Combination of a Collection with Repetitions - c++

There are a lot of links on http://stackoverflow.com for how to do combinations: Generating combinations in c++ But these links presume to draw from an infinite set without repetition. When given a finite collection which does have repetition, these algorithms construct duplicates. For example you can see the accepted solution to the linked question failing on a test case I constructed here: http://ideone.com/M7CyFc
Given the input set: vector<int> row {40, 40, 40, 50, 50, 60, 100};
I expect to see:
40 40 40
40 40 50
40 40 60
40 40 100
40 50 50
40 50 60
40 50 100
40 60 100
50 50 60
50 50 100
50 60 100
Obviously I can use the old method store the output and check for duplicates as I generate, but this requires a lot of extra memory and processing power. Is there an alternative that C++ provides me?

Combinations by definition do not respect order. This frees us to arrange the numbers in any order we see fit. Most notably we can rely on to provide a combination rank. Certainly the most logical way to rank combinations is in sorted order, so we'll be depending upon our inputs being sorted.
There is already precedent for this in the standard library. For example lower_bound which we will actually use in this solution. When used generally this may however require the user to sort before passing.
The function we will write to do this will take in iterators to the sorted collection which the next combination is to be drawn from, and iterators to the current combination. We'd also need the size but that can be derived from the distance between the combination's iterators.
template <typename InputIt, typename OutputIt>
bool next_combination(InputIt inFirst, InputIt inLast, OutputIt outFirst, OutputIt outLast) {
assert(distance(inFirst, inLast) >= distance(outFirst, outLast));
const auto front = make_reverse_iterator(outFirst);
const auto back = make_reverse_iterator(outLast);
auto it = mismatch(back, front, make_reverse_iterator(inLast)).first;
const auto result = it != front;
if (result) {
auto ub = upper_bound(inFirst, inLast, *it);
copy(ub, next(ub, distance(back, it) + 1), next(it).base());
}
return result;
}
This function is written in the format of the other algorithm functions, so any container that supports bidirectional iterators can be used with it. For our example though we'll use: const vector<unsigned int> row{ 40U, 40U, 40U, 50U, 50U, 60U, 100U }; which is, necessarily, sorted:
vector<unsigned int> it{ row.cbegin(), next(row.cbegin(), 3) };
do {
copy(it.cbegin(), it.cend(), ostream_iterator<unsigned int>(cout, " "));
cout << endl;
} while(next_combination(row.cbegin(), row.cend(), it.begin(), it.end()));
Live Example
After writing this answer I've done a bit more research and found N2639 which proposes a standardized next_combination, which was:
Actively under consideration for a future TR, when work on TR2 was deferred pending
Viewed positively at the time
Due at least one more revision before any adoption
Needed some reworking to reflect the addition of C++11 language facilities
[Source]
Using N2639's reference implementation requires mutability, so we'll use: vector<unsigned int> row{ 40U, 40U, 40U, 50U, 50U, 60U, 100U };. And our example code becomes:
vector<unsigned int>::iterator it = next(row.begin(), 3);
do {
copy(row.begin(), it, ostream_iterator<unsigned int>(cout, " "));
cout << endl;
} while(next_combination(row.begin(), it, row.end()));
Live Example

You can do something like this (maybe avoiding the recursion):
#include <iostream>
#include <vector>
#include <algorithm>
using std::cout;
using std::vector;
void perm( const vector<int> &v, vector<vector<int>> &p, vector<int> &t, int k, int d) {
for ( int i = k; i < v.size(); ++i ) {
// skip the repeted value
if ( i != k && v[i] == v[i-1]) continue;
t.push_back(v[i]);
if ( d > 0 ) perm(v,p,t,i+1,d-1);
else p.push_back(t);
t.pop_back();
}
}
int main() {
int r = 3;
vector<int> row {40, 40, 40, 50, 50, 60, 100};
vector<vector<int>> pp;
vector<int> pe;
std::sort(row.begin(),row.end()); // that's necessary
perm(row,pp,pe,0,r-1);
cout << pp.size() << '\n';
for ( auto & v : pp ) {
for ( int i : v ) {
cout << ' ' << i;
}
cout << '\n';
}
return 0;
}
Which outputs:
11
40 40 40
40 40 50
40 40 60
40 40 100
40 50 50
40 50 60
40 50 100
40 60 100
50 50 60
50 50 100
50 60 100
I know, it's far from efficient, but if you get the idea you may come out with a better implementation.

Here is a class I once wrote in my university times to handle bosons. It's quite long, but it's generally usable and seems to work well. Additionally, it also gives ranking and unranking functionality. Hope that helps -- but don't ever ask me what I was doing back then ... ;-)
struct SymmetricIndex
{
using StateType = std::vector<int>;
using IntegerType = int;
int M;
int N;
StateType Nmax;
StateType Nmin;
IntegerType _size;
std::vector<IntegerType> store;
StateType state;
IntegerType _index;
SymmetricIndex() = default;
SymmetricIndex(int _M, int _N, int _Nmax = std::numeric_limits<int>::max(), int _Nmin = 0)
: SymmetricIndex(_M, _N, std::vector<int>(_M + 1, std::min(_Nmax, _N)), StateType(_M + 1, std::max(_Nmin, 0)))
{}
SymmetricIndex(int _M, int _N, StateType const& _Nmax, StateType const& _Nmin)
: N(_N)
, M(_M)
, Nmax(_Nmax)
, Nmin(_Nmin)
, store(addressArray())
, state(M)
, _index(0)
{
reset();
_size = W(M, N);
}
friend std::ostream& operator<<(std::ostream& os, SymmetricIndex const& sym);
SymmetricIndex& reset()
{
return setBegin();
}
bool setBegin(StateType& state, StateType const& Nmax, StateType const& Nmin) const
{
int n = N;
for (int i = 0; i<M; ++i)
{
state[i] = Nmin[i];
n -= Nmin[i];
}
for (int i = 0; i<M; ++i)
{
state[i] = std::min(n + Nmin[i], Nmax[i]);
n -= Nmax[i] - Nmin[i];
if (n <= 0)
break;
}
return true;
}
SymmetricIndex& setBegin()
{
setBegin(state, Nmax, Nmin);
_index = 0;
return *this;
}
bool isBegin() const
{
return _index==0;
}
bool setEnd(StateType& state, StateType const& Nmax, StateType const& Nmin) const
{
int n = N;
for (int i = 0; i < M; ++i)
{
state[i] = Nmin[i];
n -= Nmin[i];
}
for (int i = M - 1; i >= 0; --i)
{
state[i] = std::min(n + Nmin[i], Nmax[i]);
n -= Nmax[i] - Nmin[i];
if (n <= 0)
break;
}
return true;
}
SymmetricIndex& setEnd()
{
setEnd(state, Nmax, Nmin);
_index = _size - 1;
return *this;
}
bool isEnd() const
{
return _index == _size-1;
}
IntegerType index() const
{
return _index;
}
IntegerType rank(StateType const& state) const
{
IntegerType ret = 0;
int n = 0;
for (int i = 0; i < M; ++i)
{
n += state[i];
for (int k = Nmin[i]; k < state[i]; ++k)
ret += store[(n - k) * M + i];
}
return ret;
}
IntegerType rank() const
{
return rank(state);
}
StateType unrank(IntegerType rank) const
{
StateType ret(M);
int n = N;
for (int i = M-1; i >= 0; --i)
{
int ad = 0;
int k = std::min(Nmax[i] - 1, n);
for (int j = Nmin[i]; j <= k; ++j)
ad+=store[(n - j) * M + i];
while (ad > rank && k >= Nmin[i])
{
ad -= store[(n - k) * M + i];
--k;
}
rank -= ad;
ret[i] = k+1;
n -= ret[i];
if (n <= 0)
{
return ret;
}
}
return ret;
}
IntegerType size() const
{
return _size;
}
operator StateType& ()
{
return state;
}
auto operator[](int i) -> StateType::value_type& { return state[i]; }
operator StateType const& () const
{
return state;
}
auto operator[](int i) const -> StateType::value_type const& { return state[i]; }
bool nextState(StateType& state, StateType const& Nmax, StateType const& Nmin) const
{
//co-lexicographical ordering with Nmin and Nmax:
// (1) find first position which can be decreased
// then we have state[k] = Nmin[k] for k in [0,pos]
int pos = M - 1;
for (int k = 0; k < M - 1; ++k)
{
if (state[k] > Nmin[k])
{
pos = k;
break;
}
}
// if nothing found to decrease, return
if (pos == M - 1)
{
return false;
}
// (2) find first position after pos which can be increased
// then we have state[k] = Nmin[k] for k in [0,pos]
int next = 0;
for (int k = pos + 1; k < M; ++k)
{
if (state[k] < Nmax[k])
{
next = k;
break;
}
}
if (next == 0)
{
return false;
}
--state[pos];
++state[next];
// (3) get occupation in [pos,next-1] and set to Nmin[k]
int n = 0;
for (int k = pos; k < next; ++k)
{
n += state[k] - Nmin[k];
state[k] = Nmin[k];
}
// (4) fill up from the start
for (int i = 0; i<M; ++i)
{
if (n <= 0)
break;
int add = std::min(n, Nmax[i] - state[i]);
state[i] += add;
n -= add;
}
return true;
}
SymmetricIndex& operator++()
{
bool inc = nextState(state, Nmax, Nmin);
if (inc) ++_index;
return *this;
}
SymmetricIndex operator++(int)
{
auto ret = *this;
this->operator++();
return ret;
}
bool previousState(StateType& state, StateType const& Nmax, StateType const& Nmin) const
{
////co-lexicographical ordering with Nmin and Nmax:
// (1) find first position which can be increased
// then we have state[k] = Nmax[k] for k in [0,pos-1]
int pos = M - 1;
for (int k = 0; k < M - 1; ++k)
{
if (state[k] < Nmax[k])
{
pos = k;
break;
}
}
// if nothing found to increase, return
if (pos == M - 1)
{
return false;
}
// (2) find first position after pos which can be decreased
// then we have state[k] = Nmin[k] for k in [pos+1,next]
int next = 0;
for (int k = pos + 1; k < M; ++k)
{
if (state[k] > Nmin[k])
{
next = k;
break;
}
}
if (next == 0)
{
return false;
}
++state[pos];
--state[next];
int n = 0;
for (int k = 0; k <= pos; ++k)
{
n += state[k] - Nmin[k];
state[k] = Nmin[k];
}
if (n == 0)
{
return true;
}
for (int i = next-1; i>=0; --i)
{
int add = std::min(n, Nmax[i] - state[i]);
state[i] += add;
n -= add;
if (n <= 0)
break;
}
return true;
}
SymmetricIndex operator--()
{
bool dec = previousState(state, Nmax, Nmin);
if (dec) --_index;
return *this;
}
SymmetricIndex operator--(int)
{
auto ret = *this;
this->operator--();
return ret;
}
int multinomial() const
{
auto v = const_cast<std::remove_reference<decltype(state)>::type&>(state);
return multinomial(v);
}
int multinomial(StateType& state) const
{
int ret = 1;
int n = state[0];
for (int i = 1; i < M; ++i)
{
n += state[i];
ret *= binomial(n, state[i]);
}
return ret;
}
SymmetricIndex& random(StateType const& _Nmin)
{
static std::mt19937 rng;
state = _Nmin;
int n = std::accumulate(std::begin(state), std::end(state), 0);
auto weight = [&](int i) { return state[i] < Nmax[i] ? 1 : 0; };
for (int i = n; i < N; ++i)
{
std::discrete_distribution<int> d(N, 0, N, weight);
++state[d(rng)];
}
_index = rank();
return *this;
}
SymmetricIndex& random()
{
return random(Nmin);
}
private:
IntegerType W(int m, int n) const
{
if (m < 0 || n < 0) return 0;
else if (m == 0 && n == 0) return 1;
else if (m == 0 && n > 0) return 0;
//else if (m > 0 && n < Nmin[m-1]) return 0;
else
{
//static std::map<std::tuple<int, int>, IntegerType> memo;
//auto it = memo.find(std::make_tuple(k, m));
//if (it != std::end(memo))
//{
// return it->second;
//}
IntegerType ret = 0;
for (int i = Nmin[m-1]; i <= std::min(Nmax[m-1], n); ++i)
ret += W(m - 1, n - i);
//memo[std::make_tuple(k, m)] = ret;
return ret;
}
}
IntegerType binomial(int m, int n) const
{
static std::vector<int> store;
if (store.empty())
{
std::function<IntegerType(int, int)> bin = [](int n, int k)
{
int res = 1;
if (k > n - k)
k = n - k;
for (int i = 0; i < k; ++i)
{
res *= (n - i);
res /= (i + 1);
}
return res;
};
store.resize(M*M);
for (int i = 0; i < M; ++i)
{
for (int j = 0; j < M; ++j)
{
store[i*M + j] = bin(i, j);
}
}
}
return store[m*M + n];
}
auto addressArray() const -> std::vector<int>
{
std::vector<int> ret((N + 1) * M);
for (int n = 0; n <= N; ++n)
{
for (int m = 0; m < M; ++m)
{
ret[n*M + m] = W(m, n);
}
}
return ret;
}
};
std::ostream& operator<<(std::ostream& os, SymmetricIndex const& sym)
{
for (auto const& i : sym.state)
{
os << i << " ";
}
return os;
}
Use it like
int main()
{
int M=4;
int N=3;
std::vector<int> Nmax(M, N);
std::vector<int> Nmin(M, 0);
Nmax[0]=3;
Nmax[1]=2;
Nmax[2]=1;
Nmax[3]=1;
SymmetricIndex sym(M, N, Nmax, Nmin);
while(!sym.isEnd())
{
std::cout<<sym<<" "<<sym.rank()<<std::endl;
++sym;
}
std::cout<<sym<<" "<<sym.rank()<<std::endl;
}
This will output
3 0 0 0 0 (corresponds to {40,40,40})
2 1 0 0 1 (-> {40,40,50})
1 2 0 0 2 (-> {40,50,50})
2 0 1 0 3 ...
1 1 1 0 4
0 2 1 0 5
2 0 0 1 6
1 1 0 1 7
0 2 0 1 8
1 0 1 1 9
0 1 1 1 10 (-> {50,60,100})
DEMO
Note that I assumed here an ascending mapping of your set elements (i.e. the number 40's is given by index 0, the number of 50's by index 1, and so on).
More precisely: Turn your list into a map<std::vector<int>, int> like
std::vector<int> v{40,40,40,50,50,60,100};
std::map<int, int> m;
for(auto i : v)
{
++m[i];
}
Then use
int N = 3;
int M = m.size();
std::vector<int> Nmin(M,0);
std::vector<int> Nmax;
std::vector<int> val;
for(auto i : m)
{
Nmax.push_back(m.second);
val.push_back(m.first);
}
SymmetricIndex sym(M, N, Nmax, Nmin);
as input to the SymmetricIndex class.
To print the output, use
while(!sym.isEnd())
{
for(int i=0; i<M; ++i)
{
for(int j = 0; j<sym[i]; ++j)
{
std::cout<<val[i]<<" ";
}
}
std::cout<<std::endl;
}
for(int i=0; i<M; ++i)
{
for(int j = 0; j<sym[i]; ++j)
{
std::cout<<val[i]<<" ";
}
}
std::cout<<std::endl;
All untested, but it should give the idea.

Related

compact form of many for loop in C++

I have a piece of code as follows, and the number of for loops is determined by n which is known at compile time. Each for loop iterates over the values 0 and 1. Currently, my code looks something like this
for(int in=0;in<2;in++){
for(int in_1=0;in_1<2;in_1++){
for(int in_2=0;in_2<2;in_2++){
// ... n times
for(int i2=0;i2<2;i2++){
for(int i1=0;i1<2;i1++){
d[in][in_1][in_2]...[i2][i1] =updown(in)+updown(in_1)+...+updown(i1);
}
}
// ...
}
}
}
Now my question is whether one can write it in a more compact form.
The n bits in_k can be interpreted as the representation of one integer less than 2^n.
This allows easily to work with a 1-D array (vector) d[.].
In practice, an interger j corresponds to
j = in[0] + 2*in[1] + ... + 2^n-1*in[n-1]
Moreover, a direct implementation is O(NlogN). (N = 2^n)
A recursive solution is possible, for example using
f(val, n) = updown(val%2) + f(val/2, n-1) and f(val, 0) = 0.
This would correspond to a O(N) complexity, at the condition to introduce memoization, not implemented here.
Result:
0 : 0
1 : 1
2 : 1
3 : 2
4 : 1
5 : 2
6 : 2
7 : 3
8 : 1
9 : 2
10 : 2
11 : 3
12 : 2
13 : 3
14 : 3
15 : 4
#include <iostream>
#include <vector>
int up_down (int b) {
if (b) return 1;
return 0;
}
int f(int val, int n) {
if (n < 0) return 0;
return up_down (val%2) + f(val/2, n-1);
}
int main() {
const int n = 4;
int size = 1;
for (int i = 0; i < n; ++i) size *= 2;
std::vector<int> d(size, 0);
for (int i = 0; i < size; ++i) {
d[i] = f(i, n);
}
for (int i = 0; i < size; ++i) {
std::cout << i << " : " << d[i] << '\n';
}
return 0;
}
As mentioned above, the recursive approach allows a O(N) complexity, at the condition to implement memoization.
Another possibility is to use a simple iterative approach, in order to get this O(N) complexity.
(here N represents to total number of data)
#include <iostream>
#include <vector>
int up_down (int b) {
if (b) return 1;
return 0;
}
int main() {
const int n = 4;
int size = 1;
for (int i = 0; i < n; ++i) size *= 2;
std::vector<int> d(size, 0);
int size_block = 1;
for (int i = 0; i < n; ++i) {
for (int j = size_block-1; j >= 0; --j) {
d[2*j+1] = d[j] + up_down(1);
d[2*j] = d[j] + up_down(0);
}
size_block *= 2;
}
for (int i = 0; i < size; ++i) {
std::cout << i << " : " << d[i] << '\n';
}
return 0;
}
You can refactor your code slightly like this:
for(int in=0;in<2;in++) {
auto& dn = d[in];
auto updown_n = updown(in);
for(int in_1=0;in_1<2;in_1++) {
// dn_1 == d[in][in_1]
auto& dn_1 = dn[in_1];
// updown_n_1 == updown(in)+updown(in_1)
auto updown_n_1 = updown_n + updown(in_1);
for(int in_2=0;in_2<2;in_2++) {
// dn_2 == d[in][in_1][in_2]
auto& dn_2 = dn_1[in_2];
// updown_n_2 == updown(in)+updown(in_1)+updown(in_2)
auto updown_n_2 = updown_n_1 + updown(in_2);
.
.
.
for(int i2=0;i2<2;i1++) {
// d2 == d[in][in_1][in_2]...[i2]
auto& d2 = d3[i2];
// updown_2 = updown(in)+updown(in_1)+updown(in_2)+...+updown(i2)
auto updown_2 = updown_3 + updown(i2);
for(int i1=0;i1<2;i1++) {
// d1 == d[in][in_1][in_2]...[i2][i1]
auto& d1 = d2[i1];
// updown_1 = updown(in)+updown(in_1)+updown(in_2)+...+updown(i2)+updown(i1)
auto updown_1 = updown_2 + updown(i1);
// d[in][in_1][in_2]...[i2][i1] = updown(in)+updown(in_1)+...+updown(i1);
d1 = updown_1;
}
}
}
}
}
And make this into a recursive function now:
template<std::size_t N, typename T>
void loop(T& d) {
for (int i = 0; i < 2; ++i) {
loop<N-1>(d[i], updown(i));
}
}
template<std::size_t N, typename T, typename U>
typename std::enable_if<N != 0>::type loop(T& d, U updown_result) {
for (int i = 0; i < 2; ++i) {
loop<N-1>(d[i], updown_result + updown(i));
}
}
template<std::size_t N, typename T, typename U>
typename std::enable_if<N == 0>::type loop(T& d, U updown_result) {
d = updown_result;
}
If your type is int d[2][2][2]...[2][2]; or int*****... d;, you can also stop when the type isn't an array or pointer instead of manually specifying N (or change for whatever the type of d[0][0][0]...[0][0] is)
Here's a version that does that with a recursive lambda:
auto loop = [](auto& self, auto& d, auto updown_result) -> void {
using d_t = typename std::remove_cv<typename std::remove_reference<decltype(d)>::type>::type;
if constexpr (!std::is_array<d_t>::value && !std::is_pointer<d_t>::value) {
// Last level of nesting
d = updown_result;
} else {
for (int i = 0; i < 2; ++i) {
self(self, d[i], updown_result + updown(i));
}
}
};
for (int i = 0; i < 2; ++i) {
loop(loop, d[i], updown(i));
}
I am assuming that it is a multi-dimensional matrix. You may have to solve it mathematically first and then write the respective equations in the program.

Finding unique entries of vector and deleting entries outside of interval [cMin, cMax]

I already managed to implement most of what I planned to do correctly, but somehow I struggle with the unique and cut method.
The unique method should sort the vector and delete all entries that appear more than once and the original vector should be overwritten with the shortened on. The cut method should delete all entries < cMin or > cMax.
Here is my try so far:
#include <cassert>
#include <iostream>
using std::cout;
using std::endl;
class Vector {
private:
int n;
double* coeff;
public:
Vector(int, double = 0);
~Vector();
Vector(const Vector&);
Vector& operator=(const Vector&);
int size() const;
double& operator()(int);
const double& operator()(int) const;
double max() const;
void sort();
void unique();
void cut(double Cmin, double Cmax);
void print() const;
};
Vector::Vector(int n, double init)
{
assert(n >= 0);
this->n = n;
if (n == 0) {
coeff = (double*)0;
}
else {
coeff = new double[n];
}
for (int i = 0; i < n; i++) {
coeff[i] = init;
}
}
Vector::Vector(const Vector& rhs)
{
n = rhs.n;
if (n > 0) {
coeff = new double[n];
}
else {
coeff = (double*)0;
}
for (int i = 0; i < n; i++) {
coeff[i] = rhs.coeff[i];
}
}
Vector::~Vector()
{
if (n > 0) {
delete[] coeff;
}
}
Vector& Vector::operator=(const Vector& rhs)
{
if (this != &rhs) {
if (n != rhs.n) {
if (n > 0) {
delete[] coeff;
}
n = rhs.n;
if (n > 0) {
coeff = new double[n];
}
else {
coeff = (double*)0;
}
}
for (int i = 0; i < n; i++) {
coeff[i] = rhs.coeff[i];
}
}
return *this;
}
int Vector::size() const
{
return n;
}
double& Vector::operator()(int j)
{
assert(j >= 1 && j <= n);
return coeff[j - 1];
}
const double& Vector::operator()(int j) const
{
assert(j >= 1 && j <= n);
return coeff[j - 1];
}
double Vector::max() const
{
double max = coeff[0];
for (int i = 1; i < n; i++) {
if (coeff[i] > max) {
max = coeff[i];
}
}
return max;
}
void Vector::sort()
{ //bubble-sort
double tmp = 0;
for (int i = 0; i < n - 1; i++) {
for (int j = 0; j < n - 1; j++) {
if (coeff[j] > coeff[j + 1]) {
tmp = coeff[j];
coeff[j] = coeff[j + 1];
coeff[j + 1] = tmp;
}
}
}
}
void Vector::unique()
{
sort();
int counter = 0;
Vector kopie = *this;
for (int i = 0; i < n; i++) {
if (i == 0 && coeff[i] != coeff[i + 1]) {
counter++;
}
if (i == n - 1 && coeff[i] != coeff[i - 1]) {
counter++;
}
if (i != 0 && i != n - 1 && coeff[i] != coeff[i - 1] && coeff[i] != coeff[i + 1]) {
counter++;
}
}
delete[] coeff;
coeff = new double[counter];
//to be continued...
}
void Vector::cut(double Cmin, double Cmax)
{
sort();
int counter = 0;
int j = 0;
Vector kopie = *this;
for (int i = 0; i < n; i++) {
if (coeff[i] >= Cmin && coeff[i] <= Cmax) {
counter++;
}
}
delete[] coeff;
coeff = new double[counter];
for (int i = 0; i < n; i++) {
if (kopie.coeff[i] >= Cmin && kopie.coeff[i] <= Cmax) {
coeff[j] = kopie.coeff[i];
j++;
if (j == n) {
break;
}
}
}
}
void Vector::print() const
{
for (int i = 0; i < n; i++) {
cout << coeff[i] << " ";
}
}
int main()
{
Vector X(8);
X.print();
cout << endl;
X(1) = 1.;
X(2) = 7.;
X(3) = 2.;
X(4) = 5.;
X(5) = 6.;
X(6) = 5.;
X(7) = 9.;
X(8) = 6.;
X.print();
cout << endl;
X.sort();
X.print();
cout << endl;
//X.unique();
//X.print();
//cout << endl;
X.cut(2, 6);
X.print();
cout << endl;
return 0;
}
For the unique function, rather than checking if it's legal to move the counter forward, I would just check if your current element and the next element aren't the same. If they are, set the next element's pointer to skip over the duplicate element.
Pseduocode:
For(int i = 0; i < n-1; i++) {
if(coef[i] == coef[i+1]) {
//Keep moving next element pointer until not equal. Probably use a while loop
}
}
The simplest solution that I can think of is something like this:
void Vector::unique()
{
size_t counter = 0;
double* copy = new double[n];
copy[counter++] = coeff[0]; // The first element is guaranteed to be unique
// Since coeff is sorted, copy will be sorted as well.
// Therefore, its enough to check only the last element of copy to
// the current element of coeff
for (size_t i = 1; i < n; i++)
{
if (coeff[i] != copy[counter])
{
copy[counter++] = coeff[i];
}
}
// copy already contains the data in the format that you want,
// but the reserved memory size may be larger than necessary.
// Reserve the correct amount of memory and copy the data there
delete[] coeff;
coeff = new double[counter];
std::memcpy(coeff, copy, counter*sizeof(double));
}
For cut() you can use a similar algorithm:
void Vector::cut(double Cmin, double Cmax)
{
size_t counter = 0;
double* copy = new double[n];
for (size_t i = 0; i < n; i++)
{
if (coeff[i] > Cmin && coeff[i] < Cmax)
{
copy[counter++] = coeff[i];
}
}
// Same story with memory size here as well
delete[] coeff;
coeff = new double[counter];
std::memcpy(coeff, copy, counter*sizeof(double));
}
Is there any reason why you can't use the standard library?
void Vector::unique() {
std::sort(coeff, std::next(coeff, n));
auto it = std::unique(coeff, std::next(coeff, n));
double* tmp = new double[n = std::distance(coeff, it)];
std::copy(coeff, it, tmp);
delete[] std::exchange(coeff, tmp);
}
void Vector::cut(double Cmin, double Cmax) {
auto it = std::remove_if(coeff, std::next(coeff, n),
[=] (double d) { return d < Cmin || d > Cmax; });
double* tmp = new double[n = std::distance(coeff, it)];
std::copy(coeff, it, tmp);
delete[] std::exchange(coeff, tmp);
}
To remove duplicates and sort, you can achieve it in three ways
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() );
All three has different performance. You can use one depending upon your size and number of duplicates present.
To cut you can use algorithm library
std::remove, std::remove_if
Syntax
template< class ForwardIt, class T >
constexpr ForwardIt remove( ForwardIt first, ForwardIt last, const T& value );
Possible implementation
First version
template< class ForwardIt, class T > ForwardIt remove(ForwardIt first, ForwardIt last, const T& value)
{ first = std::find(first, last, value);
if (first != last)
for(ForwardIt i = first; ++i != last; )
if (!(*i == value))
*first++ = std::move(*i); return first; }
Second version
template<class ForwardIt, class UnaryPredicate> ForwardIt remove_if(ForwardIt first, ForwardIt last, UnaryPredicate p)
{ first = std::find_if(first, last, p);
if (first != last)
for(ForwardIt i = first; ++i != last; )
if (!p(*i)) *first++ = std::move(*i);
return first; }
Examples
The following code removes all spaces from a string by shifting all non-space characters to the left and then erasing the extra. This is an example of erase-remove idiom.
Run this code
#include <algorithm>
#include <string>
#include <iostream>
#include <cctype>  
int main()
{ std::string str1 = "Text with some spaces"; str1.erase(std::remove(str1.begin(), str1.end(), ' '), str1.end());
std::cout << str1 << '\n';  
std::string str2 = "Text\n with\tsome \t whitespaces\n\n"; str2.erase(std::remove_if(str2.begin(), str2.end(), [](unsigned char x)
{return std::isspace(x);}), str2.end());
std::cout << str2 << '\n'; }
Output:
Textwithsomespaces
Textwithsomewhitespaces

c++ Karatsuba Multiplication using Vectors

So i've been trying to write out an algorithm for the Karatsuba Multiplication algorithm, and i've been attempting to use vectors as my data structure to handle the really long numbers which will be input...
My program can do smaller numbers fine, however it really struggles with larger numbers, and i get a core dump (Seg Fault). It also outputs strange results when the left hand side number is smaller than the right hand side.
Got any ideas? Heres the code.
#include <iostream>
#include <string>
#include <vector>
#define max(a,b) ((a) > (b) ? (a) : (b))
using namespace std;
vector<int> add(vector<int> lhs, vector<int> rhs) {
int length = max(lhs.size(), rhs.size());
int carry = 0;
int sum_col;
vector<int> result;
while(lhs.size() < length) {
lhs.insert(lhs.begin(), 0);
}
while(rhs.size() < length) {
rhs.insert(rhs.begin(), 0);
}
for(int i = length-1; i >= 0; i--) {
sum_col = lhs[i] + rhs[i] + carry;
carry = sum_col/10;
result.insert(result.begin(), (sum_col%10));
}
if(carry) {
result.insert(result.begin(), carry);
}
int x = 0;
while(result[x] == 0) {
x++;
}
result.erase(result.begin(), result.begin()+x);
return result;
}
vector<int> subtract(vector<int> lhs, vector<int> rhs) {
int length = max(lhs.size(), rhs.size());
int diff;
vector<int> result;
while(lhs.size() < length) {
lhs.insert(lhs.begin(), 0);
}
while(rhs.size() < length) {
rhs.insert(rhs.begin(), 0);
}
for(int i = length-1; i >= 0; i--) {
diff = lhs[i] - rhs[i];
if(diff >= 0) {
result.insert(result.begin(), diff);
} else {
int j = i - 1;
while(j >= 0) {
lhs[j] = (lhs[j] - 1) % 10;
if(lhs[j] != 9) {
break;
} else {
j--;
}
}
result.insert(result.begin(), diff+10);
}
}
int x = 0;
while(result[x] == 0) {
x++;
}
result.erase(result.begin(), result.begin()+x);
return result;
}
vector<int> multiply(vector<int> lhs, vector<int> rhs) {
int length = max(lhs.size(), rhs.size());
vector<int> result;
while(lhs.size() < length) {
lhs.insert(lhs.begin(), 0);
}
while(rhs.size() < length) {
rhs.insert(rhs.begin(), 0);
}
if(length == 1) {
int res = lhs[0]*rhs[0];
if(res >= 10) {
result.push_back(res/10);
result.push_back(res%10);
return result;
} else {
result.push_back(res);
return result;
}
}
vector<int>::const_iterator first0 = lhs.begin();
vector<int>::const_iterator last0 = lhs.begin() + (length/2);
vector<int> lhs0(first0, last0);
vector<int>::const_iterator first1 = lhs.begin() + (length/2);
vector<int>::const_iterator last1 = lhs.begin() + ((length/2) + (length-length/2));
vector<int> lhs1(first1, last1);
vector<int>::const_iterator first2 = rhs.begin();
vector<int>::const_iterator last2 = rhs.begin() + (length/2);
vector<int> rhs0(first2, last2);
vector<int>::const_iterator first3 = rhs.begin() + (length/2);
vector<int>::const_iterator last3 = rhs.begin() + ((length/2) + (length-length/2));
vector<int> rhs1(first3, last3);
vector<int> p0 = multiply(lhs0, rhs0);
vector<int> p1 = multiply(lhs1,rhs1);
vector<int> p2 = multiply(add(lhs0,lhs1),add(rhs0,rhs1));
vector<int> p3 = subtract(p2,add(p0,p1));
for(int i = 0; i < 2*(length-length/2); i++) {
p0.push_back(0);
}
for(int i = 0; i < (length-length/2); i++) {
p3.push_back(0);
}
result = add(add(p0,p1), p3);
int x = 0;
while(result[x] == 0) {
x++;
}
result.erase(result.begin(), result.begin()+x);
return result;
}
int main() {
vector<int> lhs;
vector<int> rhs;
vector<int> v;
lhs.push_back(2);
lhs.push_back(5);
lhs.push_back(2);
lhs.push_back(5);
lhs.push_back(2);
lhs.push_back(5);
lhs.push_back(2);
lhs.push_back(5);
rhs.push_back(1);
rhs.push_back(5);
rhs.push_back(1);
rhs.push_back(5);
rhs.push_back(1);
rhs.push_back(5);
rhs.push_back(1);
v = multiply(lhs, rhs);
for(size_t i = 0; i < v.size(); i++) {
cout << v[i];
}
cout << endl;
return 0;
}
There are several issues with subtract. Since you don't have any way to represent a negative number, if rhs is greater than lhs your borrow logic will access before the beginning of of the data for lhs.
You can also march past the end of result when removing leading zeros if the result is 0.
Your borrow calculation is wrong, since -1 % 10 will return -1, and not 9, if lhs[j] is 0. A better way to calculate that is add 9 (one less than the value you're dividing by), lhs[j] = (lhs[j] + 9) % 10;.
In an unrelated note, you can simplify your range iteration calculations. Since last0 and first1 have the same value, you can use last0 for both, and last1 is lhs.end(). This simpifies lhs1 to
vector<int> lhs1(last0, lhs.end());
and you can get rid of first1 and last1. Same goes for the rhs iterators.

Algorithm that builds heap

I am trying to implement build_max_heap function that creates the heap( as it is written in Cormen's "introduction do algorithms" ). But I am getting strange error and i could not localize it. My program successfully give random numbers to table, show them but after build_max_heap() I am getting strange numbers, that are probably because somewhere my program reached something out of the table, but I can not find this error. I will be glad for any help.
For example I get the table:
0 13 18 0 22 15 24 19 5 23
And my output is:
24 7 5844920 5 22 15 18 19 0 23
My code:
#include <iostream>
#include <ctime>
#include <stdlib.h>
const int n = 12; // the length of my table, i will onyl use indexes 1...n-1
struct heap
{
int *tab;
int heap_size;
};
void complete_with_random(heap &heap2)
{
srand(time(NULL));
for (int i = 1; i <= heap2.heap_size; i++)
{
heap2.tab[i] = rand() % 25;
}
heap2.tab[0] = 0;
}
void show(heap &heap2)
{
for (int i = 1; i < heap2.heap_size; i++)
{
std::cout << heap2.tab[i] << " ";
}
}
int parent(int i)
{
return i / 2;
}
int left(int i)
{
return 2 * i;
}
int right(int i)
{
return 2 * i + 1;
}
void max_heapify(heap &heap2, int i)
{
if (i >= heap2.heap_size || i == 0)
{
return;
}
int l = left(i);
int r = right(i);
int largest;
if (l <= heap2.heap_size || heap2.tab[l] > heap2.tab[i])
{
largest = l;
}
else
{
largest = i;
}
if (r <= heap2.heap_size || heap2.tab[r] > heap2.tab[i])
{
largest = r;
}
if (largest != i)
{
std::swap(heap2.tab[i], heap2.tab[largest]);
max_heapify(heap2, largest);
}
}
void build_max_heap(heap &heap2)
{
for (int i = heap2.heap_size / 2; i >= 1; i--)
{
max_heapify(heap2, i);
}
}
int main()
{
heap heap1;
heap1.tab = new int[n];
heap1.heap_size = n - 1;
complete_with_random(heap1);
show(heap1);
std::cout << std::endl;
build_max_heap(heap1);
show(heap1);
}
Indeed, the table is accessed with out-of-bounds indexes.
if (l <= heap2.heap_size || heap2.tab[l] > heap2.tab[i])
^^
I think you meant && in this condition.
The same for the next branch with r.
In case you're still having problems, below is my own implementation that you might use for reference. It was also based on Cormen et al. book, so it's using more or less the same terminology. It may have arbitrary types for the actual container, the comparison and the swap functions. It provides a public queue-like interface, including key incrementing.
Because it's part of a larger software collection, it's using a few entities that are not defined here, but I hope the algorithms are still clear. CHECK is only an assertion mechanism, you can ignore it. You may also ignore the swap member and just use std::swap.
Some parts of the code are using 1-based offsets, others 0-based, and conversion is necessary. The comments above each method indicate this.
template <
typename T,
typename ARRAY = array <T>,
typename COMP = fun::lt,
typename SWAP = fun::swap
>
class binary_heap_base
{
protected:
ARRAY a;
size_t heap_size;
SWAP swap_def;
SWAP* swap;
// 1-based
size_t parent(const size_t n) { return n / 2; }
size_t left (const size_t n) { return n * 2; }
size_t right (const size_t n) { return n * 2 + 1; }
// 1-based
void heapify(const size_t n = 1)
{
T& x = a[n - 1];
size_t l = left(n);
size_t r = right(n);
size_t select =
(l <= heap_size && COMP()(x, a[l - 1])) ?
l : n;
if (r <= heap_size && COMP()(a[select - 1], a[r - 1]))
select = r;
if (select != n)
{
(*swap)(x, a[select - 1]);
heapify(select);
}
}
// 1-based
void build()
{
heap_size = a.length();
for (size_t n = heap_size / 2; n > 0; n--)
heapify(n);
}
// 1-based
size_t advance(const size_t k)
{
size_t n = k;
while (n > 1)
{
size_t pn = parent(n);
T& p = a[pn - 1];
T& x = a[n - 1];
if (!COMP()(p, x)) break;
(*swap)(p, x);
n = pn;
}
return n;
}
public:
binary_heap_base() { init(); set_swap(); }
binary_heap_base(SWAP& s) { init(); set_swap(s); }
binary_heap_base(const ARRAY& a) { init(a); set_swap(); }
binary_heap_base(const ARRAY& a, SWAP& s) { init(a); set_swap(s); }
void init() { a.init(); build(); }
void init(const ARRAY& a) { this->a = a; build(); }
void set_swap() { swap = &swap_def; }
void set_swap(SWAP& s) { swap = &s; }
bool empty() { return heap_size == 0; }
size_t size() { return heap_size; }
size_t length() { return heap_size; }
void reserve(const size_t len) { a.reserve(len); }
const T& top()
{
CHECK (heap_size != 0, eshape());
return a[0];
}
T pop()
{
CHECK (heap_size != 0, eshape());
T x = a[0];
(*swap)(a[0], a[heap_size - 1]);
heap_size--;
heapify();
return x;
}
// 0-based
size_t up(size_t n, const T& x)
{
CHECK (n < heap_size, erange());
CHECK (!COMP()(x, a[n]), ecomp());
a[n] = x;
return advance(n + 1) - 1;
}
// 0-based
size_t push(const T& x)
{
if (heap_size == a.length())
a.push_back(x);
else
a[heap_size] = x;
return advance(++heap_size) - 1;
}
};

Iterating over subsets of any size

I can iterate over the subsets of size 1
for( int a = 0; a < size; a++ ) {
or subsets of size 2
for( int a1 = 0; a1 < size; a1++ ) {
for( int a2 = a1+1; a2 < size; a2++ ) {
or 3
for( int a1 = 0; a1 < size; a1++ ) {
for( int a2 = a1+1; a2 < size; a2++ ) {
for( int a3 = a2+1; a3 < size; a3++ ) {
But how to do this for subsets of size n?
This does the job, based on an answer by Adam Rosenfield
void iterate(int *a, int i, int size, int n)
{
int start = 0;
if( i > 0 ) start = a[i-1]+1;
for(a[i] = start; a[i] < n; a[i]++) {
if(i == n-1) {
// a is the array of indices of size n
for( int k = 0; k < size; k++ ) {
printf("%d ",a[k]);
}
printf("\n");
}
else
iterate(a, i+1, size, n);
}
}
You can use recursion:
void iterate(int *a, int i, int size, int n)
{
for(a[i] = 0; a[i] < size; a[i]++)
{
if(i == n-1)
DoStuff(a, n); // a is the array of indices of size n
else
iterate(a, i+1, size, n);
}
}
...
// Equivalent to 4 nested for loops
int a[4];
iterate(a, 0, size, 4);
You likely could do this with some recursion.
Here is something I used for a similar problem. It does not use recursion; rather, it uses a vector of indexes.
#include <vector>
template<class T>
class MultiForVar {
std::vector<T> _begin, _end, _vars;
inline int dim(){return _vars.size();}
public:
MultiForVar(std::vector<T> begin, std::vector<T> end) : _begin(begin), _end(end), _vars(_begin)
{
assert(begin.size() == end.size() and "Starting and ending vector<T> not the same size!" );
}
MultiForVar& operator ++()
{
++_vars[dim()-1];
for(int d = dim()-1; d > 0; --d)
{
if( _vars[d] >= _end[d] )
{
_vars[d] = _begin[d];
++_vars[d-1];
}
}
return *this;
}
bool done()
{
/*for(int d = 0; d < dim(); ++d)
if( _vars[d] < _end[d] )
return false;
return true;*/
return (_vars[0] >= _end[0]);
}
T operator[](int d)
{
return _vars.at(d);
}
int numDimensions(){
return dim();
}
std::vector<T>& getRaw(){
return _vars;
}
};
If I understand what you're asking correctly, another way to do it is to use bit-wise operators:
for(int i = 0; i < 1<<size; i++) {
for(int j = 0; j < size; j++) {
if(i & 1<<j) printf("%d ", a[j]);
}
printf("\n");
}
You need something the constructs the powerset of the original set. It's been a while since I've written that, but the psuedocode looks like
Powerset(a, size)
{
if(size == 0) return emptyset
subseta = Powerset(a, size-1) // Powerset of everything except last element
subsetb = appendToAll(a[size-1], subseta) // appends the last element to every set in subseta
return union(subseta, subsetb)
}