Thread: When to use private Inheritance?

Originally Posted by Bargi

My question is that, in what circumtances we should only use private inheritance?

Pretty much never. It's a sign that you're trying to implement a "has-a" relationship via an "is-a" relationship which is almost never a correct design.

There is an example on the Cprogramming article board which ends with the statement "If you need this kind of functionality, then private inheritance is the only solution to the problem!" Well, that's nonsense. There is no problem which can only be solved by private inheritance.

Using private inheritance might save you from writing a certain amount of code, in some cases, but that's hardly a good basis for a design.

I've never had a reason to even think about using it in the last 10 years of C++ programming.

Originally Posted by brewbuck

Pretty much never. It's a sign that you're trying to implement a "has-a" relationship via an "is-a" relationship which is almost never a correct design.

Private inheritence is used more often for "implemented using" relationships than for "has a" relationships.

In terms of the original question, I agree there are no cases where one should use private inheritence. Private inheritence is, invariably, one option among several that can address some need. The choice of which option to use should generally be one that represents a rational trade-off for the intended application.

If the original question concerned when one might use private inheritence, the answers would probably show some more variation.

Originally Posted by Bargi

I am learning about the private inheritance and found a little bit confusing.
In all other type of inheritance we can reuse the code in derived class but not in case of private( unless we apply some technique to use them).

The way that I see it, you can reuse the code in the derived class with private inheritance. In fact, Scott Meyers' opinion is that "private inheritance means is-implemented-in-terms-of", and is "purely an implementation technique", although it is a technique that can pretty much always be replaced by composition.

This is in contrast to public inheritance ("is-a"), in which one would inherit not only to reuse code (which might not happen at all if the base class is a pure interface), but so that the derived class can be used by existing code that is written for the base class and its derived classes, with no or minimal modification to existing code.

Originally Posted by EVOEx

Well, I've used protected inheritance quite often. Of course, not remotely as often as public inheritance. But yeah - I've never used private at all.

Out of curiosity, but why did you use protected inheritance? Did you intentionally choose it over private inheritance and composition, and if so, what were your reasons?