Andreas Raab wrote:
> PS:
> 
>>if you want to 
>>optimize then choose a central, high-level, intent revealing 
>>place (such as, in fact, SortedCollection>>addAll:) and don't 
>>ever have the client second-guessing the structure and 
>>behavior of the object in question.
> 
> 
> A good example of what I mean can be found if you browse the
> implementors of #atAllPut: - it contains specific optimizations for
> polymorphic receivers and you simply can not get any faster by "client
> side optimizations" (e.g., writing out the loop or similar). But such a
> critical place has to be identified first and if everyone is
> "optimizing" the code locally you will never find the bottleneck to
> begin with.

Agreed.

It is worse for "client side optimizations": they may be penalized regarding
speed or more worse *broken* (if using private (in a semantically sense)
methods) by making optimizations at the correct places...


Greetings,

Stephan

> 
> Cheers,
>   - Andreas
> 
> 
> 


-- 
Stephan Rudlof (sr at evolgo.de)
   "Genius doesn't work on an assembly line basis.
    You can't simply say, 'Today I will be brilliant.'"
    -- Kirk, "The Ultimate Computer", stardate 4731.3