|Re: [eigen] news from the LU / rank / solving API front|
[ Thread Index |
| More lists.tuxfamily.org/eigen Archives
- To: eigen@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [eigen] news from the LU / rank / solving API front
- From: Benoit Jacob <jacob.benoit.1@xxxxxxxxx>
- Date: Sun, 18 Oct 2009 12:17:55 -0400
- 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; bh=gHGyQzhx1gGFi34HJrxI3JGUhoPogoTVDnTwdXHJ96Y=; b=YhjV101f5ybwYmAciFUUhh7aCGCnw8VKdsNfdEwnQEyo7z/a/6U7mJBkVJYsvhcG5A GBNdBtszDQ2DnXXwdfNoyJIed8t8Zq8eXTsctBrzUELLnMLm7iZuQLtRhFl5fO8yOBLz W0NvjudNmUkhCf4tpQ4pnt1GAx7BO9SS2EC+k=
- 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; b=kTPKyaeYtY51uw4pIOOgj5zTopHDD3a/+Wf34G+L4HEZdmFkvY/oOst0Zf9p6UGW67 MVBKkhRdakrNFbIzlrrXKiWitQIMumpASPFvJvnB/X+XF+scOynf5Un693c35v0R2PqT mD4zFdTSKtMld+MXU8rXYkUPitLxpaPVbQj9g=
2009/10/18 Hauke Heibel <hauke.heibel@xxxxxxxxxxxxxx>:
> Hey Benoit,
> The class looks good! I have only very few comments. First, I am wondering
> why useThreshold is returning a reference to the LU object and that actually
> only according to the signature since the current implementation returns
OK, it was a bug that i omitted the return statement; my intention was
to allow chained calls like
but indeed that is perhaps not a good idea as that syntax would
suggest that the threshold value is only used for that call. So I
think that the name useThreshold was a bit bad. How about renaming it
to setThreshold (so it's clear that it's a permanent change) and then
> I would simply make the return type void. Second, for the threshold
> retrieval, I woul make only 'RealScalar threshold() const' public and move
> the defaultThreshold() method to the private section or even consider
> removing it. When the default threshold is set, the thresold() function does
> return it - in case it is not set, I don't see why a 'public' user should be
> interested in it.
OK, that makes sense.
> Finally, I am wondering why all the variables are 'protected'. Do we expect
> somebody to inherit from LU? If that were the case it should have a virtual
> destructor. There is not harm in declaring the variables as protected - I
> was simply wondering.
It's just a habit that I have, to make variables protected, just in
case some day someone wants to inherit. It actually happens in
practice when people want to adjust the API e.g. to make porting
easier (I know that Thomas subclassed Matrix in the past).
Why should the destructor be virtual here? Is it important, in our
context, to allow referring to Derived objects as objects of the base
(here LU) type?