Hi all,
Keep forgetting to write this mail for a while now. It is intended mostly for
(in aplhabetical order) Aaron, Marco, Martin and Sebastian, but everybody is
invited to chime in.
It is about the smart and safe d-pointer I have presented some time ago [1].
The question is should it be used in plasma and where?
Cons:
- too low level to be in plasma-frameworks (sebas)
- none else that I know of
Pros:
- safety: no access to the raw pointer
- safety: no accidental initialization errors or anything similar
- safety: no possible leaks
- convenience: forwarded constructor arguments (': d(1,2)' instead of ':
d(new Private(1,2))')
- convenience: default constructor works for no-arg Private constructor
(nothing instead of 'd(new Private())')
- convenience: no delete d;
- they are spiffy (aaron) :)
Choices:
1. (a non-choice imo) Use raw pointers like in the old days everywhere
2. Use less safe and less convenient smart-pointers like std::unique_ptr or
QScopedPointer
3. Use this d-ptr outside of the library, choose 1 or 2 for the library
4. Use it everywhere, but export the headers as if it were 1 or 2 (all should
be binary compatible with a raw pointer)
5. Use it everywhere, and install the header file for the users to be able to
use it as well. (it is a templated class, so it leaves no binary footprint by
itself)
Cheerio,
Ivan
[1] http://ivan.fomentgroup.org/blog/2013/06/22/d-ptr-the-modern-way/
--
Science gathers knowledge faster than society gathers wisdom.
-- Isaac Asimov
_______________________________________________
Plasma-devel mailing list
Plasma-devel@kde.org
https://mail.kde.org/mailman/listinfo/plasma-devel
