| Re: [eigen] updated TODO for 3.0 |
[ Thread Index |
Date Index
| More lists.tuxfamily.org/eigen Archives
]
- To: eigen@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [eigen] updated TODO for 3.0
- From: Benoit Jacob <jacob.benoit.1@xxxxxxxxx>
- Date: Wed, 6 Oct 2010 07:41:54 -0400
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:received:received:in-reply-to :references:date:message-id:subject:from:to:content-type :content-transfer-encoding; bh=Wwdvz82dd1Ao85GSmSfT6coe/QbXRtaPpjdp3TqPIXA=; b=dfHwWg/e55JGdKXD0/9YxhyJITd2khpZmwmxdExIWcJvdTosIlPsCvR9QykYvPGC1m mGfWB+rBnpXj0cq8Pu1xeLl4/OSJnPAFGYrVye4CUF8YWqP6qYey0GWBLPsYIluh8uFE LjDPyeEUU0T395Di0xcSlElUUlrB971JzD/XQ=
- 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=aelrgr5aqPmX7IeBwwizU8an9GMfB4S76GXcCNn9VQtMsXVCRqQDMlPkQo8p6GXYkX HNrLVDXGpXCzaBiixhltWAqwCIhr8nO+pfZuGxQRXSgjbAqqNuZPGly1j4qTh4B3xrb5 A7+8lwcNuJWXTa56loctzfsjImY+6HC1pdoZE=
2010/10/6 Jitse Niesen <jitse@xxxxxxxxxxxxxxxxx>:
> On Tue, 28 Sep 2010, Benoit Jacob wrote:
>
>> ---> actually i'm thinking more and more that all these
>> "intermediate" decompositions (also Hessenberg) should rather be kept
>> internal. What's your opinion?
>
> The message is old (more than a week ago) but I still want to reply to it..
>
> I don't know that much about the bidiagonal decomposition, but the
> Hessenberg and Schur decompositions are used as a first step in many matrix
> algorithms. I use Schur for instance when computing matrix functions. So my
> immediate reaction is that we should expose these decompositions. Why do you
> want to keep them internal?
Because that would nicely reduce the number of classes in need of a
thorough API check before we release Eigen 3.0... but if you have a
need for them, OK let's keep them public. Anyway they're not in a too
bad shape.
>
> If we're not sure about the correct API, we can mark it as unstable for
> v3.0. Or did we decide that we would no longer have unstable / experimental
> bits except in unsupported?
Yes, we did. Making them be internal (e.g. putting them in a nested
"ei" or "internal" namespace) is the closest thing we can do. Eigen2's
approach of using the documentation to declare classes as internal
just didn't work: people ignored that.
Benoit
>
> Cheers,
> Jitse
>
>
>