I have an assignment to read a file and output the average test scores.
It is pretty simple but I don't like how the average is done.
average = (test1 + test2 + test3 + test4 + test5) / 5.0;
Is there a way to just have it divide by the number of test scores? I couldn't find anything like this in the book or from google. Something like
average = (test + test + test + test) / ntests;
If you have the values in a vector or an array, just use std::accumulate from <numeric>:
std::vector<double> vec;
// ... fill vec with values (do not use 0; use 0.0)
double average = std::accumulate(vec.begin(), vec.end(), 0.0) / vec.size();
Step 1. Via iteration (if you want to be done) or recursion (if you want to be brave) place all test scores into an array (if you want simplicity and speed) or a linked list (if you want flexibility but slow)
Step 2. Iterate through the array/list until you reach the end; adding the contents of each cell/node as you go. Keep a count of what cell/node you are currently at as you go as well.
Step 3. Take the sum from the first variable and divide it by the second variable that kept track of where you were. This will yield the mean.
Wondering, why no one mentioned boost::accumulators. It is not the shortest of the already posted solutions, but can be more easily extended for more general statistical values. Like standard deviation or higher moments.
#include <iostream>
#include <boost/accumulators/accumulators.hpp>
#include <boost/accumulators/statistics/stats.hpp>
#include <boost/accumulators/statistics/mean.hpp>
#include <algorithm>
#include <vector>
double mean(const std::vector<double>& values) {
namespace bo = boost::accumulators;
if (values.empty()) return 0.;
bo::accumulator_set<double, bo::stats<bo::tag::mean>> acc;
acc=std::for_each(values.begin(), values.end(), acc);
return bo::mean(acc);
}
int main()
{
std::vector<double> test = { 2.,6.,4.,7. };
std::cout << "Mean: " << mean(test) << std::endl;
std::cout << "Mean: " << mean({}) << std::endl;
return 0;
}
Here is my generalization of getting the average of the elements of a container by specifying a lambda function to obtain each value and then add up:
template <typename ForwardIterator, typename F>
double inline averageOf (ForwardIterator first, ForwardIterator last, F function) {
std::vector<typename std::result_of<F(typename ForwardIterator::value_type)>::type> values;
while (first != last) {
values.emplace_back (function(*first));
++first;
}
return static_cast<double>(std::accumulate (values.begin(), values.end(), 0)) / values.size();
}
The client code I tested it with goes like
const std::list<CharmedObserver*> devotees =
charmer->getState<CharmerStateBase>(CHARMER)->getDevotees();
const int averageHitPointsOfDevotees = averageOf (devotees.begin(), devotees.end(),
[](const CharmedObserver* x)->int {return x->getCharmedBeing()->getHitPoints();});
C++11 gives nice solution:
constexpr auto countArguments() -> size_t
{
return 0;
}
template<class T1, class ... Ti>
constexpr auto countArguments(T1, Ti ...xi) -> size_t
{
return 1 + countArguments(xi...);
}
template<class T>
constexpr auto sumAruguments(T x) -> double
{
return x;
}
template<class T1, class ... Ti>
constexpr auto sumAruguments(T1 x1, Ti ...xi) -> double // decltype(x1 + sumAruguments(xi...))
{
return x1 + sumAruguments(xi...);
}
template<class...T>
constexpr auto avarage(T...xi) -> double
{
return sumAruguments(xi...) / countArguments(xi...);
}
I was unable to write it so it auto-deduce return type.
When I tried I get weird result for average(-2).
https://wandbox.org/permlink/brssPjggn64lBGVq
You can also calculate average using variable number of arguments. The principle of this a function that an unknown number of arguments is stored in a stack and we can take them.
double average(int n, ...) // where n - count of argument (number)
{
int *p = &n; // get pointer on list of number in stack
p++; // get first number
double *pp = (double *)p; // transformation of the pointer type
double sum = 0;
for ( int i = 0; i < n; pp++, i++ ) //looking all stack
sum+=(*pp); // summarize
return sum/n; //return average
}
And you can using this function like:
double av1 = average( 5, 3.0, 1.5, 5.0, 1.0, 2.0 );
double av2 = average( 2, 3.0, 1.5 );
But the number of arguments must match with the n.
Related
I have a set of integers {1,2}. I want to produce "Transform#1, Transform#2" where each element is tranformed and then result is accumulated with a delimiter.
What would be the easiest way to accomplish this? Do we have "folds", "maps" in c++?
We dont use boost.
You can use std::transform and std::accumulate
int main()
{
std::vector<int> v1 {1,2,3};
std::vector<std::string> v2;
std::transform(begin(v1), end(v1), std::back_inserter(v2), [](auto const& i) {
return std::string("Transform#") + std::to_string(i);
});
std::string s = std::accumulate(std::next(begin(v2)), end(v2), v2.at(0), [](auto const& a, auto const& b) {
return a + ", " + b;
});
std::cout << s;
}
prints Transform#1, Transform#2, Transform#3
You may want to use Range Adaptors. Boost already has them and they are coming to the standard with C++20.
Take a look at the boost::adaptors::transformed example here.
Also, check out the reference to get a better picture of what operations are supported by adaptors.
In the end, you can achieve much cleaner code and the performance difference is negligible (unlike in some other languages, where using this style of programming incurs heavy performance costs).
If you can stand a trailing separator, the following function can transform any iterable range of data { X, ..., Z } to the string "<tag>X<sep>...<sep><tag>Z<sep>".
Code
template <class InputIt>
std::string f(InputIt begin, InputIt end, std::string_view separator = ", ", std::string_view tag = "Transform#")
{
std::stringstream output;
std::transform(begin, end,
std::ostream_iterator<std::string>(output, separator.data()),
[tag](auto const& element){ return std::string{tag} + std::to_string(element); }
);
return output.str();
}
It works by transforming each element from the range into a stream iterator.
Usage
int main()
{
std::set<int> const data{1, 2, 3}; // works with vector, string, list, C-arrays, etc.
std::cout << f(begin(data), end(data)) << '\n';
// prints Transform#1, Transform#2, Transform#3,
}
Live demo
You can perform a fold using simply std::accumulate
#include <set>
#include <string>
#include <iostream>
#include <numeric>
int main()
{
auto transformation = [](int number) { return "Transform#" + std::to_string(number); };
auto transform_and_fold = [&transformation](std::string init, int number) { return std::move(init) + ", " + transformation(number); };
std::set<int> numbers{1, 2};
std::cout << std::accumulate(std::next(numbers.begin()), numbers.end(), transformation(*numbers.begin()), transform_and_fold);
}
Outputs
Transform#1, Transform#2
Assuming that I correctly understand the problem, the following straightforward implementation also looks very simple and easy.
This function works in C++11 and over:
DEMO with 5 test cases
std::string concatenate(
const std::vector<int>& indecies,
const std::string& delimiter = ", ",
const std::string& tag = "Transform#")
{
if(indecies.empty()){
return "";
}
std::string s(tag + std::to_string(indecies[0]));
for(auto it = indecies.begin()+1; it != indecies.cend(); ++it){
s += (delimiter + tag + std::to_string(*it));
}
return s;
}
(BTW, as for this function concatenate, if indecies is empty, the return value is also an empty string, not exceptions (AndreasDM's one) or UB (Everlight's one).
And if indecies has only a single element, for instance indecies={1}, then result is "Transform#1”, not "Transform#1, ”(YSC's one) or ", Transform#1”(sakra's one).
These are different from other answers and this function will be more simpler if this handling is removed.)
Although the performance may not be a focal point, the above function can be slightly optimized by pre-reserving the minimum capacity to save the resulted string by std::basic_string::reserve as follows.
Here +1 in *.size()+1 means the minimum length of a number character.
I also removed delimiter+tag in the for-loop.
This still looks simple:
DEMO with 5 test cases
std::string concatenate_fast(
const std::vector<int>& indecies,
std::string delimiter = ", ",
const std::string& tag = "Transform#")
{
if(indecies.empty()){
return "";
}
std::string s(tag + std::to_string(indecies[0]));
delimiter += tag;
s.reserve((tag.size()+1) + (indecies.size()-1)*(delimiter.size()+1));
for(auto it = indecies.begin()+1; it != indecies.cend(); ++it){
s += (delimiter + std::to_string(*it));
}
return s;
}
I have also tested the performance of these functions and some proposed answers as follows.
These tests are done by Quick C++ Benchmark within gcc-8.2, C++17 and O3 optimization.
Since std::transform_reduce is still not available in Quick C++ Benchmark, I haven’t tested it.
The above concatenate_fast shows best performance at least in these cases and concatenate is second best.
Finally, just personally, taking the balance of the readability and the performance into account, I would like to propose the above concatenate as a solution:
- Performance test with size 2 and 8. (DEMO)
- Performance test with size 16 and 32. (DEMO)
Unless you have some other requirement to preserve the intermediate tranformed list, storing it is suboptimal. You can just call std::accumulate and do both operations on the fly:
#include <cstdio>
#include <iterator>
#include <numeric>
int main ( )
{
int const input [] = { 1, 2, 3, 4, 5, 6 };
// computes sum of squares
auto const add_square = [] ( int x, int y ) { return x + y * y; };
int result = std::accumulate
( std::cbegin (input)
, std::cend (input)
, 0
, add_square
);
std::printf ( "\n%i\n", result );
return 0;
}
If you have the luxury of using C++17, there is a standard library algorithm which does exactly what you need. Here is an example:
#include <iterator>
#include <iostream>
#include <numeric>
#include <string>
int main()
{
auto input = {1, 2, 3};
std::cout << std::transform_reduce(
std::cbegin(input), std::cend(input),
std::string("Result:"),
[](const std::string & left, const std::string & right) { return left + " " + right; },
[](int value) { return "Transform#" + std::to_string(value); }
) << "\n";
}
I have a void function that takes as arguments iterators to the start and the end of a set of points (Setiterator set_begin, Setiterator set_end),a query point to perform some calculations and a iterator to the beginnig of a set of points ( where I am going to add the results)
void computedist(Setiterator set_begin, Setiterator set_end,
Vector const& query_point, DistanceIterator dist_begin )
{
std::transform(set_begin, set_end, dist_begin, calculation);
}
I have read that with std::transform I can do that calculations over the whole set of points, but I don't know how should I define the calculation to be done, as I am new to C++.
In this case, I want to compute the distance of the points to the query point:
I guess that my calculation should look like this
double calc_dist(double query_point, double point_of_the_set){
double dist;
dist = fabs(query_point - point_of_the_set);
return dist;
But I don't know how should I give the arguments to the function, since I am new to working with iterators.
Thank's!
The flavor of std::transform you are calling expects a UnaryOp function to call, to which it passes '*iterator' by const reference and expects a return value of the iterator::value_type which it writes to *dest.
#include <iostream>
#include <algorithm>
#include <vector>
int main() {
using valtype = int;
std::vector<valtype> v1 { 1, 2, 3, 4, 5 };
std::vector<valtype> v2(v1.size());
std::transform(v1.begin(), v1.end(), v2.begin(),
[] (const valtype& in) { return in * 2; });
for (auto& val : v2)
std::cout << val << '\n';
}
Live demo: http://ideone.com/CUUlvA
Here is a simple C++ question.
Description of the problem:
I have a function that takes as input an integer and returns a vector of zeros with length the input. Assume that I call the function many times with the same argument. What I want to avoid is that my function creates the vector of zeroes each time it is called. I want this to happen only the first time the function is called with the given input.
How I approached it: This brought to mind static variables. I thought of creating a static vector that holds the required zero vectors of each size, but wasn't able to figure out how to implement this. As an example I want something that "looks" like [ [0], [0,0], ...].
If there is a different way to approach such a problem please feel free to share! Also, my example with vectors is a bit specialised but replies that are more generic (concerning static variables that depend on the argument) would be greatly appreciated.
Side question:
To generalise further, is it possible to define a function that is only called once for each choice of arguments?
Thanks a lot.
You can have a map of sizes and vectors, one vector for each size:
#include <vector>
#include <map>
#include <cstddef>
std::vector<int>& get_vector(std::size_t size)
{
static std::map<size_t, std::vector<int> > vectors;
std::map<size_t, std::vector<int> >::iterator iter = vectors.find(size);
if (iter == vectors.end())
{
iter = vectors.insert(std::make_pair(size, std::vector<int>(size, 0))).first;
}
return iter->second;
}
If I understand correctly what you are trying to do, I don't think you will get the benefit you are expecting.
I wrote a quick benchmark to compare the performance of repeatedly creating a vector of zeros. The first benchmark uses the standard vector constructor. The second uses a function that only creates the vector the first time and stores it in a map:
const std::vector<int>& zeros(std::size_t size) {
static std::unordered_map<size_t, std::vector<int>> vectors;
auto find = vectors.find(size);
if (find != vectors.end())
return find->second;
auto insert = vectors.emplace(size, std::vector<int>(size));
return insert.first->second;
}
std::chrono::duration<float> benchmarkUsingMap() {
int sum = 0;
auto start = std::chrono::high_resolution_clock::now();
for (int i = 0; i != 10'000; ++i) {
auto zeros10k = zeros(10'000);
zeros10k[5342] = 1;
sum += zeros10k[5342];
}
auto end = std::chrono::high_resolution_clock::now();
std::cout << "Sum: " << sum << "\n";
return end - start;
}
std::chrono::duration<float> benchmarkWithoutUsingMap() {
int sum = 0;
auto start = std::chrono::high_resolution_clock::now();
for (int i = 0; i != 10'000; ++i) {
auto zeros10k = std::vector<int>(10'000);
zeros10k[5342] = 1;
sum += zeros10k[5342];
}
auto end = std::chrono::high_resolution_clock::now();
std::cout << "Sum: " << sum << "\n";
return end - start;
}
int main() {
std::cout << "Benchmark without map: " << benchmarkWithoutUsingMap().count() << '\n';
std::cout << "Benchmark using map: " << benchmarkUsingMap().count() << '\n';
}
Output:
Benchmark without map: 0.0188374
Benchmark using map: 0.134966
So, in this case, just creating the vector each time was almost 10x faster. This is assuming you want to create a mutable copy of the vector of zeros.
If each vector needs to be a separate instance then you will have to have a construction for each instance. Since you will have to construct each instance you can make a simple make_int_vector function like:
std::vector<int> make_int_vector(std::size_t size, int fill = 0)
{
return std::vector(size, fill);
}
The returned vector will either be moved or be elided with copy elision
What you are asking for is a cache. The hard part is how long an entry should exist in the cache. Your current requirement seems to be an eternal cache, meaning that each entry will persist for ever. For such a simple use case, à static map is enough:
template<typename T, typename U>
T cached(T (*funct)(U arg)) {
static unordered_map<U, T> c;
if (c.count(arg) == 0) {
c[arg] = funct(arg);
}
return c[arg];
}
The above is returning a value,which will require à copy. If you want to avoid the copy, just return a reference, but then, if you change one of the vectors, the next call will return the modified value.
template<typename T, typename U>
&T cached(T (*funct)(U arg)) {
static unordered_map<U, T> c;
if (c.count(arg) == 0) {
c[arg] = funct(arg);
}
return c[arg];
}
If I want to calculate the sum of a bunch of numbers retrieved from an std::istream, I can do the following:
// std::istream & is = ...
int total = std::accumulate(std::istream_iterator<int>(is),
std::istream_iterator<int>(),
0);
However, if I want to calculate their average, I need to accumulate two different results:
the total sum (std::accumulate)
the total count (std::distance)
Is there any way to "merge" these two algorithms and run them "side by side" in a single pass of an iterator range? I would like to do something like:
using std::placeholders;
int total, count;
std::tie(total, count) = merge_somehow(std::istream_iterator<int>(is),
std::istream_iterator<int>(),
std::bind(std::accumulate, _1, _2, 0),
std::distance);
double average = (double)total / count;
Is this possible?
A ready-made solution for this sort of single-pass accumulation is implemented by Boost.Accumulators. You make a single accumulator, say for sum, count and average, populate it, and then extract all three results at the end.
You cannot merge two different algorithms to be interleaved. The algorithms control the flow, and you can only have one flow. Now, in your particular case you can simulate it:
int count = 0;
int total = std::accumulate(std::istream_iterator<int>(is),
std::istream_iterator<int>(),
0,
[&](int x, int y) { ++count; return x+y; });
This is total hack, but something like this:
#include <iostream>
#include <algorithm>
#include <tuple>
#include <iterator>
#include <sstream>
namespace Custom {
template <class InputIterator, class T, class Bind, typename... Args>
std::tuple<Args...> accumulate (InputIterator first, InputIterator last,
T init, T& output, Bind bind, Args&... args)
{
while (first!=last) {
init = bind(init, *first, args...);
++first;
}
output = init;
std::tuple<Args...> tuple(args...);
return tuple;
}
}
int main() {
int total = 0, count = 0;
std::istringstream is;
is.str("1 2 3 4 5");
std::tie(count) = Custom::accumulate(std::istream_iterator<int>(is),
std::istream_iterator<int>(),
0,
total,
std::bind([&] (int a, int b, int& count) { ++count; return a + b; },
std::placeholders::_1, std::placeholders::_2, std::placeholders::_3),
count);
std::cout << total << " " << count;
return 0;
}
Have anybody written a C++ STL-compliant algorithm that combines std::transform and std::accumulate into a single pass algorithm supporting both the unary, binary and perhaps even (n-ary!) variant, say std::transformed_accumulate? I want this because I have found this pattern highly reusable in for example linear algebra for example in (l1-)norm calculations. The l1-norm calculates the sum of the absolute values of the elements.
Uhm... My bet is that you can do that by embedding your transformation into the binary predicate, tranform the element and accumulate after the transformation.
struct times2accumulator {
int operator()( int oldvalue, int newvalue ) const {
return oldvalue + 2*newvalue;
}
};
int r = std::accumulate( v.begin(), v.end(), 2, times2accumulator() );
That functor would be equivalent to:
struct times2 {
int operator()( int x ) {
return 2*x;
}
};
std::vector<int> tmp; tmp.reserve( v.size() );
std::transform( v.begin(), v.end(), std::back_inserter(tmp), times2 );
int r = std::accumulate( tmp.begin(), tmp.end(), 0 );
Of course this could be made generic, just pass the transformation functor to a generic base functor:
template <typename Transform>
struct transform_accumulator_t {
Transform t;
transform_accumulator_t( Transform t ) : t(t) {}
int operator()( int oldvalue, int newvalue ) const {
return oldvalue + t(newvalue);
}
};
// syntactic sugar:
template <typename T>
transform_accumulator_t<T> transform_accumulator( T t ) {
return transform_accumulator_t<T>(t);
}
int r = std::accumulate(v.begin(), v.end(), 0, transform_accumulator(times2));
And you could also generalize on the type in the container... or even create a more generic transform_accumulator that takes both an accumulator and a transformation functors and applies them in order. Actual implementation left as an exercise for the reader.
Although it may not exactly fit the original intent, std::inner_product is basically your binary version. You pass it an initial value, two ranges, and two functors, and it applies them as:
T acc = initial_value;
while (begin1 != end1) {
acc = binary_op1(acc, binary_op2(begin1, begin2);
++begin1;
++begin2;
return acc;
So, for your L1 you'd do something on this general order:
norm = std::inner_product(input1.begin(), input1.end(),
input2.begin(), input2.end(),
std::plus<int>(), std::abs);
Only that doesn't quite work -- right now, it's trying to pass std::abs where you really need a binary function that combines the two inputs, but I'm not sure how the two inputs are really supposed to be combined.
std::partial_sum is fairly close to your unary version, except that along with accumulating a result, it (attempts to) write out each intermediate result, not just the final result. To just get the final result, you'd have to write (and pass an instance of) a kind of do-nothing iterator that just holds a single value:
template<class T, class Dist=size_t, class Ptr = T*, class Ref = T&>
class unique_it : public std::iterator<std::random_access_iterator_tag, T, Dist, Ptr, Ref> {
T &value;
public:
unique_it(T &v) : value(v) {}
T &operator*() { return value; }
unique_it &operator++() { return *this; }
unique_it &operator+(size_t) { return *this; }
unique_it &operator++(int) { return *this; }
};
template <class T>
unique_it<T> make_res(T &v) { return unique_it<T>(v); }
With this, your L1 normalization would look something like this:
int main(){
double result=0.0;
double inputs[] = {1, -2, 3, -4, 5, -6};
std::partial_sum(
inputs, inputs+6,
make_res(result),
[](double acc, double v) {return acc + std::abs(v);});
std::cout << result << "\t";
return 0;
}
If you want to use some parallelism, I made a quick version using OpenMP :
template <class T,
class InputIterator,
class MapFunction,
class ReductionFunction>
T MapReduce_n(InputIterator in,
unsigned int size,
T baseval,
MapFunction mapper,
ReductionFunction reducer)
{
T val = baseval;
#pragma omp parallel
{
T map_val = baseval;
#pragma omp for nowait
for (auto i = 0U; i < size; ++i)
{
map_val = reducer(map_val, mapper(*(in + i)));
}
#pragma omp critical
val = reducer(val, map_val);
}
return val;
}
It is fast but there is certainly room for optimisation, especially around for (auto i = 0U; i < size; ++i) I think. (But I couldn't figure how to make an iterator-only version with OpenMP, any help would be appreciated!).
On a quick test with 1000000 elements array, and the computation iterated 1000 times to have a mean value, I made some comparisons.
Version 1 :
for (auto i = 0U; i < size; ++i)
val += std::pow(in[i][0], 2) + std::pow(in[i][1], 2);
score when compiled with:
g++ : 30 seconds
g++ -O3 : 2.6 seconds
Version 2 :
This version is the most optimized for this computation I think. (It gives the best result).
#pragma omp parallel reduction( + : val )
{
double map_val = 0.0;
#pragma omp for
for (int i=0; i < size; ++i)
{
map_val += std::pow(in[i][0], 2) + std::pow(in[i][1], 2);
}
val += map_val;
}
g++ -O3 : 0.2 seconds (it's the best one)
Version 3
This version uses the MapReduce_n function template I shown earlier :
double val = MapReduce_n(in, size, 0.0, [] (fftw_complex val)
{
return std::pow(val[0], 2.0) + std::pow(val[1], 2.0);
}, std::plus<double>());
g++ -O3 : 0.4 seconds, so there is a slight overhead for not using directly the OMP reduce directly. However, it doesn't allows custom operators, so at one point you (sadly) have to trade speed for genericity.
I am surprised noone said how to do this with Boost.Range:
accumulate(v | transformed((int(*)(int))&std::abs), 0);
where v is a Singe Pass Range (ie, any STL container). The abs overload has to be specified, otherwise this would be as elegant as Haskell.
As of C++17 there is also std::transform_reduce, which also has the benefit of being parallelizable.
https://en.cppreference.com/w/cpp/algorithm/transform_reduce