I've always wondered why generics mess this up

Miscellaneous Topics
#21

The java code K.I.L.E.R posted, and the one you claimed was in ArrayList (can’t find it personally) is not valid java code in the sense that it would always throw a runtime exception and fail.

“(String)new Object();” will compile, but it will always fail at runtime, just like “(String[])new Object[];” will.
And I don’t think anyone would say that “(String)new Object();” is valid java code.

#22

[quote]The java code K.I.L.E.R posted, and the one you claimed was in ArrayList (can’t find it personally) is not valid java code in the sense that it would always throw a runtime exception and fail.
[/quote]
ArrayList is doing this in its ArrayList(int initialCapacity) constructor. But the cast there is for a generic type E (I’m assuming K.I.L.E.R was trying to do the same), which won’t produce any exception (because no cast will be made).

Anyway, such a cast is valid java code in the sense that it will compile just fine. Agreed on it being stupid and dangerous though. :wink:

#23

So why doesn’t Sun just follow C# and implement a real version of generics?

#24

[quote]So why doesn’t Sun just follow C# and implement a real version of generics?
[/quote]
Backward compatibility. You either have a completely new set of classes (such as Collections) for generic use, or insist that all existing code is at least recompiled.
Sun wanted existing byte code to keep working.

#25

Can’t we have both?

#26

[quote]Can’t we have both?
[/quote]
Of course. But we don’t. And that is something that should put a certain amount of fear in each of use w.r.t. Sun’s ongoing (in)ability not to screw-up java. But that’s been discussed at length here already.

#27

Unfortunately this leads to frustration.
How big of an impact will .NET 2.0 have on Sun?
Will Sun be pushed further to improve Java at a faster pace?