TITLE: another interpretation of mixin inheritance


PROBLEM: <4am53j$eht@netlab.cs.rpi.edu>

A mixin is an abstract class that represents a specific interface and is 
used to build up larger interfaces. Take for example the interfaces
'ReadAble' and 'WriteAble' represented by the classes:


RESPONSE: stuart@bmsi.com (Stuart D. Gathman), 20 Dec 95

I guess the definition varies.  What you call mixins, I have always
called multiple inheritance - objects that comform to more than one
interface.

What I call mixins are partial implementations of a single interface.  You
can quickly build a complete object without writing any methods by combining
packages of prewritten methods.

This explanation and example is much simpler than Max Skallers (though
not exploring all the nuances), so I believe that c.l.c++.m readers may
benefit.

[ I'll repost some of Skaller's stuff in the near future. -adc ]

E.g.

  // interface
  struct SerialInterface {
    int read(char *buf,int) = 0;
    int write(const char *buf,int) = 0;
  };		

  // partial implementations
  struct BufferedRead: virtual public SerialInterface{
    int read(char *buf,int);
  };			

  struct PolledRead: virtual public SerialInterface{
    int read(char *buf,int);
  };			

  struct BufferedWrite: virtual public SerialInterface{
    int write(const char *buf,int);
  };			

  struct PolledWrite: virtual public SerialInterface{
    int write(const char *buf,int);
  };			

  // finally, the application creates a customized SerialInterface
  // without writing any code:

  struct MyComm: BufferedRead, PolledWrite { };

  MyComm com1;

  com1.read(buf,buflen);	// reads are interrupt driven
  com1.write(buf,buflen);	// writes are polled