From help-octave-request at bevo dot che dot wisc dot edu Tue Apr 29 16:47:09 2003
Subject: Re: Save-load problem
From: "Henry F. Mollet" <mollet at pacbell dot net>
To: "John W. Eaton" <jwe at bevo dot che dot wisc dot edu>
CC: Pablo Barrera Gonz=?ISO-8859-1?B?4Q==?=lez <pbarrera at tsc dot uc3m dot es>,    Octave_post <help-octave@bevo.che.wisc.edu>
Date: Tue, 29 Apr 2003 14:50:19 -0700

>From a beginners point of view, I had to scratch my head if saving and
re-loading does not produce the same results. Obviously, I was not aware
that Octave was saving a range rather than the 1x129 matrix I anticipated. I
presume that's what MATLAB does as I was merely trying an example in the
introduction of a MATLAB text and combined it with save/load on the previous
page. I thought that the "range" statement (A=0:pi/16:8*pi;) was a quick way
to produce the matrix.

Four work-around solutions were provided:
1) save -float-binary session3.mat
2) Add default_save_format="ascii"; to ~/.octaverc
3) a(1)=a(1)            <- the trick
4) save_precision = 16

I do not doubt JWE's response that there are good reasons to save as a range
but should this be the default for beginners to trip over? None of
work-around solutions were obvious to me. Some of them also have potential
problems according to John's reply. Perhaps then default save should mean
save -matrix, and one would use save -range for saving the range?
Henry



on 4/29/03 11:33 AM, John W. Eaton at jwe at bevo dot che dot wisc dot edu wrote:

> On 29-Apr-2003, Henry F. Mollet <mollet at pacbell dot net> wrote:
> 
> | But I still don't understand. If it were a matter of loosing precision,
> | would not the elements of vector A be a little different.
> 
> The problem is that for a range, the base, limit, and increment of the
> range are stored instead of all the individual elements.  Then when
> you load the file with this information and recompute the size, Octave
> has to recompute the size given the numbers read from the file.  If
> they are off slightly, you can generate a different number of
> elements.
> 
> The following worked for me:
> 
> save_precision = 16
> A = 0:pi/16:8*pi;
> save foo.dat A
> clear
> load foo.dat
> size (A)  ==> 1, 129.
> 
> but this is not a very good solution (you might also be able to get
> bad results if you use a value for save_precision that is too large).
> 
> At this point, probably it would be best for Octave to default to a
> binary format.  I don't think it makes sense to save all elements of
> the range, but if that's what you want, I think you will have to
> either modify Octave or use the Matlab binary format which I don't
> think supports ranges as a special case.
> 
> I like the space savings that are possible by representing ranges
> using the base, limit, and increment, and it would be nice to preserve
> that when saving and loading files, but perhaps when saving in ASCII,
> it would be best to convert to matrices.
> 
> Comments?
> 
> jwe



-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.

Octave's home on the web:  http://www.octave.org
How to fund new projects:  http://www.octave.org/funding.html
Subscription information:  http://www.octave.org/archive.html
-------------------------------------------------------------