typename usege problem ...

[466]

Hi all,
I have this member in a template class:

bool opsInsert(int nId, T pObj){
pairRet = insert(typename std::map<int,T>::value_type(nId, pObj)); }

which does compile with gcc but not with cl.exe (Windows).
If I remove the typename keyword, the code does not compile with gcc,
but compiles witc cl.exe....

The question: which is the standard way to call this function? Is there
any workaround to make compile with both compilers?

I know this is a compiler issue, but I hope you will help me ...

thanx,
stefan

 

[Rolf Magnus]

Hi all,
I have this member in a template class:

bool opsInsert(int nId, T pObj){
pairRet = insert(typename std::map<int,T>::value_type(nId, pObj)); }

which does compile with gcc but not with cl.exe (Windows).
If I remove the typename keyword, the code does not compile with gcc,
but compiles witc cl.exe....

The question: which is the standard way to call this function?

The standard requires the typename keyword in this place.

Is there any workaround to make compile with both compilers?

Maybe use a macro, like:

#if compiler is vc++
#define TYPENAME
#else
#define TYPENAME typename
#endif

bool opsInsert(int nId, T pObj){
pairRet = insert(TYPENAME std::map<int,T>::value_type(nId, pObj)); }

It's ugly, but works.

 

[466]

The standard requires the typename keyword in this place.


Maybe use a macro, like:

#if compiler is vc++
#define TYPENAME
#else
#define TYPENAME typename
#endif

bool opsInsert(int nId, T pObj){
pairRet = insert(TYPENAME std::map<int,T>::value_type(nId, pObj)); }

It's ugly, but works.

Sure, as a ultimate solution ...
Thanx.

 

[Ron Natalie]

Hi all,
I have this member in a template class:

bool opsInsert(int nId, T pObj){
pairRet = insert(typename std::map<int,T>::value_type(nId, pObj)); }

which does compile with gcc but not with cl.exe (Windows).
If I remove the typename keyword, the code does not compile with gcc,
but compiles witc cl.exe....

Get a version of the windows compiler that was published in the last
five years.

 

[jim_taylor]

Maybe use a macro, like:

#if compiler is vc++
#define TYPENAME
#else
#define TYPENAME typename
#endif

You could use the macro
#define typename
as basically the older compilers don't know the typename keyword.
Be careful to make your #if specific to the current (and older)
versions of your compiler, as you don't want it to break when you do
upgrade...
Note also that the standard also allows typename as a synonym
(basically) for class in the template arguments,which also will not
work in yuour old compiler. i.e. you can write
template <typename MyType>

 

[Bo Persson]

You could use the macro
#define typename
as basically the older compilers don't know the typename keyword.

Except that in this case we recognize a certain compiler, that requires
typename in some places, doesn't care in other places, and flat out refuses
to recognize some cases where typename *is* required.

A very useful advice - use a compiler from this millennium - has alredy been
given by another poster.


Bo Persson

 

[Mark P]

Hi all,
I have this member in a template class:

bool opsInsert(int nId, T pObj){
pairRet = insert(typename std::map<int,T>::value_type(nId, pObj)); }

which does compile with gcc but not with cl.exe (Windows).
If I remove the typename keyword, the code does not compile with gcc,
but compiles witc cl.exe....

The question: which is the standard way to call this function? Is there
any workaround to make compile with both compilers?

I know this is a compiler issue, but I hope you will help me ...

thanx,
stefan

As has been noted, the standard requires "typename". However you could
replace the expression by the actual type of value_type, namely
std:air< const int, T>. That should not require "typename".