• i_don_t_know 2 months ago

    The implementation forwards unknown messages to the first mixin that implements the message, if any. I think forwarding messages through #perform: is very much slower than sending messages directly.

    Are there Smalltalk implementations that optimize #perform:? How? Are there implementations that look for forwarding patterns of #doesNotUnderstand: + #perform:, and that add the destination to the inline cache of the original send?