Re: [boost] Re: MPL containers and algorithms

From: Fernando Cacciola (fcacciola_at_[hidden])
Date: 2002-07-31 09:11:52

----- Original Message -----
From: "Paul Mensonides" <pmenso57_at_[hidden]>
To: <boost_at_[hidden]>
Sent: Wednesday, July 31, 2002 2:36 AM
Subject: Re: [boost] Re: MPL containers and algorithms

> [snipped]
>
> Yes, but with the STL there are specific needs for specific container
types such
> as access time, insertion time, memory use, etc..  Various types of
optimization
> for various needs.  These have come about over decades of study, and the
need
> for them proven in the field.

OK.

>  These same arguments do not apply here.
Why?

>  I don't
> think that we need multiple *sequence* types.

Fair enough, but you should notice that your entire argument seems to based
in the axiomatic statement that I remarked above.
IMO, the bottom line of this discussion is that Andrei and you think that
compile-time programming differs from run-time programming so much that, as
you said, the conventional run-time needs (such as efficient random access
time, head-tail appending time, insertion time, permutation time, etc...)
don't apply to compile-time.

You have to prove, or at least, explain a lot more deeply, this point
precisely. Why different sequence-types are undoubtly useful in run-time,
but not in compile-time? It all boils down to this.

BTW:
I haven't got the time the actually use MPL during the review.
Based on the paper and the examples -which I looked closely- I vote to
accept the library.
I haven't read the reference documentation, though; only the paper
I'm quite familiar with metaprogramming, so I couldn't see the paper from
the perspective of a brand new user, but from my background I found it very
very good!

Fernando Cacciola
Sierra s.r.l.
fcacciola_at_[hidden]
www.gosierra.com