C++Talk.NET

Ulrich Eckhardt · Guest

Fellows,

I think the topic itself is not something new and there are already
containers made for storage of polymorphic types (e.g. the one in Boost)
but there I had the idea of a different approach: instead of wrapping
pointers and cloning objects on demand, I'd rather drop the ability to
insert object.
Sounds strange? Yes, maybe the idea is not useful but please read on, I must
also confess that I've had lots of coffee today and that this is a
braindump and not the result of proper research.

The idea is that you create objects inside the container instead of creating
them externally and copying them into the container. For the particular
cases I have in mind (i.e. my work) this is all I need (yes, apart from
destroying and iteration). Now, the difference is that you can't detach
them from the container. You can also make them non-copyable, even
nonclonable (ever wanted a list<ostream> for logging?). Of course, this is
only feasible for node-based containers (linked lists, trees), creating a
polymorphic vector that still guarantees contiguous storage is a)
unnecessary (you only need the contiguous for array replacement, and there
are no polymorphic arrays) and b) would only be possible if you knew the
size of the objects in advance, which you don't.

Now, how would this be used?

typedef polymorphic_list<ostream> streams;

streams s;
// create an object of type 'ofstream' and pass "fou.txt" to its ctor
// considering the varying amount of possible arguments create()
// needs to be overloaded
s.create<ofstream>("fou.txt");

// create new object at the beginning
// this will probably be a bit tricky because one needs to
// distinguish between an iterator and an argument to the ctor
s.create<ostringstream>( s.begin());

streams::iterator it = s.begin();

// '*it' yields an ostream&
*it << "Hallo monde!";

// destroy the last element
s.erase(*it);

Credits: this was inspired by someone that strongly opposed the use of
containers because of their overhead while copying object into them and out
of them. I forgot who it was, the only things I remember was that I
otherwise didn't agree with that person at all but that it inspired me to
think of how it would be like when this transferring into and out of was
removed from containers.

OK, so much for now, back to work...

Uli

[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]

Thant Tessman · Guest

Ulrich Eckhardt wrote:

[...]

Mirek Fidler · Guest