Re: ublas: vector<>::upper_bound()

From: jhrwalter (walter_at_[hidden])
Date: 2002-03-14 05:16:00

--- In boost_at_y..., Kresimir Fresl <fresl_at_g...> wrote:
> 
> jhrwalter wrote:
> 
> > May be upper_bound()/lower_bound() are bad names.
> 
> It seems that they are. At least, they are misleading.
> 
> > These member are essential for the iterators of binary operation 
> > expression templates: [...]
> 
> I think (hope) that I see what you mean: if
> ---------------------------------------------------------------
>      numerics::unit_vector<double> u3 (8, 3);
>      numerics::unit_vector<double> u4 (8, 5);
> 
>      cout << (u3+u4).begin().index() << "  "
>              << (u3+u4).end().index() << endl;
> ----------------------------------------------------------------
> is compiled with -DNDEBUG (which on g++ means
> `use expression templates'), the output is: `3  6',
> and when NDEBUG is not defined (ie., `don't use ETs'),
> output is `0  8'.

Correct, that's the intended behaviour ;-)
 
> > OTOH do you know a better name for the existing beasts?
>  > [lower_bound()/upper_bound()]
> 
> Hmm. Maybe simply `first()' and `last()'?

Ok, I'll change them to find_first()/find_last(), if you agree.
 
> >>[2] `find()', `lower_bound()' and `upper_bound()' are listed
> >>in interface of `numerics::vector<>' in docs, but I didn't find
> >>their description (AFAIK functions `find()' etc. are described
> >>only in docs for `compressed_array<>'; their descriptions are
> >>similar to those for `std::map<>').
> 
> > Yep. I'm not sure, if these members aren't an implementation 
detail 
> > only.
> 
> Well, I have no idea how to use them in my `real' code ;o).
> But I guess that they must be `public' (ie. part of the interface) 
so
> that expression templates can access them. 
 
I've never checked, if private access works with static polymorphism 
(under which compiler ;-) also, but I'll do that ASAP.
 
> Maybe you should mark
> them `For internal use only' in docs to prevent people like me
> to play with them ;o).
> 
> > Thanks for your effort.
> 
> Nothing at all. I hope that soon we'll have efficient and reliable
> vector/matrix library.
> 
> fres
 
Regards 
 
Joerg