Design question: where to create objects as result of a function

[Alex]

Hi everyone,

I have a question that implies thinking about some design issues with C
++.

I have a method like this:

class C {
....
public:
void get_something(std::vector<Something*>& results);
}

It returns a vector of objects that should be created (or filled in
with data) by the class C.

The problem is, if the Something objects are created inside
get_something, I will have to delete them later outside the function,
and I must avoid that for several reasons.

I don't know how many objects will be in the vector, so I can't create
the objects in advance and pass a vector of references to be filled in
by the function.

What is your advice for this kind of situations?

Thanks.

[Triple-DES]

On 29 Aug, 10:54, "Alf P. Steinbach" <[e-mail]:

* Alex:

And, also, I need to use references/pointers to Something, not
variables in the stack...

Why.

Wild guess: His class is a base class with virtual functions. Or does
not have value semantics.

In which case the "politically correct" advice is to return an
std::vector< boost::shared_ptr<T

DP

[James Kanze]

On Aug 29, 2:09 pm, Triple-DES <[e-mail]:

On 29 Aug, 10:54, "Alf P. Steinbach" <[e-mail]:

* Alex:

And, also, I need to use references/pointers to Something, not
variables in the stack...

Why.

Wild guess: His class is a base class with virtual functions.
Or does not have value semantics.

In which case the "politically correct" advice is to return an
std::vector< boost::shared_ptr<T

Which is usually bad advice. But without knowing the role of
the objects in his vector, we can only speculate. The most
likely "correct" solution is what Alf proposed: don't use
pointers at all. And just return the vector. Just about
anything else counts as a special case, which requires its own
special rules.

--
James Kanze (GABI Software) email:ja*********@gmail.com
Conseils en informatique orientée objet/
Beratung in objektorientierter Datenverarbeitung
9 place Sémard, 78210 St.-Cyr-l'École, France, [phone]

[Jerry Coffin]

In article <f656b761-4573-4c5f-842b-4b41466536c0
@k36g2000pri.googlegroups.com>, ac******@gmail.com says...

Hi everyone,

I have a question that implies thinking about some design issues with C
++.

I have a method like this:

class C {
...
public:
void get_something(std::vector<Something*>& results);
}

It returns a vector of objects that should be created (or filled in
with data) by the class C.

The problem is, if the Something objects are created inside
get_something, I will have to delete them later outside the function,
and I must avoid that for several reasons.

I don't know how many objects will be in the vector, so I can't create
the objects in advance and pass a vector of references to be filled in
by the function.

What is your advice for this kind of situations?

Have get_something take an iterator instead of a vector. Call it
something like:

std::vector<Somethingresults;

C c;
c.get_something(std::back_inserter(results));

Personally, I'd look at C (especially C::get_something) with an eye to
whether you can reasonably turn it into a functor to (for example)
produce the results one at a time, so you'd use it with a standard
algorithm.

--
Later,
Jerry.

The universe is a figment of its own imagination.