tma@mas1.UUCP (Tim Atkins) (06/09/89)
For the life of me, I don't understand why SomeClass x[100]; needs to invoke SomeClass::SomeClass when simpleCType x[100]; has no need to do such. My intuition would be that in the former case I have merely allocated space for 100 instances of SomeClass and that no initialization was asked for or even desirable. Can someone explain this feature to me? I fail to see the current C++ behaviour as a desirable feature. thanks, Tim Atkins
jima@hplsla.HP.COM (Jim Adcock) (06/10/89)
Well,
SomeClass x1;
needs to invoke SomeClass::SomeClass whereas
simpleCType x1;
has no need to do such.
A better question is why does:
SomeClass x[100];
with:
inline SomeClass::SomeClass(){}
cause 100 invocations of this do-nothing "initializer" whereas x1 does not
cause even 1 invocation of the do-nothing "initializer".shankar@hpclscu.HP.COM (Shankar Unni) (06/13/89)
> A better question is why does: > SomeClass x[100]; > with: > inline SomeClass::SomeClass(){} > cause 100 invocations of this do-nothing "initializer" whereas x1 does not > cause even 1 invocation of the do-nothing "initializer". Lazy code generation... No, I take that back. This is a solvable problem, but I suppose that the ROI for doing it the most efficient way was not seen as making it worthwhile to do. Who knows.. In general, it's easy to do the trivial case of the null constructor (I mean a constructor with an empty body). If the constructor has even 1 statement, it becomes pretty hard. A better solution would have been an array constructor syntax. Maybe something like: SomeClass::SomeClass[int count]() { ... } Naah! :-) (Parsing C++ is already a pain in the ***; no need to make it any worse). But seriously, it would address this problem. Only one function is called with a pointer to the array and a number of elements, and whatever other parameters the user specifies. ---- Shankar.
simpson@poseidon.uucp (Scott Simpson) (06/13/89)
> simpleCType x1; > >has no need to do such. >A better question is why does: > > SomeClass x[100]; > >with: > >inline SomeClass::SomeClass(){} > >cause 100 invocations of this do-nothing "initializer" whereas x1 does not >cause even 1 invocation of the do-nothing "initializer". . . . . . Stupid inews line count "feature" . . . . . Because the SomeClass initializer may do a "new" and you want 100 objects allocated. See Stroustrup section 5.5.5 page 162. What do I win? :-) Scott Simpson TRW Space and Defense Sector oberon!trwarcadia!simpson (UUCP) trwarcadia!simpson@usc.edu (Internet)
jima@hplsla.HP.COM (Jim Adcock) (06/15/89)
> >inline SomeClass::SomeClass(){} > > > Because the SomeClass initializer may do a "new" and you want 100 objects > allocated. See Stroustrup section 5.5.5 page 162. What do I win? :-) How can the SomeClass initializer do a "new" when I pointed out its not doing anything? I was arguing that I would at least expect a compiler to recognize when a class has a no-op initializer and not bother calling it 100 times ???