如何创建过滤向量的迭代器?

2022-01-24 00:00:00 lambda containers filter iterator c++

假设我有一个名为 SpotDeal 的向量 spot_deals,它是一个类:

Suppose I have a vector named spot_deals of SpotDeal that is a class:

class SpotDeal
{
public:
    int deal_id_; // primary key, and vector is sorted by id
    string ccy_pair_; // ccy pair, e.g. GBPUSD, AUDUSD
    double amount_;
}

假设我需要将 spot_deals 的两个子集传递给函数 foo 进行一些计算.但是,我可以制作副本,这会耗费内存和时间.实际上 foo 只需要交易的迭代器.那么我可以创建 vector 的 2 个迭代器,即 it1it2 并将它们传递给 foo 吗?

Say I need to pass two subset of spot_deals to a function foo for some computation. I could make copies, however, that would cost memory and time. Actually foo only needs iterators of deals. So can I make 2 iterators of vector<SpotDeal>, namely it1 and it2 and pass them to foo?

spot_deals 的两个子集可以被 ccy_pair_ 过滤,例如GBPUSD 和 AUDUSD 的交易,或其他条件.所以我正在寻找一种方法来定义一个由向量??和 lambda 函数定义的迭代器(虽然可以等效地是一个仿函数).

The two subset of spot_deals could be filtered by ccy_pair_, e.g. deals of GBPUSD and AUDUSD, or by other conditions. So I'm looking for a way to define an iterator defined by a vector and a lambda function (could equivalently be a functor though).

有没有办法编写一个辅助函数 make_filtered_iterator 以便我可以有类似下面的东西?

Is there a way to write a helper function make_filtered_iterator so that I can have something like below?

auto it1 = make_filtered_iterator(spot_deals, filter_lambda1);
auto it2 = make_filtered_iterator(spot_deals, filter_lambda2);
foo(it1, it2);

推荐答案

答案肯定是是".STL 风格的 C++ 迭代器可以用来做各种技巧.一个常见但基本的方法是为 std::map 创建一个迭代器,在取消引用时只给出键或值.

The answer is certainly "yes." C++ iterators in the STL style can be made to do all sorts of tricks. A common but basic one is making an iterator for std::map which when dereferenced gives only the key or the value.

在您的特定情况下,一个简单的实现可能是这样的:

In your particular case, a simple implementation might be like this:

template <typename BaseIterator>
struct filtered_iterator : BaseIterator
{
    typedef std::function<bool (const value_type&)> filter_type;

    filtered_iterator() = default;
    filtered_iterator(filter_type filter, BaseIterator base, BaseIterator end = {})
        : BaseIterator(base), _end(end), _filter(filter_type) {
        while (*this != _end && !_filter(**this)) {
            ++*this;
        }
    }

    filtered_iterator& operator++() {
        do {
            BaseIterator::operator++();
        } while (*this != _end && !_filter(**this));
    }

    filtered_iterator operator++(int) {
        filtered_iterator copy = *this;
        ++*this;
        return copy;
    }

private:
    BaseIterator _end;
    filter_type _filter;
};

template <typename BaseIterator>
filtered_iterator<BaseIterator> make_filtered_iterator(
        typename filtered_iterator<BaseIterator>::filter_type filter,
        BaseIterator base, BaseIterator end = {}) {
    return {filter, base, end};
}

我为 end 设置了一个默认值,因为通常您可以为此使用默认构造的迭代器.但在某些情况下,您可能只想过滤容器的一个子集,在这种情况下,指定结尾会很容易.

I set a default value for end because typically you can use a default-constructed iterator for that. But in some cases you might want to filter only a subset of the container, in which case specifying the end makes it easy.

相关文章