Using _t vs _type suffixes in typedefs

[Chris Fairles]

I'm wondering if theres any rationale behind using _t vs _type in
typedefs. I took a look in the standard and it doesn't say much
besides things like "size_type must be size_t", "difference_type must
be ptrdiff_t".

Looking at some libraries in Boost (Boost.GIL specifically) the
concepts describe types mainly using _type (value_type etc) but the
odd _t sneaks in there (const_t, layout_t).

One thing I notice is that _t is used more often with concrete types
whereas _type is used to re-declare template parameters etc. as its
something you usually want to pull out with scope resolution
(my_concrete_type<T>::value_type).

Any thoughts/opinions?

Cheers,
Chris

 

[Victor Bazarov]

I'm wondering if theres any rationale behind using _t vs _type in
typedefs. I took a look in the standard and it doesn't say much
besides things like "size_type must be size_t", "difference_type must
be ptrdiff_t".

I believe it's more of a nod in C direction. They use _t. 'size_t',
'ptrdiff_t', 'wchar_t' all are defined in C [as well]. It's easier
to have them named the same in the two related languages.

The C++ types that don't have counterparts in C (like 'size_type'
member of classes), don't have to be shortened.

Looking at some libraries in Boost (Boost.GIL specifically) the
concepts describe types mainly using _type (value_type etc) but the
odd _t sneaks in there (const_t, layout_t).

The 'value_type' typedef is part of the Standard containers.

One thing I notice is that _t is used more often with concrete types
whereas _type is used to re-declare template parameters etc. as its
something you usually want to pull out with scope resolution
(my_concrete_type<T>::value_type).

Any thoughts/opinions?

See above.

V

 

[Gennaro Prota]

The C++ types that don't have counterparts in C (like 'size_type'
member of classes), don't have to be shortened.

It's also worth noticing that *_t is reserved by POSIX.