[boost] Re: overloads: request for discussion

From: Aleksey Gurtovoy (agurtovoy_at_[hidden])
Date: 2005-01-28 02:16:26

Alexander Nasonov writes:
>> > What if you have 400? Can mpl::vector or mpl::set contain 400
>> > elements?
>> 
>> 'mpl::set' definitely can. Structurally, it's not very different
>> from an "overload set".
>
> I wonder how to create such a huge set. By inserting elements 400
> times?

That would be one way.

> How much compilation time would it take then?

Depends on the compiler. On my G5, CodeWarrior 9.2 swallowed the
following in ~15 sec:

    typedef copy<
          range_c<int,0,400>
        , inserter< set<>, insert<_1,_2> >
        >::type r;

>
>> > I'm afraid not. Some overloads library work even on higher numbers
>> > (up to 700 on my laptop).
>> 
>> Interesting. How do you generate these in the first place?
>
> typedef overload_set<overloads, range_c<int,1,701> > huge_set;
> struct overloads contains 700 call operators.

Hand-written?

>
> You can't do much with it. Only special algorithms that is designed with
> this number in mind and which are supposed to return a moderate amount
> of element. For example, typical FSM may have 20 states and 30 events,
> that is, up to 600 transitions. First algorithm to apply to an overload
> set of transitions is searching for unique states. It should return an
> overload set with size 20 where each signature contains unique state.
> 20 is moderate number and can be represented by vector20.
>
> This is how this algorithm can transform transitions into overloads
> with unique states inside:
>
> in : overload_set<overloads, range_c<int,1,601> >
> out: overload_set<overloads, vector20< /*...*/> >
>
> Something similar to has_key would be nice too. I have one idea how to
> imlement fast detection of a given signature but I have to check it.
>
>> 
>> > 400 is not a random number. Imagine, you have 20 classes in Shape
>> > hierarchy. Intersestion of all of them requires 20x20=400
>> > has_intersection functions.
>> >
>> > BTW, I remember Dave mentioned that you generate FSM using mpl::vectorN,
>> > with N higher then default 50. How high is N?
>> 
>> ~70, IIRC.
>
> Is it a number of transitions?

Yep.

-- 
Aleksey Gurtovoy
MetaCommunications Engineering