DP's are compressed in OO

Post #175,932 by tuberculosis 9/23/04 1:09:43 PM Reply	DP's are compressed in OO I think the big win is that OO moves the decision point to object construction. At that point you have evaluated your criteria and made your decision. Calls to polymorphic messages piggyback on the decision made at construction time. Thus, sans-OO you have to remake the decision for every operation. Writing the decision making code is potentially error-prone and is quite likely the cause of many bugs. So you have to make n decisions for n polymorphic operations. With OO, however, you only have one copy of the decision code and so you can do n operations with only 1 decision. I suspect this is a big part of the increased reliability in OO. That was lovely cheese. --Wallace, The Wrong Trousers
Post #178,369 by tablizer 10/7/04 11:07:36 PM Reply	Polymorphism relies on a lie Calls to polymorphic messages piggyback on the decision made at construction time. Thus, sans-OO you have to remake the decision for every operation. Writing the decision making code is potentially error-prone and is quite likely the cause of many bugs. So you have to make n decisions for n polymorphic operations. But that assumes that the world devides nicely up into a single "subtype" taxonomy. Often it does not. Subtypes are a poor approximation of reality, and you cannot compile the entire world. Sets would be a better fit to the real world. OO made a mistake by allying much closer to trees than to sets. Further, in Smalltalk and the like such dispatching is still made at run-time. ________________ oop.ismad.com
Post #178,374 by admin 10/7/04 11:23:58 PM Reply	Re: Polymorphism relies on a lie But that assumes that the world devides nicely up into a single "subtype" taxonomy. No, it doesn't. Stop confusing inheritance with polymorphism.You've been told this time and time again. Further, in Smalltalk and the like such dispatching is still made at run-time. And this is different from your "eval" how, exactly? Smalltalk optimizes its runtime dispatch to an extreme level that is possible only because of the meta-information provided by the classes. This is impossible with simple evals. Regards, -scott anderson "Welcome to Rivendell, Mr. Anderson..."
Post #179,112 by tablizer 10/13/04 1:10:40 PM Reply	Mostly, it is Stop confusing inheritance with polymorphism. In MOST cases, or at least in the comparisons being made here, polymorphism assumes a tree or tree-like taxonomy, or at least a single-dimensional list of mutually-exclusive choices. The "taxonomy" may not be explicitly given to the language, but it is there. And this is different from your "eval" how I thought we were comparing mostly case statements here. ________________ oop.ismad.com
Post #179,127 by admin 10/13/04 1:57:31 PM Reply	To the naive programmer, perhaps. Of C++, Smalltalk, Python, Perl, and Java, just C++ uses implementation inheritance only. C++ has multiple inheritance, however, which can mitigate the issue by using mix-ins or strictly virtual classes inherited to simulate interfaces. Regards, -scott anderson "Welcome to Rivendell, Mr. Anderson..."
Post #178,384 by FuManChu 10/8/04 1:56:34 AM Reply	Something tells me you'd like prototype OO as a midground
Post #178,385 by pwhysall 10/8/04 1:57:41 AM Reply	No-one can eat that much crow Peter [link\|http://www.debian.org\|Shill For Hire] [link\|http://www.kuro5hin.org\|There is no K5 Cabal] [link\|http://guildenstern.dyndns.org\|Blog]

Welcome to IWETHEY!