|
Boost : |
From: Beman Dawes (bdawes_at_[hidden])
Date: 2001-11-28 08:29:07
At 06:59 PM 11/27/2001, David Abrahams wrote:
>
>----- Original Message -----
>From: "Jonathan H Lundquist" <jhl_at_[hidden]>
>
>
>> Having implemented Andrei's typelists in MSVC, I can state without
>> reservation that a working version without partial specialization was
not
>> significantly more complicated than the original.
>>
>> More importantly, I could get my mind around the problem in order to do
>so.
>> Don't know if I could with mpl or not, but the 127 files asserted by
>Andrei
>> certainly sounds daunting...
>
>While some things in MPL are undeniably too complicated, we shouldn't be
>too
>impressed by the number of files. Every single algorithm/component is
>implemented in a separate file. Boost organization has evolved in that
>direction, and I think Aleksey was just anticipating the review comments.
While I understand the motivation of those who like to decompose headers
into many separate files, we are seeing here another unintended
consequence. It is causing the library to appear far more complicated that
it really is.
(The other problem is that new users include the wrong header, and get
confusing error messages as a result. The fix for that is to provide an
overall header which includes the others, but that doesn't address the
increased apparent complexity problem above.)
--Beman
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk