Pimpl-ization

[Jan Holesovsky]

Hi Kohei,

Kohei Yoshida píše v St 10. 12. 2014 v 21:09 -0500:

> My belief was that this was not a controvercial change, but it appears
> that it is.  If there is a strong objection I'll stop doing it.  If not,
> I'd like to continue pursing this since it seems to improve both of the
> 2 aforementioned points, and I've only covered perhaps 10-20% of all
> possible candidates.

I don't have objection if _you_ are doing that :-)  It's fine for me
when it is targeted, focused on the goals you've outlined.

But I did not know the reasons for pimpl-ization previously, and my fear
was that this might become a favorite Easy Hack where we'd have loads of
people pimpl-izing just everything; which was my concern from 2 reasons:

* debuggability / code reading perspective: one more level of
  indirection that you need to go through to see what's going on

* cost of new/delete of the Impl class + cost of the pImpl-> calls

If it is not going to become such an Easy Hack, I am OK with that of
course.

All the best,
Kendy