Re: [Boost-users] Boost.mpl suggestion: variation of find_if

From: David Abrahams (dave_at_[hidden])
Date: 2005-06-15 11:40:30

Vyacheslav Kononenko <groups_at_[hidden]> writes:

> --- Aleksey Gurtovoy <agurtovoy_at_[hidden]> wrote:
>
>> Vyacheslav Kononenko writes:
>> > I could not come with usable name for such thing
>> but it could be
>> > usable anyway.  So it is like find_if but applies
>> a function to an
>> > sequence and return first non void type that
>> function returned. So
>> > implementation would be trivial:
>> >
>> > template< typename Sequence, typename Func >
>> > struct find_type : fold< Sequence, void_, if_<
>> > is_void_< _1 >, Func< _2 >, _1 > {};
>> 
>> Or, which is likely to be more efficient,
>> 
>>     template< typename Seq, typename F > struct
>> find_type
>>     {
>>         typedef typename find_if< 
>>               transform_view<Seq,F>
>>             , is_not_void_<_1>
>>             >::type iter_;
>> 
>>         typedef typename eval_if< 
>>               is_same< typename iter_::base,
>> typename end<Seq>::type >
>>             , void_
>>             , deref<iter_>
>>             >::type type;
>>     };
>
> Correct my if I am wrong but your will invoke F on
> every element when mine will not.

transform_view is lazy, so F will only be invoked until it returns a
non-void_ result.

>> > If it can be implemented by current set of
>> algorithms
>> > please point me out.
>> 
>> Doesn't your sketch above employ an algorithm from
>> the current set
>> ('fold')? In any case, at the moment this feels too
>> specialized to me
>> to be considered for inclusion in the library. At
>> the very least, it
>> needs a name ;).
>
> There is another topic about mpl for_each and I think
> my algo would be usable there. But for the name it is
> an author work to make names isn't it?

You appear to be the author of this algorithm ;-)

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com