Reputation: 35408
I need an "elegant" way of initializing a vector in the declaration phase with the content of another one and a few extra elements.
What I want to solve is:
Let's consider the following (example) declaration with initialization:
const std::vector<std::string> c90_types = {
"char",
"signed char",
"unsigned char",
"short",
"unsigned short",
"int",
"unsigned int",
"long",
"unsigned long",
"float",
"double",
"long double"
};
const std::vector<std::string> c99_types = {
"char",
"signed char",
"unsigned char",
"short",
"unsigned short",
"int",
"unsigned int",
"long",
"unsigned long",
"float",
"double",
"long double",
"long long",
"unsigned long long",
"intmax_t",
"uintmax_t"
};
as you can see c99_types
has a subset which is exactly c90_types
. I want to avoid the situation where I need to change the subset and then change manually the "superset" too, just to avoid the extra step that might introduce bugs :)
As a side note, I don't want to write code like:
second.insert(second.begin(), first.begin(), first.end());
second.push_back(something);
Any good and clean solutions to this?
Upvotes: 17
Views: 2631
Reputation: 2965
There is a trick called "I want to initialize a const variable with something elaborate." that became possible with C++11, shamelessly stolen from Javascript.
const std::vector<std::string> c90_types = {
"char",
// and so on, and so forth....
};
const std::vector<std::string> c99_types = ([&](){
const auto additional_types = { // initializer_list<const char *>, but it does not matter.
"long long",
"unsigned long long",
"intmax_t",
"uintmax_t"
};
std::vector<std::string> vec{c90_types};
vec.insert(vec.end(), additional_types.begin(), additional_types.end());
return vec;
})();
Pack your initialization logic into an unnamed lambda, and call it right away, copy-initializing your const variable.
vec
is moved, not copied.
Upvotes: 18
Reputation: 13288
You can use boost::join
:
#include <vector>
#include <boost/range/join.hpp>
const std::vector<std::string> c90_types = {
"char",
"signed char",
"unsigned char",
"short",
"unsigned short",
"int",
"unsigned int",
"long",
"unsigned long",
"float",
"double",
"long double"
};
auto range = boost::join(c90_types, std::vector<std::string>{
"long long",
"unsigned long long",
"intmax_t",
"uintmax_t"
});
const std::vector<std::string> c99_types(range.begin(), range.end());
Upvotes: 4
Reputation: 38218
std::array
This can probably be cleaned up and improved a lot, but it's at least a starting point (it uses Jonathan Wakely's redi::index_tuple
:
template<typename T, std::size_t N, unsigned... I, typename ...U>
inline auto
append_array_helper(const std::array<T, N>& array, redi::index_tuple<I...>, U&&... elements) -> std::array<T, N + sizeof...(elements)>
{
return std::array<T, N + sizeof...(elements)>{ std::get<I>(array)..., std::forward<U>(elements)... };
}
template<typename T, std::size_t N, typename ...U>
inline auto
append_array(const std::array<T, N>& array, U&&... elements) -> std::array<T, N + sizeof...(elements)>
{
return append_array_helper(array, typename redi::make_index_tuple<N>::type(), std::forward<U>(elements)...);
}
const std::array<std::string, 12> c90_types = {
"char",
"signed char",
"unsigned char",
"short",
"unsigned short",
"int",
"unsigned int",
"long",
"unsigned long",
"float",
"double",
"long double"
};
const std::array<std::string, 16> c99_types = append_array(
c90_types,
"long long",
"unsigned long long",
"intmax_t",
"uintmax_t"
);
If you don't want to specify the array size, you can use the following method:
template<typename T, typename... U>
constexpr auto make_array(U&&... elements) -> std::array<T, sizeof...(elements)>
{
return { std::forward<U>(elements)... };
}
const auto c90_types = make_array<std::string>(
"char",
"signed char",
"unsigned char",
"short",
"unsigned short",
"int",
"unsigned int",
"long",
"unsigned long",
"float",
"double",
"long double"
);
...
Not my favorite, but it's simple and easy to understand and edit:
#define C90_TYPES \
"char", \
"signed char", \
"unsigned char", \
"short", \
"unsigned short", \
"int", \
"unsigned int", \
"long", \
"unsigned long", \
"float", \
"double", \
"long double"
#define C99_TYPES \
C90_TYPES, \
"long long", \
"unsigned long long", \
"intmax_t", \
"uintmax_t"
const std::vector<std::string> c90_types = {
C90_TYPES
};
const std::vector<std::string> c99_types = {
C99_TYPES
};
Upvotes: 5
Reputation: 7320
You could define the biggest vector (here that would be c99_types) first, and then construct the others with iterators from the largest one.
Here is an example :
const vector<int> a{1,2,3,4};
const vector<int> b{begin(a), begin(a)+2}; // b is {1,2}
So you could write:
const std::vector<std::string> c99_types = {
"char",
"signed char",
"unsigned char",
"short",
"unsigned short",
"int",
"unsigned int",
"long",
"unsigned long",
"float",
"double",
"long double",
"long long",
"unsigned long long",
"intmax_t",
"uintmax_t"
};
const std::vector<std::string> c90_types{begin(c99_types), begin(c99_types)+12};
Upvotes: 7
Reputation: 217283
With extra code, you may still have const vector:
std::vector<std::string> make_c99_type()
{
auto res = c90_types;
const std::vector<std::string> extra_c99_types = {
"long long",
"unsigned long long",
"intmax_t",
"uintmax_t"
};
res.insert(res.end(), extra_c99_types.begin(), extra_c99_types.end());
return res;
}
const std::vector<std::string> c99_types = make_c99_type();
Upvotes: 3