Adding arity traits

[gchatelet]

First minimal addition to std.traits.
Adding an arity template returning the number of argument of a function.

Maybe it should go to std.functional instead ?

[dnadlinger]

Imho, the way variadic parameters are handled (i.e. ignored) should be documented, and the description should refer to »parameters« instead of »arguments«.

Also consider leaving off the type of the constant (it should probably be size_t), it will be inferred anyway.

Anyway, I'm not quite sure whether an isolated addition like this makes much sense. I could imagine that such a template would fit well into a set of other callable type handlers/manipulators, but as it is now, I don't see much benefit over ParameterTypeTuple!().length.

[gchatelet]

This patch is motivated by the following discussion on the forum. @andralex suggested to add small pull requests and proposed to start with arity.

Yes there is no real benefit over ParameterTypeTuple!().length, it just states your intents clearer.

Also

• I agree with the uint type, I was supposed to remove it but I forgot.
• I will also update the documentation accordingly to your comments ( parameters / variadic )

We can debate about whether this should be a coherent proposal or just some piecemeal additions.

[andralex]

@klickverbot: uint is better than size_t because the latter measures sizes of objects in memory whereas uint is a better fit as a "natural number".

[andralex]

Probably variadic functions should have uint.max as arity, or not define arity at all. I'd prefer the latter.

[dnadlinger]

@andralex: typeof(TypeTuple!(1, 2, 3).length) is size_t (well, ulong instead of uint on D_LP64, that is). I thought that this might be a problem for compiling on x86_64, but forgot about all the integer range checking magic for compile-time constants.

[gchatelet]

Which one of the following is recommended for the variadic function unittest ?

• static assert(__traits(compiles,arity!variadicFoo)==false);
• static assert(is(typeof(arity!variadicFoo))==false);

[jmdavis]

You can go with either:

static assert(__traits(compiles,arity!variadicFoo)==false);
static assert(is(typeof(arity!variadicFoo))==false);

Some people prefer one, some the other. __traits is arguably a bit clearer, but it's longer, and anyone hacking on Phobos needs to understand the is(typeof... synatx as well, since it's quite common, so on some level, the clarity gain is irrelevant. Personally, I used to prefer __traits, but increasingly, I've just been using is(typeof..., because it's shorter.

[jmdavis]

@andralex I actually would have argued size_t over uint, because that's what almost all length stuff is (the primary exception being stuff that needs to always be ulong).

[andralex]

Alright size_t it is. @gchatelet could you please make the change so I can pull. Thanks!

[gchatelet]

@andralex : done !
Thank you all for reviewing my first submission : )