itertools.tee equivalent in Boost::Range?
Asked Answered
M

1

9

Python's itertools has tee for n-plicating iterables:

def tee(iterable, n=2):
    it = iter(iterable)
    deques = [collections.deque() for i in range(n)]
    def gen(mydeque):
        while True:
            if not mydeque:             # when the local deque is empty
                newval = next(it)       # fetch a new value and
                for d in deques:        # load it to all the deques
                    d.append(newval)
            yield mydeque.popleft()
    return tuple(gen(d) for d in deques)

I couldn't find the equivalent in Boost::Range. Am I missing something or should I just roll my own?

Malicious answered 1/11, 2012 at 16:52 Comment(3)
Maybe add a better plain text description of what you want, not everybody might understand what yield or list expressions do in python.Papacy
Do you have a use case for this?Microbe
It's been a while, but the idea was to n-plicate an iterator to a set of consumers (e.g. different threads) that would apply different operations. Similar to the tee command in Unix. I ended up using a completely different approach where instead of duplicating the input I just apply the operations sequentially (i.e. multiplexing them).Malicious
T
2

itertools.tee is well-suited for single pass iterables which are ubiquitous in Python. For instance Generators are single pass, and they are often used.

But if you already have list/deque you will not use itertools.tee for it, because it would involve superfluous duplication - you can just iterate over original list/deque over and over again.

C++ also has concept of single pass ranges, for instance Input Iterator, but they are not so ubiquitous. It is consequence of another set of aims of typical C++ program - give to user maximum as possible maintaining best performance. It is another mindset if you wish.

To illustrate this let's compare boost::transformed and itertools.imap (or generator expressions):

They both provide view of input sequence via given "prism". itertools.imap returns single pass iterable, while boost::transformed returns range view which has same category as input range - i.e., if you would pass Random Access Range as input you would get Random Access Range as the result.

Another fact is that C++ employs value semantics by default, while python has pointer semantics. It means that if copy iterator in C++, and "bump" it several times - original iterator will not be changed (though it can be invalidated if it is single pass range, but it is not the point).

But, sometimes you do want to accumulate values from single pass range and look at them several times. In such case, most common solution is to accumulate values to some container explicitly, by hands. For instance:

vector<int> cache(first,last);

However, tee-like wrappers are still possible in C++, here is proof-of-concept. Usage is:

auto forward_range = tee_range(first,last);

tee_range takes single pass range as argument and returns forward range (which is multi-pass) (there is also make_tee_iterator, which works at iterator level). So, you can take copies of that range and iterate it several times:

auto r = forward_range;
auto out = ostream_iterator<int>(cout," ");
copy(forward_range,out);
copy(forward_range,out);
copy(r,out);

Therer is also improvenment over itertools.tee - internally, only one deque is used to cache values.

live demo:

#include <boost/range/adaptor/transformed.hpp>
#include <boost/iterator/iterator_facade.hpp>
#include <boost/smart_ptr/make_shared.hpp>
#include <boost/range/iterator_range.hpp>
#include <boost/smart_ptr/shared_ptr.hpp>
#include <boost/container/vector.hpp>
#include <boost/container/deque.hpp>
#include <boost/range/algorithm.hpp>
#include <algorithm>
#include <iterator>
#include <cassert>
#include <limits>

template<typename InputIterator>
class tee_iterator : public boost::iterator_facade
    <
        tee_iterator<InputIterator>,
        const typename std::iterator_traits<InputIterator>::value_type,
        boost::forward_traversal_tag
    >
{
    typedef typename std::iterator_traits<InputIterator>::value_type Value;
    typedef unsigned Index;
    struct Data
    {
        boost::container::deque<Value> values;
        boost::container::vector<tee_iterator*> iterators;
        InputIterator current,end;
        Index min_index, current_index;
        Index poped_from_front;
        //
        Data(InputIterator first,InputIterator last)
            : current(first), end(last), min_index(0), current_index(0), poped_from_front(0)
        {}
        ~Data()
        {
            assert(iterators.empty());
        }
    };
    boost::shared_ptr<Data> shared_data;
    Index index;
    static Index get_index(tee_iterator *p)
    {
        return p->index;
    }
public:
    tee_iterator()
        : index(std::numeric_limits<Index>::max())
    {}
    tee_iterator(InputIterator first,InputIterator last)
        : shared_data(boost::make_shared<Data>(first,last)), index(0)
    {
        shared_data->iterators.push_back(this);
    }
    tee_iterator(const tee_iterator &x)
        : shared_data(x.shared_data), index(x.index)
    {
        if(shared_data)
            shared_data->iterators.push_back(this);
    }
    friend void swap(tee_iterator &l,tee_iterator &r)
    {
        using std::swap;
        *boost::find(l.shared_data->iterators,&l) = &r;
        *boost::find(r.shared_data->iterators,&r) = &l;
        swap(l.shared_data,r.shared_data);
        swap(l.index,r.index);
    }
    tee_iterator &operator=(tee_iterator x)
    {
        swap(x,*this);
    }
    ~tee_iterator()
    {
        if(shared_data)
        {
            erase_from_iterators();
            if(!shared_data->iterators.empty())
            {
                using boost::adaptors::transformed;
                shared_data->min_index = *boost::min_element(shared_data->iterators | transformed(&get_index));
                Index to_pop = shared_data->min_index - shared_data->poped_from_front;
                if(to_pop>0)
                {
                    shared_data->values.erase(shared_data->values.begin(), shared_data->values.begin()+to_pop);
                    shared_data->poped_from_front += to_pop;
                }
            }
        }
    }
private:
    friend class boost::iterator_core_access;
    void erase_from_iterators()
    {
        shared_data->iterators.erase(boost::find(shared_data->iterators,this));
    }
    bool last_min_index() const
    {
        return boost::count
        (
            shared_data->iterators | boost::adaptors::transformed(&get_index),
            shared_data->min_index
        )==1;
    }
    Index obtained() const
    {
        return Index(shared_data->poped_from_front + shared_data->values.size());
    }
    void increment()
    {
        if((shared_data->min_index == index) && last_min_index())
        {
            shared_data->values.pop_front();
            ++shared_data->min_index;
            ++shared_data->poped_from_front;
        }
        ++index;
        if(obtained() <= index)
        {
            ++shared_data->current;
            if(shared_data->current != shared_data->end)
            {
                shared_data->values.push_back(*shared_data->current);
            }
            else
            {
                erase_from_iterators();
                index=std::numeric_limits<Index>::max();
                shared_data.reset();
            }
        }
    }
    bool equal(const tee_iterator& other) const
    {
        return (shared_data.get()==other.shared_data.get()) && (index == other.index);
    }
    const Value &dereference() const
    {
        if((index==0) && (obtained() <= index))
        {
            shared_data->values.push_back(*(shared_data->current));
        }
        assert( (index-shared_data->poped_from_front) < shared_data->values.size());
        return shared_data->values[index-shared_data->poped_from_front];
    }
};

template<typename InputIterator>
tee_iterator<InputIterator> make_tee_iterator(InputIterator first,InputIterator last)
{
    return tee_iterator<InputIterator>(first,last);
}

template<typename InputIterator>
boost::iterator_range< tee_iterator<InputIterator> > tee_range(InputIterator first,InputIterator last)
{
    return boost::iterator_range< tee_iterator<InputIterator> >
    (
        tee_iterator<InputIterator>(first,last),
        tee_iterator<InputIterator>()
    );
}
// _______________________________________________________ //

#include <iostream>
#include <ostream>
#include <sstream>

int main()
{
    using namespace std;
    stringstream ss;
    ss << "1 2 3 4 5";
    istream_iterator<int> first(ss /*cin*/ ),last;
    typedef boost::iterator_range< tee_iterator< istream_iterator<int> > > Range; // C++98
    Range r1 = tee_range(first,last);
    Range r2 = r1, r3 = r1;
    boost::copy(r1,ostream_iterator<int>(cout," "));
    cout << endl;
    boost::copy(r2,ostream_iterator<int>(cout," "));
    cout << endl;
    boost::copy(r2,ostream_iterator<int>(cout," "));
}

Output is:

1 2 3 4 5
1 2 3 4 5
1 2 3 4 5 

Boost.Spirit has Multi Pass iterator which has similar aims.

The multi_pass iterator will convert any input iterator into a forward iterator suitable for use with Spirit.Qi. multi_pass will buffer data when needed and will discard the buffer when its contents is not needed anymore. This happens either if only one copy of the iterator exists or if no backtracking can occur.

Thais answered 17/3, 2013 at 21:46 Comment(0)

© 2022 - 2024 — McMap. All rights reserved.