|Re: [eigen] Does c++ use of "norm" bother anyone else?|
[ Thread Index |
| More lists.tuxfamily.org/eigen Archives
- To: eigen@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [eigen] Does c++ use of "norm" bother anyone else?
- From: Benoit Jacob <jacob.benoit.1@xxxxxxxxx>
- Date: Thu, 18 Jun 2009 13:40:34 +0200
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:received:in-reply-to:references :date:message-id:subject:from:to:content-type :content-transfer-encoding; bh=lZ+qb1C5TEVc6tCTCDoHx6ouSlFsGnNCR5wgsBOKQ0c=; b=K5hHYFviYeQ8u2exLEY6IIkMpjNfgdkNIeL28CqNHxSRdCvDJgLKf6/LAgTUCsA2I1 IpFMs1EFaQq6XEP5xnz2kl7oTfhC4qczZQ0hTAx3EZuk7AUHR8TPfS9FiUMSMNFAtCod M6Zxv+DWmLVO2RKAGb3+thttHh0y6BH0/eBeA=
- Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :content-type:content-transfer-encoding; b=Ll+rtqLIg/VZEADOS03Jym4AJbOxLq80xgDqYfM5P9l/PP7ClMedOZWOv5+oZ4PMsj CXaN+yGeclbyagXusvkIsnkCryI4s40lwJKjfIHKlTTfikAthBUkpQL+8vcIaMsbcQ8K 2bEAXwrKQdNyt/b3V62Il+cPQGSZIkbA3lAZY=
2009/6/18 Markus Moll <markus.moll@xxxxxxxxxxxxxxxx>
> On Thursday 18 June 2009 11:44:47 Tim Hutt wrote:
> > >> norm( complex<float>(3,4) ) == 25
> > Wow that's really bad. I've never heard of this norm being used in
> > mathematics. 'Norm' on its own virtually always means the L2 norm.
> Quite often it just means "any norm" (given that in finite dimensions, all
> norms are equivalent). But in math it certainly means that it satisfies all
> properties of a norm, which L2-squared certainly does not (not scaling
> properly, triangle inequality violated).
Indeed, std::complex norm() isn't a "norm" in this sense. But it still
is called "norm" in other mathematical contexts -- it's not just the
c++ people who were on crack, it's the whole mathematics tradition :)
(Optional part: the other sense of "norm" -- skip if not interested:
let K/k be a Galois extension, let x in K, then the norm of x with
respect to this extension is the product of the sigma(x) for all sigma
in the Galois group of K/k. So for complex numbers, they are a Galois
extension of the real numbers, of degree 2, the Galois group Gal(C/R)
has 2 elements, the identity and complex conjugation, so the norm of a
complex number x is the product x * conj(x) which is what
std::complex calls norm(). More references:
> So in my opinion, it is a good idea to correct this mistake in C++ and not
> support norm(), or maybe support it if a certain macro is defined (something
> like EIGEN_COMPLEX_PROVIDE_DEFECTIVE_NORM).
as I explain, it's not per se a mistake or a defect. But it still is
very misleading to have both. I don't have a strong opinion, it
depends whether our priority is to have a drop-in replacement for
std::complex, or to start from scratch our ideal complex class.
Actually I think that norm() isn't a method in std::complex, it's only
a global function in std:: , right ? Then I think that it's safe to
implement std::norm also for Eigen::Complex, as it's documented there
what it does. A user using it probably wants it to work as with
std::complex -- as the custom Eigen::Complex API would be implemented
as methods in Eigen::Complex, not as global functions in std.