Boost logo

Boost :

From: Andy Little (andy_at_[hidden])
Date: 2004-07-23 05:52:52


"Arkadiy Vertleyb" <vertleyb_at_[hidden]> wrote
> "Andy Little" <andy_at_[hidden]> wrote
>
> I don't think you really want lvalues. You would only want to preserve
> lvalue if you want to do something like:
>
> a /b = 5;
>
> and I don't think that is what you want.

Ah yes ... I am ever ignorant of the what the term 'l_value' means.
Only thing I can think of is stream output via <<, May be wrong but
ostream& is then an l_value?
Whether that causes problems I have no idea.

Whatever. I would hope it tries to get as close as it can to the decltype
and auto proposals as possible, because..
bearing in mind that it may quite cumbersome to use in larger programs, I
still see it as being very useful for demos and experimenting with
ie to say "if we had decltype or whatever, you could do this, this and this.
" . as a prelude to the real thing.
That anyway is how I would probably use it.( And in fact would find it very
useful to "look into the future"). I would tend to avoid it in larger
programs due to compile times. (Only because other libraries drags things
down enough already). I think spirit and lambda already have their own type
deduction schemes, making it easy to implement there. IOW It might be useful
to grab another library ie UBLAS etc and try to implement the mechanism
there to find/ show what problems other users may have in implementing it.

The "User interface" seems to be the part that is lacking at the moment...
but I guess thats old news.

regards
Andy Little


Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk