[BUG]Collection>>removeAll:

Richard A. O'Keefe ok at cs.otago.ac.nz
Thu Aug 29 02:03:03 UTC 2002


Stephan Rudlof <sr at evolgo.de> wrote:
	Alternative: Use of a value holder (preserving identity) holding the
	collection and the dependency mechanism.
	
I really don't like the attitude I think I am seeing here:

    It is a good thing to push arbitrary amounts of complexity
    onto programmers instead of fixing a method in a collection class.

We now know that if OrderedCollection>>removeIndex: were more efficient
(so that oc removeAll: oc copy took linear time instead of quadratic time)
there needn't be any hole in the coverage of OrderedCollection>>removeAll:
_even_without_changing_one_single_character_of_removeAll:_itself.

This completely disposes of the argument that programmers could be expected
to know about the hole in coverage.

So it is clearly unreasonable to expect Smalltalk programmers to make
_their_ code complicated to compensate for a bug which may very possibly
have been introduced in 2000 and is not shared by all Smalltalks and
whose correction could make the system _more_ efficient.




More information about the Squeak-dev mailing list