Without D1-style operator overloads, the `in` operator for classes doesn't make sense

Don prosthetictelevisions at teletubby.medical.com
Mon Jan 6 10:55:15 UTC 2020

In DMD 2.088.0, the D1 operator overloads were deprecated, in 
favour of the D2 syntax which has been available for years.

For example, opAdd_r is replaced with opBinaryRight(string 
op)(...) if (op == "+")

But this is not purely a syntactic change. Some functionality is 
lost: the D2 syntax is a template; and therefore, it is not a 
virtual function.
It was argued that using virtual functions for operators is an 
obscure feature which hardly anybody would use. So it is not a 
significant loss. And I agree with that, *for the operators which 
D inherited from C*.

Unfortunately, this argument is not valid for two operators which 
are unique to D: opCat and opIn. The problem is most acute with 
the `in` operator.

Suppose you make an object-oriented container class library.
Let's say you have a Set. This may have derived classes HashSet, 
BloomFilterSet, etc.

One of the most basic operations is asking if an element is a 
member of the set. For sure you would want to use the `in` 
operator. But you can't, because it isn't virtual.

So, I think we've got ourselves into a situation which doesn't 
make sense. It's pretty much always a mistake to define a 
D2-style 'in' operator for a class. If you are using classes, you 
would want it to be virtual; if you don't need it to be virtual, 
you shouldn't be using classes.

So I feel that by introducing this deprecation, we've dropped 
support for OOP container classes.
If this is part of a grand conspiracy to remove classes from D, I 
won't stand in the way :) But apparently we've done this by 
accident, which is rather alarming.

More information about the Digitalmars-d mailing list