken. When you read up on std::includes you will find that it requires the ranges it is operating on to be ordered by some comparison function (defaults to operator<).

Now compare this with the source of the iterators which is a template call unordered_set.

You should not use std::includes on unordered_set because the element sequence of the latter does not satisfy the preconditions of the former, thus resulting in an invalid algorithm.

Meaningless benchmark

There are several problems with your benchmarking code (some of which have already been addressed):

  • executed only once: to reduce the impact of noise in the measurements the benchmark is usually done several times and then the average is calculated
  • too short measurement period: This is somewhat connected the previous one. If the runtime of an algorithm is very short (close to the time measurement granularity limit) there will be big steps for very little change so one would try to increase the runtime (again my doing multiple runs)
  • too little problem size: the effect of hashing is the advantage of \$O(1)\$ element access compared to \$O(log n)\$ (for trees) or even \$O(n)\$ for linear structures. However, these are only asymptotic complexities which become meaningful as n grows towards infinity. For the little problem sizes you are presenting here they hardly matter and are dominated by surrounding (constant complexity) code
  • you print results during measuring which will most likely account for most of the time you are measuring because printing is very slow.

C++ 11

If you've got C++11 then why not use the whole package?

Use range-based for to clean up messy iterator loops:

for (std::unordered_map <int, std::unordered_set<int>>::const_iterator SetTwo = S2.begin(); SetTwo != S2.end(); ++SetTwo)

can become

for (auto const &element: S2)

However notice that element is already the element pointed to by the dereferenced iterator.

Function Interface

Now that we got away from the iterator we should clean up the interface of IsIn. This function is far too specialized by taking an iterator. Instead take the type you want directly:

bool IsIn(const std::unordered_set<int>& haystack, std::unordered_set<int>& needles) {
    for (auto const& needle: needles)
        if(haystack.find(needle) == haystack.end()) 
            return false;
    return true;

I renamed the variables to make their intention more clearly (though they are not fitting for the current task). The function should probably be renamed too.

Unused Variable

What is the intention of variable S1? You initialize it and then just copy some of its values into SetOne. You could have just initialized SetOne and be done.

Wrong container type

Using an unordered_map for S2 is totally overkill. You have a continuous range of indices starting at 0 and it seems like you want to iterate this in exactly this order. You can use a vector for this and initialize it with the desired values:

std::vector<std::unordered_set<int>> S2{
    { { 1 }, { 2 }, { 3} }, //  S2 is a subset of S1
    { { 3}, { 4 }, { 5} }, // S2 is a subset of S1
    { { 2 }, { 4 } }, // S2 is  a subset of S1
    { { 3 }, { 4 }}, // S2 is  a subset of S1
    { { 3 },  { 5 } }, // S2 is a subset of S1
    { { 1 }, { 2 }, { 8 } }, //  S2 is a subset of S1
    { { 3 }, { 4 }, { 6 } }, // S2 is not a subset of S1
    { 2 }, {7 } }, // S2 is  a subset of S1
    { 7 }, { 8 } }, // S2 is  a subset of S1
    { 5 }, { 6 } }, // S2 is not a subset of S1

I would propose renaming S2 to testSets or some other name that indicates that there are multiple sets that should be tested. It may even be advisable to create two vectors, one with subsets and one that has no subsets.

Inadequate loop

In your second method you are using a while loop instead of a for loop. Benchmarkwise this might skew your results because you don't want to measure differences in the loops, so you should be using the same code there.

The for loop is clearly the better alternative here because you intend to loop over the whole container anyways.

share|improve this answer
Wow, It's precious information, I learned a lot here, thank you for sharing that, thanks for your time, thank you very much – phoenix Jun 16 '14 at 15:06

I just have one nitpick, not related to your primary request.

You don't need to declare these first:

float time1, time2;

Variables should be declared or initialized as close in scope as possible. This will ease maintenance as you can keep better track of variables, such as if any are no longer in use.

In this case, you can just initialize both variables right away:

float time1 = (float)clock() / CLOCKS_PER_SEC;

float time2 = (float)clock() / CLOCKS_PER_SEC;

In addition, you could use static_cast<>() over C-style casting:

static_cast<float>(clock()) / CLOCKS_PER_SEC;

See this post for more information on different types of casts.

share|improve this answer
Thank you so much, It is new information for me, I appropriate it – phoenix Jun 16 '14 at 15:02

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.