Re: [boost] [metaparse] Practical usefulness

Subject: Re: [boost] [metaparse] Practical usefulness
From: Niall Douglas (s_sourceforge_at_[hidden])
Date: 2015-06-02 21:42:01

On 2 Jun 2015 at 17:42, Michael Caisse wrote:

> > I think back before the C++ 11 STL the question of usefulness was
> > more important than now as Boost becomes more heterogeneous. I note
> > how many reviewers of Metaparse said they are not experts in that
> > field, or would even expect to use Metaparse any time soon
> > themselves.
> 
> Have you ever heard or seen a library like metaparse?

To be honest I had always assumed the concept wasn't feasible as a 
practical library. I was very glad to learn I was wrong.

> None of us are experts in the field because I think it is new. 

This is ancillory to the argument that it has niche use cases, and 
that new C++ libraries have a trend of becoming more specialised and 
less general.

> I also think this is exactly the type of library that Boost has been 
> known to promote in the past. It is like MPL or Fusion. I'm excited to 
> see something like it again.

I am also positive on the library Michael. I had been hoping to give 
it an unconditional yes vote in fact, all the boxes were ticked for 
me except the silly easy to fix ones.

I think though there is an enormous difference in the C++ field of 
practice today compared to pre-Fusion. Before Fusion few of us had 
the wisdom to not push C++ as far as it could go. Now we've all been 
burned trying to use these fancy libraries in big real world 
applications, few of us are itching to repeat that experience. There 
has been a big shift back towards simpler is better, and you probably 
saw some of that in the recent argumentation about next gen futures 
where there was a faction who felt nothing is fundamentally wrong 
with present futures, and a faction who think them too big for what 
they ought to be. And that's just futures, which are ridiculously 
simple compared to anything Eric or you are writing every day, though 
I've noticed arguments about library primitives are often the most 
vicious.

It's a big reason I always ask about scalability benchmarking for 
compile times and compiler memory use for the major compilers for TMP 
libraries. Too few do what Louis did before writing a single line of 
finished library where he benchmarked every TMP technique for time 
and space so he knew what not to do.

Louis should really write those TMP technique benchmarks into a book 
and publish it and make some money for himself. I'm sure that Pearson 
editor lady I was talking to at C++ Now would consider such a book, 
if not I have some tech publisher who reached out to me not too long 
ago somewhere in my records.

Niall

-- 
ned Productions Limited Consulting
http://www.nedproductions.biz/ 
http://ie.linkedin.com/in/nialldouglas/