Convert a vector<unsigned char> to vector<unsigned short>

Question

Hi all. I'm getting data from a binary file, reading from file and writing in a vector of unsigned char. I can't edit it, because I'm using a external library.

But the data that I'm reading from file is a 16 bits image, and I'd like to put the data in a vector of unsigned short

Maybe I can do a cast for it?

Rgds.

Answer 1

vector<unsigned char> a = ...;
vector<unsigned short> b(a.begin(), a.end());

But you want valarrays for simple data vectors.

Answer 2

A generic approach (not bullet proof):

#include <vector>
#include <iostream>
#include <iterator>
#include <algorithm>

typedef unsigned char u8;
typedef unsigned short u16;

u16 combine_two_bytes(u8 a, u8 b) {
    return a | (b << 8);
}

template<typename InIter, typename OutIter, typename InT, typename OutT>
void combine_pairs(InIter in, InIter in_end, OutIter out, OutT (*func)(InT, InT)) {
    while(1) {
        if(in == in_end) {
            break;
        }

        InT &left = *in++;

        if(in == in_end) {
            break;
        }

        InT &right = *in++;

        *out++ = func(left, right);
    }
}

int main() {
    using namespace std;    // lazy

    u8 input[] = { 1, 2, 3, 4, 5, 6, 7, 8 };
    const size_t in_size = sizeof(input) / sizeof(*input);
    u16 output[in_size / 2];

    cout << "Original: ";
    copy(input, input + in_size, ostream_iterator<int>(cout, " "));
    cout << endl;

    combine_pairs(input, input + in_size, output, combine_two_bytes);

    cout << "Transformed: ";
    copy(output, output + in_size / 2, ostream_iterator<int>(cout, " "));
    cout << endl;

    return 0;
}

Answer 3

Assuming the binary data in your file is in little-endian order, I would do this the simple way:

vector<unsigned char> a = ...;

std::vector<unsigned short> b;
b.reserve( a.size() / sizeof(unsigned short) );

for( std::vector<unsigned char>::const_iterator i=a.begin(); i!=a.end(); i+=2 )
{
    unsigned short shortValue = *(i+1);
    shortValue <<= 8;
    shortValue |= *i;
    b.push_back( shortValue );
}

If the data in your file is big-endian you will need to compose the short value the other way round. You should also guard against the number of entries in "a" not being a multiple of 2.

Answer 4

disclaimer: I don't have a compiler right now:

vector<unsigned char> vec = getVector();
vector<unsigned short> sv(reinterpret_cast<unsigned short*>(&vec[0]), 
                          reinterpret_cast<unsigned short*>(&vec[vec.size()]));

Answer 5

If you just want to convert from one type to the other then use the standard constructor. As long as the iterators value type is auto convertible to the destination vectors value type the compiler will do the auto conversion between the two types. Just use the standard constructor

#include <vector>
#include <algorithm>
#include <iterator>

int main()
{
    std::vector<unsigned char>      a;
    a.push_back((unsigned char)12);
    a.push_back((unsigned char)13);
    a.push_back((unsigned char)14);

    std::vector<unsigned short>     b(a.begin(),a.end());

    // Print out the vector
    std::copy(b.begin(),b.end(),std::ostream_iterator<unsigned short>(std::cout,"\t"));
}

> g++ t.cpp
> ./a.out
12  13 14

If you actually want to convert two bytes into one then some work is required. But it depends if the input data is actually the same endianess as the machine you are on. If you know that it is the same endianess that you just need to cast the input type.

#include <vector>
#include <algorithm>
#include <iterator>

int main()
{
    std::vector<unsigned char>      a;

    // Make sure that the size is correct.
    // ie. An Odd number indicates that something is not quite correct.
    //
    std::vector<unsigned short>     b(static_cast<unsigned short*>(&a[0]),
                                      static_cast<unsigned short*>(&a[a.size()]));

    // Print out the vector
    std::copy(b.begin(),b.end(),std::ostream_iterator<unsigned short>(std::cout,"\t"));
}

Alternatively if you actually need to combine two values into a single value where the endianess is not the same as the target architecture, you can write a special iterator. Something like this:

#include <Converter.h>

int main()
{
    std::vector<unsigned char>      a;

    // Make sure that the size is correct.
    // ie. An Odd number indicates that something is not quite correct.
    //
    std::vector<unsigned short>     b(make_Converter(a.begin()),make_Converter(a.end()));

    // Print out the vector
    std::copy(b.begin(),b.end(),std::ostream_iterator<unsigned short>(std::cout,"\t"));
}

Converter.h

#include <vector>
#include <iostream>
#include <iterator>

template<typename I>
struct Converter
{
    I   iterator;

    typedef typename std::input_iterator_tag                    iterator_category;
    typedef typename std::iterator_traits<I>::value_type        value_type;
    typedef typename std::iterator_traits<I>::difference_type   difference_type;
    typedef typename std::iterator_traits<I>::pointer           pointer;
    typedef typename std::iterator_traits<I>::reference         reference;

    Converter(I iter)
        :iterator(iter)
    {}

    Converter& operator++()
    {
        iterator++;
        return *this;
    }

    Converter operator++(int)
    {
        Converter   tmp(*this);
        this->operator++();

        return (tmp);
    }

    value_type operator*()
    {
        /*
         * The actual calculation done here will depend on the underlying hardware.
         */
        typename std::iterator_traits<I>::value_type val(*iterator);
        val << 8;
        iterator++;
        val |= (*iterator);

        return val;
    }

    bool operator!=(Converter const& rhs)
    {
        return iterator != rhs.iterator;
    }
};

template<typename I>
Converter<I> make_Converter(I iter)
{
    return Converter<I>(iter);
}