Unique Integer Types -- a useful C++ trick

[Christopher Brian Colohan]

This has probably been done a zillion times before, but it is new to
me (and some of the folks I work with), and also quite useful.

Let me know if there are obvious improvements that I have missed. :-)

http://www.colohan.com/docs/unique_integers.html

Chris
--
Chris Colohan Email: ch***@colohan.ca PGP: finger co*****@cs.cmu.edu
Web: www.colohan.com Phone: [phone]

[Victor Bazarov]

Christopher Brian Colohan wrote:

This has probably been done a zillion times before, but it is new to
me (and some of the folks I work with), and also quite useful.

Let me know if there are obvious improvements that I have missed. :-)

http://www.colohan.com/docs/unique_integers.html

Preventing conversion-construction from anything is usually done using
the 'explicit' keyword. I think you might find looking into it useful.
It would be nice not to have the number stated as the template argument
when declaring typedefs. I am not sure it's possible, but I suspect
that it is.

In any case, interesting work.

V

[Jonathan Turkanis]

"Victor Bazarov" <v.********@comAcast.net> wrote in message
news:O9****************@newsread1.dllstx09.us.to.v erio.net...

Christopher Brian Colohan wrote:

This has probably been done a zillion times before, but it is new to
me (and some of the folks I work with), and also quite useful.

Let me know if there are obvious improvements that I have missed. :-)

http://www.colohan.com/docs/unique_integers.html

It would be nice not to have the number stated as the template argument
when declaring typedefs. I am not sure it's possible, but I suspect
that it is.

It can almost be done with CRTP:

struct set_num_t : unique_int<set_num_t> { };

The trouble is you lose the constructor taking an unsigned int and the private
constructor taking a bool. If you don't mind macros, you could do

#define UNIQUE_INT(name) \
struct name : unique_int<name> { \
private: name(bool); \
public: name(unsigned v = 0u) : unique_int<name>(v) { }\
}; \
/**/

UNIQUE_INT(set_num_t)
UNIQUE_INT(entry_num_t)
UNIQUE_INT(lru_entry_t)

Jonathan

[Asfand Yar Qazi]

>

The trouble is you lose the constructor taking an unsigned int and the private
constructor taking a bool. If you don't mind macros, you could do

#define UNIQUE_INT(name) \
struct name : unique_int<name> { \
private: name(bool); \
public: name(unsigned v = 0u) : unique_int<name>(v) { }\
}; \
/**/

UNIQUE_INT(set_num_t)
UNIQUE_INT(entry_num_t)
UNIQUE_INT(lru_entry_t)

Macros are cool. It would have been great if Bjarne had improved the
preprocessor as well when he invented the C++ language... perhaps it
is still possible with the new C++ standard coming up?
--
http://www.it-is-truth.org/

[Jonathan Turkanis]

"Asfand Yar Qazi" <se*@message.body.com> wrote in message
news:30*************@uni-berlin.de...

The trouble is you lose the constructor taking an unsigned int and the private constructor taking a bool. If you don't mind macros, you could do

#define UNIQUE_INT(name) \
struct name : unique_int<name> { \
private: name(bool); \
public: name(unsigned v = 0u) : unique_int<name>(v) { }\
}; \
/**/

UNIQUE_INT(set_num_t)
UNIQUE_INT(entry_num_t)
UNIQUE_INT(lru_entry_t)

Macros are cool. It would have been great if Bjarne had improved the
preprocessor as well when he invented the C++ language... perhaps it
is still possible with the new C++ standard coming up?

There are definitely going to be some improvements, because variadics and
several other features from C99 are going to be added. Other features, such as
scoped macros, are under consideration.

Jonathan