kicad-developers team mailing list archive
Mailing list archive
Re: RS274X file format.
Wayne Stambaugh <stambaughw@...>
Sat, 02 May 2009 08:47:02 -0400
Thunderbird 220.127.116.11 (Windows/20090302)
Dick Hollenbeck wrote:
> Dick Hollenbeck wrote:
>>> This is why I have been using wxList and wxArray instead of C++ list and
>> I don't share your enthusiasm for these two containers.
>> There is little standard about C++ if <vector> is not.
>> In fact I would even disagree with you. wxString is reasonable because
>> it is unicode and work with the gui functions. Beyond that using
>> standard C++ containers is entirely safe and likely gives you an easier
>> path to a different GUI library. Say somebody wanted to go to QT
> Well I came back here and see that I said nothing about C++ list. If
> the object being put on a list is an object derived from our
> base_struct, then we have our DLIST<> stuff and that is the list type to
> be used.
The code I fixed in CVPcb was not base_struct derived. It was yet
another previous/next pointer implementation that was handled by inline
code. I am pretty sure I have seen some code in Eeschema that follows
the same pattern.
> C++ list actually requires two memory allocations for each node so I
> *never* use it, ever. In fact I never use any linked list framework
> which requires more than a single memory allocation per node. Vector
> and the boost ptr_vector are way better than C++ list for non
> base_struct derived object containers.
I'll take a look at Boost ptr_vector and C++ vector as a replacement for
> For anything of complexity, ptr_vector wins the day, and it is not even
> a close call, because all such instantiations share common code based on
> std::vector<void*>, and the object can be destroyed automatically via
> pointer only at container destruction time. Copy constructors don't
> enter in because only a pointer is in the container, not the actual
> object copy.
Thanks for the information. It saves me a lot of time figuring it out
on my own.