Jeff Garland wrote:
Sohail Somani wrote:
Oh gosh yes. Having things in headers because it makes it 'easier to compile' hurts compile times in the long run.
Get a faster build machine and/or use pre-compiled headers -- really.
Its just not good practice and I suspect a non-empty subset of new boost users are also new to C++.
5 years ago I believed it was a serious issue too -- I've substantially changed my mind. And not in small part by using asio (and other boost libs) as header only. It's extremely nice to not need to figure out how to build and link. You can get started with so trivially with an all header library -- and I think that frequently it's trouble getting over that first hump that stops people from using libs. So if the barrier is lower more people will use the libs.
Sorry, but I disagree. The effect of header only libraries on build times is still unacceptable and not everybody can convince their bosses that they need more powerful machines every six months or so. One of the mistakes I came to regret most in the project I'm working on is to have a very small template metaprogram in a header that's used just about everywhere. But I learned my lesson and now I'm happy that I hid Boost.DateTime behind a class of my own. Note that we're using Borland's compiler, which is considered a very fast one and that we do use pre-compiled headers; this, by the way, are not a panacaea, among other things because a sizable portion of your savings in sheer compilation time is lost in much increased dependency checking. As to binary libraries stopping people using Boost, I don't think that's much of a problem per se: after all people programming in C++ and C have been accustomed to static and shared libraries for ages. Rather what may constitute a problem is the complexity and general unfriendliness of the Boost Build system. Still, the time taken to find out how to build libraries for one's own use is mostly spent once, and can be delegated to a single team member; excessive build times impact the whole team all the time. One instructive exercise is to run g++ -H -P -E on a C++ source file chosen at random; it doesn't even need to use Boost, it's sufficient to include, say, <vector> on Windows. The resulting tree of included headers is disenheartening and a clear symptom of what I consider a very serious problem. Cheers, Nicola Musatti