0

我试图获得表现得像真假面具的范围视图。为了进行逻辑运算,我想实现掩码的 ands 和 ors。我有一个工作编译时间或:

struct make_or_mask_fn
{
    template<typename... Msks>
    auto operator()(Msks&&... msks) const
    {
        CONCEPT_ASSERT((Range<Msks>() || ...));
        return ranges::view::zip(std::forward<Msks>(msks)...) |
               ranges::view::transform(
                   [](auto&& range_item) -> bool {
                       return tuple_or(range_item);
                   });
    }

private:
    template<typename... T>
    static bool variable_length_or(const T... v)
    {
        return (v || ...);
    }
    template<typename... T, std::size_t... Idx>
    static bool tuple_or(const std::tuple<T...> t,
                         std::index_sequence<Idx...>)
    {
        return variable_length_or(std::get<Idx>(t)...);
    }
    template<typename... T>
    static bool tuple_or(const std::tuple<T...> t)
    {
        return tuple_or(t, std::index_sequence_for<T...>{});
    }
};
RANGES_INLINE_VARIABLE(make_or_mask_fn, make_or_masker)

我可以称之为很好

std::vector<bool> mask1 = ...
std::vector<bool> mask2 = ...
std::vector<bool> mask3 = ...
auto or_of_masks = make_or_masker(mask1, mask2, mask3);

目前这不能做的是构建一些在编译时未知的掩码。我目前的尝试是接受一个vector范围,检查它的大小,然后调用可变参数模板或从之前:

struct make_vector_or_mask_fn
{
    template<typename Msk>
    auto operator()(std::vector<Msk> msks)
        const // TODO const and reference types
    {
        CONCEPT_ASSERT(Range<Msk>());
        // todo return range with all true (an or of zero elements is true)
        assert(msks.size() != 0);
        if(msks.size() == 1)
            return or_ranges(msks[0]);
        if(msks.size() == 2)
            return or_ranges(msks[0], msks[1]);
        if(msks.size() == 3)
            return or_ranges(msks[0], msks[1], msks[2]);
        /// TODO: go until ... maybe 8 and recurse afterwards
    }
private:
    template<typename... Msks>
    static auto or_ranges(Msks&&... msks)
    {
        CONCEPT_ASSERT((Range<Msks>() || ...));
        return ranges::view::zip(std::forward<Msks>(msks)...) |
               ranges::view::transform(
                   [](auto&& range_item) -> bool {
                       return tuple_or(range_item);
                   });
    }
    template<typename... T>
    static bool variable_length_or(const T... v)
    {
        return (v || ...);
    }
    template<typename... T, std::size_t... Idx>
    static bool tuple_or(const std::tuple<T...> t,
                         std::index_sequence<Idx...>)
    {
        return variable_length_or(std::get<Idx>(t)...);
    }
    template<typename... T>
    static bool tuple_or(const std::tuple<T...> t)
    {
        return tuple_or(t, std::index_sequence_for<T...>{});
    }
};
RANGES_INLINE_VARIABLE(make_vector_or_mask_fn, make_vector_or_masker)

这不会编译并出现以下错误:

../include/range/v3/view/mask.hpp:221:25: error: 'auto' in return type deduced as 'ranges::v3::transform_view<ranges::v3::zip_view<ranges::v3::iterator_range<__gnu_cxx::__normal_iterator<int *, std::vector<int, std::allocator<int> > >, __gnu_cxx::__normal_iterator<int *, std::vector<int, std::allocator<int> > > >, ranges::v3::iterator_range<__gnu_cxx::__normal_iterator<int *, std::vector<int, std::allocator<int> > >,
  __gnu_cxx::__normal_iterator<int *, std::vector<int, std::allocator<int> > > > >, (lambda at ../include/range/v3/view/mask.hpp:188:32)>' here but deduced as 'ranges::v3::transform_view<ranges::v3::zip_view<ranges::v3::iterator_range<__gnu_cxx::__normal_iterator<int *, std::vector<int, std::allocator<int> > >, __gnu_cxx::__normal_iterator<int *, std::vector<int, std::allocator<int> > > > >,
  (lambda at ../include/range/v3/view/mask.hpp:188:32)>' in earlier return statement
                    return or_ranges(msks[0], msks[1]);
                    ^

据我了解,这告诉我我or_ranges根据参数的数量返回不同的类型。(zip我使用的保留了压缩在一起的知识)。

所以我想知道,如何键入擦除范围内的内容?

4

1 回答 1

1

返回ranges::v3::any_input_view < bool >而不是auto.

即使在优化的构建中,这也会慢 10 倍。类型擦除是昂贵的。

一个手动解决方案知道输入范围有一个长度,并在分发之前获取缓冲块中的数据,这可能会使性能更接近非类型擦除。我希望它仍然会明显变慢。

于 2018-08-29T20:16:57.627 回答