|
Boost : |
From: David Abrahams (dave_at_[hidden])
Date: 2005-05-22 21:01:37
"Arkadiy Vertleyb" <vertleyb_at_[hidden]> writes:
> "David Abrahams" <dave_at_[hidden]> wrote
>
>> Hmm, I'm no longer certain of that. It's at least LIMIT_SIZE * M,
>> where M is the number of distinct result types of size().
>
> OK, since we are not certain which one is better, I prefer the simplest one
> (the one you suggested first).
Well, there are two ways to get an idea of which is better:
1. Decide if size() will return the same value for multiple different
types. If so, it's likely the 2nd one is better. If not, it's
likely the 1st one is.
2. Time it ;-)
> Since, as you mentioned, somebody at Microsoft is looking at this
> (thanks a lot, BTW), it might not matter after all.
It's very close to the wire for them. There's no guarantee they'll
make it, FWIW.
> For now this one would do, and, if we have to live with it in the
> long run, we can see what can be done to optimize it.
>
> I uploaded the new version, take it at usual place,
> http://boost-sandbox.sourceforge.net/vault/, typeof.zip -- the
> problem should be fixed, and I also added similar code to the tests.
Thanks. I should probably download it and try that one eventually,
but I am currently attacking other problems in my code using my 2nd
implementation.
> I havn't changed anything else -- I think other things, like
> removing _TPL, etc., can (and should) wait until after the review...
Makes sense.
-- Dave Abrahams Boost Consulting www.boost-consulting.com
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk