I just want to ask if there are any uses for passing variables in cpp as template arguments
template<int a> struct foo {
int x = a;
};
int main() {
foo<2> bar;
std::cout << bar.x;
}
Something like this compiles, works and cout's 2 but the same thing can be done by doing
struct foo {
int x;
foo(int a) : x(a) {}
};
int main() {
foo bar(2);
std::cout << bar.x;
}
So what is the point of using variables in template arguments? I can also see a big flaw in using the first method: the variable a uses memory and isn't destructed after x is changed, as it would be after the constructor is called in the second example. It might be helpful if you showed some reasonable uses for that.
When you pass a variable through a template argument, it can be used in compile time.
For example, if you need to create a statically sized array in your class, you could use the template argument to pass the size of your array:
template <int TSize>
class Foo {
[...] // Do whatever you need to do with mData.
private:
std::array<int, TSize> mData;
};
There are many uses for constants in template parameters.
Static Sizes
This is how you would start implementing something like a std::array.
template <typename T, size_t SIZE>
struct Array {
T data[SIZE];
}
Template parameters are always usable in a constexpr context, so they can be used as sizes for statically sized arrays.
Providing Compile-Time Parameters to Algorithms
Another use is parametrizing algorithms like in the following code sample.
We have a uint32_t in ARGB order but to store it in a file, we might need to reorder it to BGRA or RGBA. We know the order at compile time, so we could use an ArgbOrder template variable.
enum class ArgbOrder { ARGB, RGBA, BGRA };
struct ChannelOffsets {
unsigned a;
unsigned r;
unsigned g;
unsigned b;
};
// and we can get a constexpr lookup table from this enum
constexpr ChannelOffsets byteShiftAmountsOf(ArgbOrder format)
{
...
}
template <ArgbOrder order>
void encodeArgb(uint32_t argb, uint8_t out[4])
{
// We can generate the shift amounts at compile time.
constexpr detail::ChannelOffsets shifts = shiftAmountsOf(order);
out[0] = static_cast<u8>(argb >> shifts.a);
out[1] = static_cast<u8>(argb >> shifts.r);
out[2] = static_cast<u8>(argb >> shifts.g);
out[3] = static_cast<u8>(argb >> shifts.b);
}
void example() {
encodeArgb<ArgbOrder::BGRA>(12345);
}
In this example, we can select the appropriate lookup table at compile time and have zero runtime cost. All that needs to happen at runtime is 4 shifts.
Feature Toggles
We can use bool template variables to toggle features in our code, like for example:
template <bool handleZeroSpecially>
int div(int x, int y) {
if constexpr (handleZeroSpecially) {
return y == 0 ? 0 : x / y;
}
else {
return x / y;
}
}
Related
I am looking for a convenient to create a C++ class where some member variables are only present if a template flag is set. As a simple example, let's assume I want to toggle an averageSum in an performance sensitive calculation, i.e.
struct Foo {
// Some data and functions..
void operator+=(const Foo& _other) {}
};
template<bool sumAverages>
class Calculator {
public:
// Some member variables...
// Those should only be present if sumAverages is true
int count = 0;
Foo resultSum;
void calculate(/* some arguments */) {
// Calculation of result...
Foo result;
// This should only be calculated if sumAverages is true
++count;
resultSum += result;
// Possibly some post processing...
}
};
One way would be using preprocessor defines, but those are rather inconvenient especially if I need both versions in the same binary. So I am looking for an alternative using templates and if constexpr and something like the following Conditional class:
template<bool active, class T>
struct Conditional;
template<class T>
struct Conditional<true, T> : public T {};
template<class T>
struct Conditional<false, T> {};
My first shot was this:
template<bool sumAverages>
class Calculator {
public:
int count = 0;
Conditional<sumAverages, Foo> resultSum;
void calculate(/* some arguments */) {
Foo result;
if constexpr(sumAverages) {
++count;
resultSum += result;
}
}
};
The if constexpr should incur no run time cost and as it is dependent on a template variable should allow non-compiling code in the false case (e.g. in this example Conditional<false, Foo> does not define a += operator, still it compiles). So this part is more or less perfect. However the variables count and resultSum are still somewhat present. In particular, as one can not derive from a fundamental type, the Conditional class does not allow to toggle the int dependent on the template. Furthermore every Conditional<false, T> variable still occupies one byte possibly bloating small classes. This could be solvable by the new [[no_unique_address]] attribute, however my current compiler chooses to ignore it in all my tests, still using at leas one byte per variable.
To improve things I tried inheriting the variables like this
struct OptionalMembers {
int count;
Foo resultSum;
};
template<bool sumAverages>
class Calculator : public Conditional<sumAverages, OptionalMembers> {
public:
void calculate(/* some arguments */) {
Foo result;
if constexpr(sumAverages) {
++OptionalMembers::count;
OptionalMembers::resultSum += result;
}
}
};
This should come at no space cost as inheriting from am empty class should do literally nothing, right? A possible disadvantage is that one cannot freely set the order of the variables (the inherited variables always come first).
My questions are:
Do you see any problems using the approaches described above?
Are there better options to de(activate) variables like this?
There are a different ways to solve this, one straightforward one would be using template specialization:
#include <iostream>
template <bool b> struct Calculator {
int calculate(int i, int j) { return i + j; }
};
template <> struct Calculator<true> {
int sum;
int calculate(int i, int j) { return sum = i + j; }
};
int main(int argc, char **argv) {
Calculator<false> cx;
cx.calculate(3, 4);
/* std::cout << cx.sum << '\n'; <- will not compile */
Calculator<true> cy;
cy.calculate(3, 4);
std::cout << cy.sum << '\n';
return 0;
}
Another solution would be to use mixin-like types to add features to your calculator type:
#include <iostream>
#include <type_traits>
struct SumMixin {
int sum;
};
template <typename... Mixins> struct Calculator : public Mixins... {
int calculate(int i, int j) {
if constexpr (is_deriving_from<SumMixin>()) {
return SumMixin::sum = i + j;
} else {
return i + j;
}
}
private:
template <typename Mixin> static constexpr bool is_deriving_from() {
return std::disjunction_v<std::is_same<Mixin, Mixins>...>;
}
};
int main(int argc, char **argv) {
Calculator<> cx;
cx.calculate(3, 4);
/* std::cout << cx.sum << '\n'; <- will not compile */
Calculator<SumMixin> cy;
cy.calculate(3, 4);
std::cout << cy.sum << '\n';
return 0;
}
I'm not sure why the array creation in the function passes but not the one in the class even though array size is a compile time computable value.
template<int N>
int getPow()
{
int power = 1;
while(power < N)
power <<= 1;
return power;
}
template<int N>
class Test
{
private:
int data[getPow<N>()];
};
void testfun()
{
int test[getPow<2>()]; // passes
Test<10> t1; // Fails????
}
As getPow is not constexpr, it cannot be used in places which require constant expression (as C-array size).
int test[getPow<2>()]; // passes
. You unfortunately use VLA extension. It should not pass.
You might solve your issue with:
template <unsigned N>
constexpr unsigned getPow()
{
return 1 << N;
}
I want to calculate factorial at the compile-time. I found some way to solve the problem, but I want to know if there any another solution for this problem without using enum s. Here the solution using enum s.
#include <iostream>
template <int n>
struct fact
{
enum{value = n*fact<n-1>::value};
};
template<>
struct fact<1>
{
enum{value = 1};
};
int main()
{
std::cout << fact<10>::value;
}
If there is no another solution, please describe why the enum s are must.
While there are alternative notations, It's written that way because more compilers accept that enum-style notation. The language supports const integral-type class members with an inline initialization, but some compilers aren't compliant with the standard in that regard. On compilers that are compliant in this regard, the following works just fine:
#include <iostream>
template <unsigned int n>
struct fact
{
static const unsigned int value = n*fact<n-1>::value;
};
template<>
struct fact<0>
{
static const unsigned int value = 1;
};
int main()
{
std::cout << fact<10>::value << "\n";
}
Replace,
enum{value};
with,
static int const value; // or unsigned int
enums are must because they suppose to be resolved at compile time. Which assures that whatever result you calculated must have been done at compile time. Other such type is static int const (means any integral type).
To illustrate:
enum E {
X = strlen(s); // is an error, because X is a compile time constant
};
Alternatively, you can use static const members:
template <unsigned int n>
struct fact { static const unsigned int value = n * fact<n-1>::value; }
I have this problem (histogramming). I've a real space: [a,b] partitioned in some way ([a0=a, a1, a2, ..., b]). The partitioning may be with equal space (a1 - a0 = a2 - a1 = ...) or variables.
I need a class that handle this, with some methods to say given a value in which bin of the partition it belongs; other methods to find the center of a particular bin and more.
During the program I don't like to instantiate a class only to call these simple function like
Binner binner(binning);
binner.get_bin(1.3);
binner.get_centerbin(2);
so I tried to write a static class using template to do something like that:
Binner<binning>::get_bin(1.3);
Binner<binning>::get_centerbin(2);
is it a good idea? There are other way to do it? Now I've free functions like
double get_bin(double bin, Binning binning); // a lot of if/else inside
but I think it's too error prone.
Here my implementation:
enum Binning {CELL, LARGE, BE};
const double binning_LARGE[] = {0, 1.2, 1.425, 1.550, 1.800, 2.5};
const double binning_BE[] = {0, 1.425, 1.550, 2.5};
template<Binning binning>
class Binner
{
public:
static const double* bins;
static const int n;
static int get_bin(double value);
};
template<> const double* myclass<LARGE>::bins = binning_LARGE;
template<> const double* myclass<BE>::bins = binning_BE;
template<> const int myclass<LARGE>::n = sizeof(binning_LARGE) / sizeof(double);
template<> const int myclass<BE>::n = sizeof(binning_BE) / sizeof(double);
template<Binning binning> int myclass<binning>::get_bin(double value)
{
return find_if(bins, bins + n,
bind2nd(greater<double>(), value)) - bins - 1;
}
template<> int myclass<CELL>::get_bin(double value)
{
return static_cast<int>(value / 0.025);
}
is it a good implementation / design?
Is there a way to avoid the n field using std::vector? How?
Is there a way to parametrize the 0.025? I know that double can't be template parameter but can I write something similar to this:
Binner<0.025> binner;
other / advices?
Edit:
For the third point Why I can't do that:
template<Binning binning, int N=100>
class Binner
{
public:
static const double* bins;
static const int n;
static int bin(double value);
};
...
template<Binning binning, int N> int Binner<CELL, N>::bin(double value)
{
return static_cast<int>(value / (2.5 / N));
}
IMHO, your design is ok, if you do not want to instantiate a class. Indeed, it seems a kind of template metaprogramming to me. Whether this makes sense depends on how you are planning to reuse this template.
Using a std::vector would allow you to get rid of the variable to hold the array size, for sure. Now, if this is good for your design, I don't know... it would move some complexity out of your template definition, to the binning definition (which now you can initialize very simply)...
Finally, you can instantiate your template passing a constant to it:
template < Binning binning, unsigned long N, unsigned long M>
class ... {
<using N>
}
have you considered a traits class? Typically if you have static information that you want to separate from the behaviour in a class, you might consider creating a traits class that encapsulates that.
So I'd start with the default behaviour:
enum Binning {CELL, LARGE, BE};
template <Binning binning>
struct BinTraits
{
// default behaviour
int get_bin(double value) { return value / 0.025; }
};
Then I'd provide the specialisations:
const double binning_LARGE[] = {0, 1.2, 1.425, 1.550, 1.800, 2.5};
const double binning_BE[] = {0, 1.425, 1.550, 2.5};
template <typename RandomAccessCollectionT>
int get_bin_impl(double value, RandomAccessCollectionT collection, unsigned size)
{
return find_if(collection, collection + size,
bind2nd(greater<double>(), value)) - collection - 1;
}
template <>
struct BinTraits<LARGE>
{
int get_bin(double value) { return get_bin_impl(value, binning_LARGE, sizeof(binning_LARGE) / sizeof(binning_LARGE[0])); }
};
template <>
struct BinTraits<BE>
{
int get_bin(double value) { return get_bin_impl(value, binning_BE, sizeof(binning_BE) / sizeof(binning_BE[0])); }
};
Then I'd put the actual container behaviour in another class that requires binning behaviour (lets call it HashTable):
template <typename BinTraits>
class HashTable
{
public:
void insert(double value)
{
int bin = BinTraits::get_bin(value);
_bins[bin].insert(value);
}
// _bin is a multimap or something
};
Looking at the usage of find_if and bind2nd as well as functors, it seems as if you are quite knowledgeable about STL and some advanced C++ concepts; yet, what you are trying to do seems to be over-engineering. While I can't fully understand what you are trying to do but it seems that you could do away with templates completely and use just a class (instantiated with different values) and method parameters.
How should I change the code below so that Array<Index> array; is enough and the SIZE is automatically deduced from the enum?
Even if the enum changes, it is guaranteed that it contains SIZE referring to the correct size.
template <typename Enum, int N>
class Array {
public:
int& operator[](Enum index) { return array[index]; }
private:
int array[N];
};
enum Index { X, Y, SIZE };
int main() {
Array<Index, SIZE> array;
array[X] = 1;
return 0;
}
UPDATE: As for "Array<type> means you're creating an array of Type objects" (Jerry) and "the name of class template is a bit misleading" (Nawaz): actually I am creating CustomSqlQueryModel<TableColumns>. The above is just a simplified code, nothing more. Jerry and Nawaz are rigth: this simplified code is unfortunate.
You can write a traits class. This requires a bit of extra work each time you define a new enum type, but no extra work for each occurrence of Array<Index> in user code:
template<class Enum>
struct ArrayTraits;
template<class Enum>
struct Array {
int& operator[](Enum index) { return array[index]; }
private:
int array[ArrayTraits<Enum>::size];
};
enum Index { X, Y, SIZE };
template<>
struct ArrayTraits<Index> {
enum { size = SIZE };
};
int main() {
Array<Index> array;
array[X] = 1;
return 0;
}
One of the advantages of this is you can specialize the traits for external enums you don't control, as long as you know how to get the max size.
As stated, I don't think you can. If, however, you change it to something like:
struct Index {
enum { X, Y, SIZE};
};
Then your template could be something like:
template <class Enum>
class Array {
// ...
private:
int array[Enum::SIZE];
};
...and if the type you pass as Enum doesn't include some positive constant named SIZE,the instantiation won't compile. For the purpose at hand, you'd really kind of prefer that Index was a namespace, but since a namespace isn't a type, I don't think you can use it as a template argument.
I should add, however, that I'm not sure I like this idea at all -- most people are going to think Array<type> means you're creating an array of Type objects, and this is clearly something entirely different from that...
If you want only the size to be template argument, not the type , as from your example it seems that the type of the array would be always int, then why don't you implement this:
template <int size>
class Array {
public:
int& operator[](int index) { return array[index]; }
//Note this addition!
int operator[](int index) const { return array[index]; }
private:
int array[size];
};
int main() {
Array<10> array;
array[0] = 1;
array[1] = 2;
return 0;
}
Note this addition: it's better if you implement const version of operator[] too, so that const Array<> can use it to access the array elements, otherwise your class wouldn't work for const Array<>.