J
James Kanze
Not quite. There's also a compiler generated copy constructor.
Not quite. There's also a compiler generated copy constructor.
B
BobR
James Kanze wrote in message...
/* """ quote
Not quite. There's also a compiler generated copy constructor.
""" */
Thanks James. I posted late night, and started to re-post an addition, but,
was talking about constructors only. Copy-constructors are, to me, an
separate column. Like the old Chinese restaurant menu, one from column A,
one from column B, etc. (Ctor, CCtor, op=, Dtor). (what, two from column A?
that will be another $2.75! only one in col D, the Dessert-tor (when you
finish that, the meal is over.)).
Without an constructor, a copy-constructor seems pretty useless (how would
you get an instance to copy? <G>).
/* """ quote
Not quite. There's also a compiler generated copy constructor.
""" */
Thanks James. I posted late night, and started to re-post an addition, but,
was talking about constructors only. Copy-constructors are, to me, an
separate column. Like the old Chinese restaurant menu, one from column A,
one from column B, etc. (Ctor, CCtor, op=, Dtor). (what, two from column A?
that will be another $2.75! only one in col D, the Dessert-tor (when you
finish that, the meal is over.)).
Without an constructor, a copy-constructor seems pretty useless (how would
you get an instance to copy? <G>).
J
JohnQ
I don't use structs like that much (at all?). When I need that kind of
functionality, I think "class".
I believe you.
John
J
James Kanze
I know what you mean. From a language point of view, it may
always be a constructor, but from a logical point of view, I,
too, tend to think of creating a new object as somehow different
from creating a copy. I even have a few special cases where I
keep count of the copies, and the "real" destruction---something
like freeing the resource in RAII---only takes place when the
last copy is destructed. (Boost::shared_ptr is a special case
of this particular idiom.)
More likely, either everything from column B, or nothing
.Quite. If you have a user defined copy-constructor (which
inhibits the automatic generation of a default constructor),
you'd better have at least one other user defined constructor as
well.
G
Geo
J
JohnQ
Geo said:
J
James Kanze
[...]
They have their acts together. It's you that don't seem to
understand.
And pay an extremely high performance penalty for it. More than
a few applications can't afford that.
J
JohnQ
[...]
"They have their acts together. It's you that don't seem to
understand."
I won't be buying that Brooklyn bridge either, thank you.
"And pay an extremely high performance penalty for it."
You're assuming that byte-aligning struct members implies non-optimum
alignment. That's a wrong assumption. And perhaps you're thinking that
everything has to be byte-aligned or not, which isn't the case either.
"More than a few applications can't afford that."
There is no penalty if you define your stucts correctly so that the data
members align. Secondly, the vast majority of applications are not
performance constrained. Indeed, they are IO constrained if anything, so not
having to marshal will speed up the program (on the "native" platform).
John
J
James Kanze
J
JohnQ
J
James Kanze
More reality. I've been at this business long enough to
recognize incompetence when I see it. If you don't understand
the necessity of marshalling, or of different processor
architectures, you're never going to write any quality software.
Primary what? Different people have different requirements, so
there are different architectures out there to solve them. I
work mainly on large scale servers, and I certainly don't want
to be forced to use an ARM architecture, just because it's the
most widespread. (I don't even really know the ARM
architecture, so little is it relevant to my field of work,
where Sparc is by far the dominant architecture.)
Not really. Not anything very useful, anyway.
Or trying to support users who are writing real applications,
and need to manage complexity.
J
JohnQ
"More reality. I've been at this business long enough to
recognize incompetence when I see it. If you don't understand
the necessity of marshalling, or of different processor
architectures, you're never going to write any quality software."
You're marshalling everywhere, all the time and calling me stupid? Have you
considered getting some bran in your diet? Maybe you wouldn't be so ornery
then.
"Primary what? Different people have different requirements, so
there are different architectures out there to solve them. I
work mainly on large scale servers, and I certainly don't want
to be forced to use an ARM architecture, just because it's the
most widespread. (I don't even really know the ARM
architecture, so little is it relevant to my field of work,
where Sparc is by far the dominant architecture.)"
Nevermind, you don't get it.
"Not really. Not anything very useful, anyway."
You fo right ahead and keep maintaining the complexity. That's your
perogative.
"Or trying to support users who are writing real applications,
and need to manage complexity."
Broken record.
John