Does C++ supports variable template parameters ?

[Fan Yang]

I'm reading Modern C++ Design, and it is saying "Variable template
parameters simply don't exist."
But I find VC7.1 & VC8 support this feature.Who can tell me that which is
right -_-b
Many thanks.

 

[Alf P. Steinbach]

* Fan Yang:

I'm reading Modern C++ Design, and it is saying "Variable template
parameters simply don't exist."
But I find VC7.1 & VC8 support this feature.Who can tell me that which is
right -_-b

Have you heard about "context"? Some statements have context.

 

[dasjotre]

I'm reading Modern C++ Design, and it is saying "Variable template
parameters simply don't exist."

aka variadic template parameters

But I find VC7.1 & VC8 support this feature.Who can tell me that which is
right -_-b

I am not aware of any compiler supporting variadic template parameters.

 

[Fan Yang]

I am sorry, but what's the context?

* Fan Yang:

Have you heard about "context"? Some statements have context.

--
A: Because it messes up the order in which people normally read text.
Q: Why is it such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?

 

[Noah Roberts]

aka variadic template parameters


I am not aware of any compiler supporting variadic template parameters.

However, in practice this type of template can be created for some
upper number of N. Look at many examples from TR1 and Boost (tuple,
MPL, array, etc...). I'd be surprised if that book doesn't describe
these techneques but I haven't read the whole thing and certainly don't
know it by heart so I can't say what part the OP is talking about.

 

[Pete Becker]

However, in practice this type of template can be created for some
upper number of N. Look at many examples from TR1 and Boost (tuple,
MPL, array, etc...). I'd be surprised if that book doesn't describe
these techneques but I haven't read the whole thing and certainly don't
know it by heart so I can't say what part the OP is talking about.

Yes, you can do it, but it's ghastly. To support 0..N arguments you have
to write N+1 templates. Which is why C++0x will probably have
variable-length argument lists for templates.

--

-- Pete
Roundhouse Consulting, Ltd. (www.versatilecoding.com)
Author of "The Standard C++ Library Extensions: a Tutorial and
Reference." (www.petebecker.com/tr1book)

 

[Noah Roberts]

Noah Roberts wrote:

Yes, you can do it, but it's ghastly. To support 0..N arguments you have
to write N+1 templates. Which is why C++0x will probably have
variable-length argument lists for templates.

From what I've seen it's more like N/X with X default parameters. For

instance, mpl's vector is implemented in terms of a vector10, vector20,
vector30...which have 10, 20, and 30 default parameters.

 

[dasjotre]

However, in practice this type of template can be created for some
upper number of N. Look at many examples from TR1 and Boost (tuple,
MPL, array, etc...). I'd be surprised if that book doesn't describe
these techneques but I haven't read the whole thing and certainly don't
know it by heart so I can't say what part the OP is talking about.

In 'Modern C++ Design', he uses that technique to design type lists
and function wrappers. It is tedious and best done with something like
m4 or boost:reprocessor. Than again, having type lists, I see no
reason why you couldn't use them to emulate variadic template
parameters, up to the maximum length of the type lists.

But that is not a true variadic template like the ones you have in D.

 

[Pete Becker]

instance, mpl's vector is implemented in terms of a vector10, vector20,
vector30...which have 10, 20, and 30 default parameters.

I haven't looked at the details there, but from a glance it seems like
that's just part of a divide and conquer approach for the preprocessor,
where each of those templates in turn generates 10 template classes.

Sooner or later you have to decide which template arguments are real and
which ones aren't.

Of course, any program that needs to pass more than ten template
arguments is out of control. <g>

--

-- Pete
Roundhouse Consulting, Ltd. (www.versatilecoding.com)
Author of "The Standard C++ Library Extensions: a Tutorial and
Reference." (www.petebecker.com/tr1book)

 

[Noah Roberts]

I haven't looked at the details there, but from a glance it seems like
that's just part of a divide and conquer approach for the preprocessor,
where each of those templates in turn generates 10 template classes.

Possible, but it would be surprising. Compiler output from errors
indicate the final types contain the default values; at least when
using MPL vector. But I haven't done a lot of looking either.

Sooner or later you have to decide which template arguments are real and
which ones aren't.

Of course, any program that needs to pass more than ten template
arguments is out of control. <g>

Well, the basic dimensional analysis idiom requires something like 8 or
9 so more than 10 isn't really outside reason in metaprogramming.

 

[kwikius]

Of course, any program that needs to pass more than ten template
arguments is out of control. <g>

My compiler goes to 11...

(Just For the Spinal Tap fans ;-) )

regards
Andy Little

 

[Noah Roberts]

My compiler goes to 11...

(Just For the Spinal Tap fans ;-) )

What's funny is my amp DOES go to 11

 

[m_pahlevanzadeh]

use variadic