From: Tobias Reif <tobiasreif@...>
Date: 2001-09-04T07:29:22+09:00
Subject: [ruby-talk:20819] Re: RCR: non-bang equivalent to []=

Hi Dave,

> Being realistic, though, wouldn't you expect
> 
>   a[1] = 2
> 
> to modify the object referenced by 'a'. 


[]= may not appear in a new concept+convention; it could stay like it 
is, and behave like everyone knows.
A new concept+convention may introduce its equivalent.

[...]Similarly for 'delete' and
 > company.

Absolutely; there *are* exceptions that make sense. As David Black said: 
#max, #first, #last sure don't make a copy.
What exactly their name is; which characters are appended to their name 
or not, is just a detail. They probybly wouldn't have both versions.


> There _are_ issues with '!', particularly to do with return values
> when the receiver isn't changed, but I don't think you should take the
> '!' as a part of the semantic content of the method being called. It
> really is no more than a hint that something "dangerous" is going on.


Any stringent concept plus semantically rich naming convention would 
help to determine how a method acts in this aspect, thus improving the 
Principle Of Least Surprise, especially when working with methods out of 
the thousands of classes and libraries that are and will be written.

Tobi

-- 
Tobias Reif
http://www.pinkjuice.com/myDigitalProfile.xhtml

go_to('www.ruby-lang.org').get(ruby).play.create.have_fun
http://www.pinkjuice.com/ruby/