std::vector<const MyType> not allowed

[muzicmakr]

I'm porting some code from windows to mac, and there are some
instances of std::vector<const MyType>, that compiled just fine on the
pc, but won't compile under gcc.

I'd never tried to do this particular construct myself, and after some
searching online, I found a post somewhere saying this isn't legal c++
because the standard containers need types that are assignable.

My questions-- is that correct? Why does visual studio allow it if
it's wrong? And is there some other way to try to preserve the intent
of the original? (Otherwise I'm just going to strip out the
consts.)

Thanks for your time!

mich

[Victor Bazarov]

mu*******@yahoo.com wrote:

I'm porting some code from windows to mac, and there are some
instances of std::vector<const MyType>, that compiled just fine on the
pc, but won't compile under gcc.

I'd never tried to do this particular construct myself, and after some
searching online, I found a post somewhere saying this isn't legal c++
because the standard containers need types that are assignable.

My questions-- is that correct? Why does visual studio allow it if
it's wrong? And is there some other way to try to preserve the intent
of the original? (Otherwise I'm just going to strip out the
consts.)

The general requirement is for the stored type to be assignable and
copy-constructible. But there can be cases where the assignable
requirement can be lifted. I can't think of one right now, but it is
not outside of the realm of possibility... For example, if you use the form

std::vector<mytypev(10);

to create a vector of 10 objects, your 'mytype' should also be
default-constructible. But if you use the form

mytype prototype(somearguments);
std::vector<mytypev(10, prototype);

there is no requirement for default-constructibility...

V
--
Please remove capital 'A's when replying by e-mail
I do not respond to top-posted replies, please don't ask

[Sam]

mu*******@yahoo.com writes:

I'm porting some code from windows to mac, and there are some
instances of std::vector<const MyType>, that compiled just fine on the
pc, but won't compile under gcc.

I'd never tried to do this particular construct myself, and after some
searching online, I found a post somewhere saying this isn't legal c++
because the standard containers need types that are assignable.

My questions-- is that correct?

That is correct.

Why does visual studio allow it if
it's wrong?

Because the very last thing that Microsoft wants you to do is write portable
code that complies with commodity standards that are not controlled by
Microsoft. Microsoft would like nothing more than to force you to write code
that compiles only on Windows.

And is there some other way to try to preserve the intent
of the original? (Otherwise I'm just going to strip out the
consts.)

The "intent of the original" would be to declare the array as a plain,
garden-variety std::vector<MyType>, and, in functions or a context that
should not be able to modify the contents of the vector, pass it by
reference as a const std::vector<MyType&. Any code will have to access the
vector using const_iterator, not iterator (you'll probably have to rewrite a
whole bunch of stuff).

>
Thanks for your time!

mich

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)

iEYEABECAAYFAkigyhMACgkQx9p3GYHlUOK9vACdGVEZA9Mcca Iz/azyhWgNhVxR
y/0An07qbB7Ni4cwZVnwYU1ZXR4nxSb7
=vSnM
-----END PGP SIGNATURE-----

[James Kanze]

On Aug 12, 12:08 am, [e-mail] wrote:

I'm porting some code from windows to mac, and there are some
instances of std::vector<const MyType>, that compiled just
fine on the pc, but won't compile under gcc.

I'd never tried to do this particular construct myself, and
after some searching online, I found a post somewhere saying
this isn't legal c++ because the standard containers need
types that are assignable.

My questions-- is that correct? Why does visual studio allow
it if it's wrong?

It's undefined behavior. It may cause an error when compiling,
it may compile, but do strange things during execution, or it
may work just fine.

And is there some other way to try to preserve the intent of
the original? (Otherwise I'm just going to strip out the
consts.)

What is the intent of the original? Given the semantics of
vector, having a non-const vector of const objects doesn't make
much sense. And if you declare an std::vector< MyType const,
there's no way you can get a non-const reference to any of the
elements.

--
James Kanze (GABI Software) email:ja*********@gmail.com
Conseils en informatique orientée objet/
Beratung in objektorientierter Datenverarbeitung
9 place Sémard, 78210 St.-Cyr-l'École, France, [phone]

[Lance Diduck]

My questions-- is that correct? *Why does visual studio allow it if

it's wrong? *

MSVC probably allowed it because your code never uses a member
function that cared whether T was const or not. This is an example of
a struct that in one version allows a const whereas the other does not
template<class T>
struct Foo{
Foo(){}
#ifdef ALLOW_CONST_T
Foo(T const&v):var(v){}
#else
Foo(T const&v){
var=v;
}
#endif
T var;
};

Whether I define ALLOW_CONST_T or not does not change the behavior of
Foo(other than performance in certian cases of T) but one version
allows Foo<const X and the other does not. And that only applies if
the user of Foo actually invokes that particular constructor.
So what is likely happening is that the particular mixture of user
code and how the container was implemented never actully hit upon
something that cared about the constness of T, whereas in the gcc
version it did.
Lance

[muzicmakr]

Thanks for all the informative replies, they were very helpful.

mich

[pjp]

On Aug 11, 7:24*pm, Sam <[e-mail]:

*application_pgp-signature_part
< 1KViewDownload

[e-mail] writes:

I'm porting some code from windows to mac, and there are some
instances of std::vector<const MyType>, that compiled just fine on the
pc, but won't compile under gcc.

I'd never tried to do this particular construct myself, and after some
searching online, I found a post somewhere saying this isn't legal c++
because the standard containers need types that are assignable.

My questions-- is that correct?

That is correct.

* * * * * * * * * * * * * * * * *Why does visual studio allow it if
it's wrong?

Because the very last thing that Microsoft wants you to do is write portable
code that complies with commodity standards that are not controlled by
Microsoft. Microsoft would like nothing more than to force you to write code
that compiles only on Windows.

Nice guess, if a bit paranoid, but you're wrong. Dinkumware added the
capability to declare a container of const type because -- are you
ready for this? -- customers demanded it. And now the latest revision
of the C++ Standard calls for *all* implementations to support vectors
of const type for all operations that make sense.

So much for vendor lock in.

P.J. Plauger
Dinkumware, Ltd.
http://www.dinkumware.com

* * * * * * And is there some other way to try to preserve the intent
of the original? *(Otherwise I'm just going to strip out the
consts.)

The "intent of the original" would be to declare the array as a plain,
garden-variety std::vector<MyType>, and, in functions or a context that
should not be able to modify the contents of the vector, pass it by
reference as a const std::vector<MyType&. Any code will have to access the
vector using const_iterator, not iterator (you'll probably have to rewrite a
whole bunch of stuff).

Thanks for your time!

mich- Hide quoted text -

- Show quoted text -