Re: [Boost-Users] norm_inf type-related error

From: jhr.walter_at_[hidden]
Date: 2003-06-05 03:52:02

----- Original Message -----
From: cerenoc
To: Boost-Users_at_[hidden]
Sent: Thursday, June 05, 2003 7:50 AM
Subject: [Boost-Users] norm_inf type-related error

> Can somebody help me with this?  The following simple piece of code
> compiles fine if m, v, and s are declared as type <double> but not
> for type <int> as shown.  Why is that?

Mathematically speaking, integers usually don't fulfill the requirements for
the field of a vector space.

> Can I cast vectors and
> matrices from int to double easily?

Copy yes, cast no. In your case something like

> ...define some matrix <int>m...
> boost::numeric::ublas::vector<int> v = row(m, 1);
> int s = norm_inf(v);

double d = norm_inf (boost::numeric::ublas::vector<double>(v));

should work.

> The complaint is as follows:
>
> /tmp/ccA7KBD2.o: In function
> `boost::numeric::ublas::type_traits<int>::norm_inf(int const &)':

[snip more unresolved references]

This is a bug, which is fixed in CVS. You now should get a runtime error
then (oops: I just realized that it probably would be better to generate a
compile time error ;-(

Thanks for your feedback,

Joerg