Boost has a function to const-qualify?

MB

14 Jan 2006 14 Jan '06

9:11 p.m.

Hi, Does boost has anything like : template< class T > inline const T& const_(const T& x) { return x; } I guess that boost::cref is different... Regards, MB p-stade.sourceforge.net

Show replies by date

Sebastian Redl

14 Jan 14 Jan

9:36 p.m.

MB wrote:

...

Hi,

Does boost has anything like :

template< class T > inline const T& const_(const T& x) { return x; }

I guess that boost::cref is different...

What exactly is the use of this? Did you perhaps mean to make the parameter non-const? If so, why would you need a function for something that's an implicit cast anyway? Sebastian Redl

MB

10:45 p.m.

Sebastian Redl wrote:

...

MB wrote:

...
Hi,

Does boost has anything like :

template< class T > inline const T& const_(const T& x) { return x; }

I guess that boost::cref is different...

What exactly is the use of this? Did you perhaps mean to make the parameter non-const? If so, why would you need a function for something that's an implicit cast anyway?

Hi, A short story is: struct X { }; void f(X& x); void f(const X& x); void test() { X x; { // initialize x. } f(const_(x)); } Here is the long story: http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2004/n1674.pdf Regards, MB p-stade.sourceforge.net

Sebastian Redl

15 Jan 15 Jan

10:48 a.m.

MB wrote:

...

A short story is:

struct X { }; void f(X& x); void f(const X& x);

void test() { X x; { // initialize x. }

f(const_(x)); }

Here is the long story: http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2004/n1674.pdf

I don't see the connection. Are you looking for a way to simplify the code in listing 4? Sebastian Redl

MB

4:29 p.m.

Sebastian Redl wrote:

...

MB wrote:

...
A short story is:

struct X { }; void f(X& x); void f(const X& x);

void test() { X x; { // initialize x. }

f(const_(x)); }

Here is the long story: http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2004/n1674.pdf

I don't see the connection. Are you looking for a way to simplify the code in listing 4?

Yes. 'const_' is something like an object-generator like 'make_pair'. Regards, MB p-stade.sourceforge.net

Felipe Magno de Almeida

2:34 p.m.

On 1/14/06, MB wrote: [snip]

...

Here is the long story: http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2004/n1674.pdf

Offtopic: Maybe the problem doesnt lie on const/non-const member functions? And the ugly syntax to choose the const when using a non-const class? Maybe this proposal shouldnt be more general? Not applying only to iterator/const_iterator. IMO, if it was, it would probably have more chance of acceptance.

...

Regards, MB p-stade.sourceforge.net

_______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users

Best regards, -- Felipe Magno de Almeida

Ai Azuma

3:43 p.m.

2006/1/15, Felipe Magno de Almeida :

...

[snip]

Offtopic: Maybe the problem doesnt lie on const/non-const member functions? And the ugly syntax to choose the const when using a non-const class? Maybe this proposal shouldnt be more general? Not applying only to iterator/const_iterator. IMO, if it was, it would probably have more chance of acceptance.

IMHO, cbegin/cend proposal may be too specific as you mentioned. There may be other cases where const/non-const member function overload resolution by programmers is desirable. And inline cv-qualification conversion may become a more generic approach in such cases. For example, consider the following code. // a vector class with copy-on-write template< class T > class vector_with_cow { ..... T &operator[]( size_type index ); T const &operator[]( size_type index ) const; ..... }; void f( vector_with_cow< int > &v ) { if( const_( v )[0] == 0 ){ // do not require deep copy return; } // write some values into v, require deep copy } Best regards, -- Ai Azuma

jens-theisen＠gmx.de

6:20 p.m.

Ai wrote:

...

For example, consider the following code. [...]

...

if( const_( v )[0] == 0 ){ // do not require deep copy

Or const_(some_string).begin() // please keep my iterators valid Jens

MB

4:33 p.m.

Felipe Magno de Almeida wrote:

...

On 1/14/06, MB wrote:

[snip]

...
Here is the long story: http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2004/n1674.pdf

Offtopic: Maybe the problem doesnt lie on const/non-const member functions? And the ugly syntax to choose the const when using a non-const class? Maybe this proposal shouldnt be more general? Not applying only to iterator/const_iterator.

I agree. Boost.Range has 'const_begin/end' for this purpose. But I think it could be applied to more. I renamed 'const_range' to 'const_'.

...

IMO, if it was, it would probably have more chance of acceptance.

If 'const_' is really missing from boost, I might do. :-) Regards, MB p-stade.sourceforge.net

jens-theisen＠gmx.de

2:53 p.m.

MB wrote:

...

Sebastian Redl wrote:

...

A short story is:

...

struct X { }; void f(X& x); void f(const X& x);

...

void test() { X x; { // initialize x. }

...

f(const_(x)); }

So it's purpose is to abbreviate a const_cast. Still useful. Jens

Felipe Magno de Almeida

7:16 p.m.

On 15 Jan 2006 14:53:00 +0000, Jens Theisen wrote: [snip]

...

So it's purpose is to abbreviate a const_cast. Still useful.

It's more readable and documents better the code too. Since const_cast can take consts away, but const_ only makes it const, and easier to use in generic code, since you dont need to rewrite the instance's type to be able to add const to it.

...

Jens

-- Felipe Magno de Almeida

Eric Niebler

10:25 p.m.

Felipe Magno de Almeida wrote:

...

On 15 Jan 2006 14:53:00 +0000, Jens Theisen wrote:

...
So it's purpose is to abbreviate a const_cast. Still useful.

It's more readable and documents better the code too. Since const_cast can take consts away, but const_ only makes it const

Then perhaps it should be called "add_const". -- Eric Niebler Boost Consulting www.boost-consulting.com

Victor A. Wagner Jr.

16 Jan 16 Jan

7:29 a.m.

At 15:25 2006-01-15, Eric Niebler wrote:

...

Felipe Magno de Almeida wrote:

...
On 15 Jan 2006 14:53:00 +0000, Jens Theisen wrote:

...
So it's purpose is to abbreviate a const_cast. Still useful.

It's more readable and documents better the code too. Since const_cast can take consts away, but const_ only makes it const

Then perhaps it should be called "add_const".

didn't I read about something called that (or something close) in the MPL book?

...

-- Eric Niebler Boost Consulting www.boost-consulting.com _______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users

Victor A. Wagner Jr. http://rudbek.com The five most dangerous words in the English language: "There oughta be a law"

Sebastian Redl

10:15 a.m.

Victor A. Wagner Jr. wrote:

...

didn't I read about something called that (or something close) in the MPL book?

Yes, but that's only for types (code example not authoritative - I have no idea how the thing is really called): template<typename T> void funct(T bla) { typename boost::mpl::add_const<T>::type constant(10); // ... } The type of constant is "const T". Sebastian Redl

MB

4:19 p.m.

Sebastian Redl wrote:

...

Victor A. Wagner Jr. wrote:

...
didn't I read about something called that (or something close) in the MPL book?

Yes, but that's only for types (code example not authoritative - I have no idea how the thing is really called):

template<typename T> void funct(T bla) { typename boost::mpl::add_const<T>::type constant(10); // ... }

The type of constant is "const T".

Sebastian Redl

I think 'add_const' is reserved for Boost.TypeTraits. I post this tiny proposal to Boost.Devel anyway. Regards, MB p-stade.sourceforge.net

Thorsten Ottosen

4:34 p.m.

MB wrote:

...

Sebastian Redl wrote:

I think 'add_const' is reserved for Boost.TypeTraits.

I post this tiny proposal to Boost.Devel anyway.

When we discussed the wg21 proposal many months ago and I added const_begin() etc to boost.range, Dave Abrahams used this function template< class T > inline T const& as_const( T const& r ) { return r; } Similarly, we might add as_volatile(); as_const_volatile(); as_mutable(); The latter would be a cast from rvalue to lvalue like it happens in boost.foreach. -Thorsten

MB

6:44 p.m.

Thorsten Ottosen wrote:

...

MB wrote:

...
Sebastian Redl wrote:

I think 'add_const' is reserved for Boost.TypeTraits.

I post this tiny proposal to Boost.Devel anyway.

When we discussed the wg21 proposal many months ago and I added const_begin() etc to boost.range, Dave Abrahams used this function

template< class T > inline T const& as_const( T const& r ) { return r; }

I found the thread. http://lists.boost.org/Archives/boost/2004/09/72340.php Maybe the time has come? :-) A range algorithm like Boost.StringAlgorithm no longer allows users to call 'const_begin/end'. Now that something like 'as_const' is really needed. boost::for_each( rng | const_qualified | // protect rng! reversed | null_terminated | tab_expanded | utf8_decoded | transformed(toZero()), &do_something ); I think 'const_rbegin' is obviously blob ;-)

...

Similarly, we might add

as_volatile(); as_const_volatile(); as_mutable();

The latter would be a cast from rvalue to lvalue like it happens in boost.foreach.

Beyond my sight. Is it possible such conversion? Regards, MB p-stade.sourceforge.net

Eric Niebler

7:43 p.m.

Thorsten Ottosen wrote:

...

as_mutable();

The latter would be a cast from rvalue to lvalue like it happens in boost.foreach.

Not sure what you're referring to. Boost.Foreach does no such rvalue-to-lvalue cast. -- Eric Niebler Boost Consulting www.boost-consulting.com

Thorsten Ottosen

9:52 p.m.

Eric Niebler wrote:

...

Thorsten Ottosen wrote:

...
as_mutable();

The latter would be a cast from rvalue to lvalue like it happens in boost.foreach.

Not sure what you're referring to. Boost.Foreach does no such rvalue-to-lvalue cast.

So this wouldn't compile: vector<int> foo(); ... BOOST_FOREACH( int& r, foo() ) ++r ? -Thorsten

Eric Niebler

10:07 p.m.

Thorsten Ottosen wrote:

...

Eric Niebler wrote:

...
Thorsten Ottosen wrote:

...
as_mutable();

The latter would be a cast from rvalue to lvalue like it happens in boost.foreach.

Not sure what you're referring to. Boost.Foreach does no such rvalue-to-lvalue cast.

So this wouldn't compile:

vector<int> foo(); ... BOOST_FOREACH( int& r, foo() ) ++r

?

No, it doesn't compile, and it shouldn't. What would be the point of mutating the unnamed temporary vector returned by foo()? -- Eric Niebler Boost Consulting www.boost-consulting.com

Thorsten Ottosen

10:57 p.m.

Eric Niebler wrote:

...

Thorsten Ottosen wrote:

...
Eric Niebler wrote:

...

...
...
Not sure what you're referring to. Boost.Foreach does no such rvalue-to-lvalue cast.

So this wouldn't compile:

vector<int> foo(); ... BOOST_FOREACH( int& r, foo() ) ++r

?

No, it doesn't compile, and it shouldn't. What would be the point of mutating the unnamed temporary vector returned by foo()?

I would probably be efficiency. Assume it was a vector of strings and that you had to append "_" to each of the strings and then pass the string to some other function. Or maybe I wanted to do the following: foo::copy( foo::unique( boost::as_mutable( foo() ) ), some_where ); -Thorsten

6878

Age (days ago)

6880

Last active (days ago)

List overview

Download

20 comments

8 participants

participants (8)

Ai Azuma
Eric Niebler
Felipe Magno de Almeida
jens-theisen＠gmx.de
MB
Sebastian Redl
Thorsten Ottosen
Victor A. Wagner Jr.