Re: [Boost-users] filesystem headers

From: Emil Dotchevski (emil_at_[hidden])
Date: 2008-05-29 20:37:10

On Thu, May 29, 2008 at 1:45 PM, John Femiani <JOHN.FEMIANI_at_[hidden]> wrote:
>> Emil Dotchevski wrote:
>> In my opinion it is pointless to measure compile times.
>> Instead, headers should be designed in a way that allows user
>> code to not "see"
>> anything that is not strictly necessary to compile a
>> particular user cpp file.
>
> So are you recommending that path should not be part of a public
> interface, since a path can be represented by a string?
>
> Or are you saying that it is not worth measuring the cost of all-header
> libraries that include too much per header since they should be
> refactored anyways?

I am sorry, pointless was perhaps too strong of a word there. My point
is that while my concerns do have something to do with compilation
times, this is somewhat secondary to my desire (in principle) to lower
the level of coupling (in boost::filesystem in particular) even if it
had no effect on compilation times.

-- 
Emil Dotchevski
Reverge Studios, Inc.
http://www.revergestudios.com/reblog/index.php?n=ReCode