Prev: C++ analog of setvbuf() ?
Next: Need to analyze C source code and determine loops (program analysis)
From: restor on 8 Dec 2009 07:37
> Subject says it all really. I have been in Javaland for a while where
> you do have this capability. I am aware of Meyers Effective STL item 2
> ("beware of the illusion of container-independent code") but I don't
> think item 2 applies when all you want to do is iterate over a
> container. This is the simple case that people want to do most of the
> time (IMHO). Of course if you want to do something else like reverse
> iterate etc then with C++ you can. But it does seem mysterious to me
> why it is more awkward to do a generic iteration for the most common
> case.

Hi, it is not clear what you mean by "generic iteration." Suppose all
containers
did inherit from a common type. Could you give an example how this
"generic iteration"
would look like?

Regards,
&rzej

--
[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]

From: Joe Gottman on 8 Dec 2009 07:39
Andrew wrote:
> Subject says it all really. I have been in Javaland for a while where
> you do have this capability. I am aware of Meyers Effective STL item 2
> ("beware of the illusion of container-independent code") but I don't
> think item 2 applies when all you want to do is iterate over a
> container. This is the simple case that people want to do most of the
> time (IMHO). Of course if you want to do something else like reverse
> iterate etc then with C++ you can. But it does seem mysterious to me
> why it is more awkward to do a generic iteration for the most common
> case.
>
> I realise that in the upcoming standard the use of auto to declare the
> iterator will make it slightly easier.

C++ distinguishes between run-time polymorphism (from inheritance)
and compile-time polymorphism (from templates). The STL was built
around compile-time polymorphism for several reasons:

1) It's faster. Small template functions are easily inlined for
added speed.
2) It can be used with non-class objects. Under your scheme, all
iterators and containers would have to inherit from some abstract
Container and iterator classes, but the STL algorithms work fine with
built-in arrays and pointers, which don't inherit from anything. Also,
vector iterators are often pointers.


Joe Gottman

--
[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]

From: Seungbeom Kim on 8 Dec 2009 07:30
Andrew wrote:
> Subject says it all really. I have been in Javaland for a while where
> you do have this capability. I am aware of Meyers Effective STL item 2
> ("beware of the illusion of container-independent code") but I don't
> think item 2 applies when all you want to do is iterate over a
> container. This is the simple case that people want to do most of the
> time (IMHO). Of course if you want to do something else like reverse
> iterate etc then with C++ you can. But it does seem mysterious to me
> why it is more awkward to do a generic iteration for the most common
> case.

All the standard containers provide begin(), end(), and typenames
[const_]iterator (and value_type), and the iterators share the common
interface (operator*() and operator++()). With these you can write

for (C::iterator i = c.begin(); i != c.end(); ++i) {
C::value_type& v = *i;
// ...
}

for every standard container type C and a container object c of type C.
Isn't this a "generic iteration"? What makes you think the standard
containers don't allow generic iteration?


# To be pedantic, what you're using is probably not "STL containers"
but "standard containers", even though the latter originated from STL.

--
Seungbeom Kim

[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]

From: Frank Bergemann on 8 Dec 2009 08:55
Andrew schrieb:
> Subject says it all really. I have been in Javaland for a while where
> you do have this capability.

I never understood the idea to have an inheritance tree, where
everything has a common root type of object.
A teacher of a C++ course(!) was telling in a break, that he likes the
this inheritance approach - he told, that java is more consequent in this.
But i was argueing, that i can't follow this approach.
Because it's is a matter of selection - based on purpose - which
distinctions/extensions you do and on what level in a hierarchy.
There is nothing like a common (static) structure, which fits everything.

> I am aware of Meyers Effective STL item 2
> ("beware of the illusion of container-independent code") but I don't
> think item 2 applies when all you want to do is iterate over a
> container.

And this is a SINGLE issue, that you address here!
BTW: in C++ there is no implementation for iteration in the container
itself. There is _support_ for this provided by the container class. But
adapters (iterators) are used to facilitate for this. I think it's an
example for the FTSE (fundamental theorem of software enginnering) which
applies here: another level of indirection. The combination is the
solution and it offers more variants.

> This is the simple case that people want to do most of the
> time (IMHO). Of course if you want to do something else like reverse
> iterate etc then with C++ you can. But it does seem mysterious to me
> why it is more awkward to do a generic iteration for the most common
> case.

You might want implement your own iterator for some purpose - to play
nicely with an STL container.

> I realise that in the upcoming standard the use of auto to declare the
> iterator will make it slightly easier.

a matter of taste.
C++ is a language, where you wanna know, what's going on.
Auto IMHO is using the help of the compiler for writing (generic) code
easier for dependent types.

>
> Regards,
>
> Andrew Marlow
>

rgds,
Frank

--
[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]

From: peter koch larsen on 8 Dec 2009 09:20
On 7 Dec., 20:52, Andrew <marlow.and...(a)googlemail.com> wrote:
> Subject says it all really. I have been in Javaland for a while where
> you do have this capability. I am aware of Meyers Effective STL item 2
> ("beware of the illusion of container-independent code") but I don't
> think item 2 applies when all you want to do is iterate over a
> container. This is the simple case that people want to do most of the
> time (IMHO). Of course if you want to do something else like reverse
> iterate etc then with C++ you can. But it does seem mysterious to me
> why it is more awkward to do a generic iteration for the most common
> case.
>
> I realise that in the upcoming standard the use of auto to declare the
> iterator will make it slightly easier.

So what is it you want to do that you can't do with a simple template-
function (that is probably already made for you and can be found in
e.g. <algorithm>)?
Requiring iterators to have virtual functions would not only increase
the memory- and cpu-footprint of your program; it would also prohibit
using ordinary pointers as iterators and require the elements
themselves to be derived from a common baseclass. If not, how would
you dereference an iterator?

--
[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]

 | 
Pages: 1
Prev: C++ analog of setvbuf() ?
Next: Need to analyze C source code and determine loops (program analysis)