Re: [boost] Re: MPL containers and algorithms

From: Peter Dimov (pdimov_at_[hidden])
Date: 2002-08-01 10:30:13

From: "Aleksey Gurtovoy" <agurtovoy_at_[hidden]>
> Andrei Alexandrescu wrote:
> > I asked for concrete, real-world examples, of tasks that suck with
> > s-expressions and are stellar with something else. (To wit, with the STL
> you
> > can easily provide clear examples of tasks that are cool with vector and
> bad
> > with list.) Got no response. I am not convinced.
>
> Let me remind you that this list is not a "real-time" chat forum. People
who
> contribute here live all over the world, in different time zones, and,
> finally, have other responsibilities besides answering someone's post in a
> rush mode.

I wouldn't want to interrupt your wonderful discussion ;-) but
++not_convinced;

There are some places where the multiple sequence approach doesn't ring true
to me. For example, if I make a sequence out of two iterators by using
iterator_range, does this produce a vector if the inputs are vector
iterators? Or in general, when a function returns a sequence, what are the
properties of this sequence?

This is not to say that MPL should drop the multiple sequence types right
away... but I think that you should keep an open mind, analyze the common
use patterns, and if the multiple sequences turn out to not provide the
advertised benefits (Dave's example of O(1) list access), or if nobody uses
anything but list... unnecessary complexity should go.