'octave' and 'matio'

classic Classic list List threaded Threaded
14 messages Options
Reply | Threaded
Open this post in threaded view
|

'octave' and 'matio'

Sergei Steshenko
Hello,

I've recently also started building 'scilab' and doing it stumbled upon
'matio' library: http://sourceforge.net/projects/matio - 'scilab' uses it.

Is there a good reason for 'octave' not to use it ?

Thanks,
  Sergei.
_______________________________________________
Help-octave mailing list
[hidden email]
https://mailman.cae.wisc.edu/listinfo/help-octave
Reply | Threaded
Open this post in threaded view
|

Re: 'octave' and 'matio'

Jordi Gutiérrez Hermoso-2
On 3 August 2011 01:10, Sergei Steshenko <[hidden email]> wrote:
> I've recently also started building 'scilab' and doing it stumbled upon
> 'matio' library: http://sourceforge.net/projects/matio - 'scilab' uses it.
>
> Is there a good reason for 'octave' not to use it ?

Pardon my ignorance, but what are we missing in Octave regarding mat files?
_______________________________________________
Help-octave mailing list
[hidden email]
https://mailman.cae.wisc.edu/listinfo/help-octave
Reply | Threaded
Open this post in threaded view
|

Re: 'octave' and 'matio'

Sergei Steshenko


--- On Wed, 8/3/11, Jordi Gutiérrez Hermoso <[hidden email]> wrote:

> From: Jordi Gutiérrez Hermoso <[hidden email]>
> Subject: Re: 'octave' and 'matio'
> To: "Sergei Steshenko" <[hidden email]>
> Cc: [hidden email]
> Date: Wednesday, August 3, 2011, 12:08 AM
> On 3 August 2011 01:10, Sergei
> Steshenko <[hidden email]>
> wrote:
> > I've recently also started building 'scilab' and doing
> it stumbled upon
> > 'matio' library: http://sourceforge.net/projects/matio - 'scilab' uses
> it.
> >
> > Is there a good reason for 'octave' not to use it ?
>
> Pardon my ignorance, but what are we missing in Octave
> regarding mat files?
>

I have no idea - just asked. I feel some NIH syndrome smell though.

Also, you could probably shed some light on

https://mailman.cae.wisc.edu/pipermail/help-octave/2011-August/046850.html

- if it is related to Matlab IO.

Thanks,
  Sergei.
_______________________________________________
Help-octave mailing list
[hidden email]
https://mailman.cae.wisc.edu/listinfo/help-octave
Reply | Threaded
Open this post in threaded view
|

Re: 'octave' and 'matio'

Jordi Gutiérrez Hermoso-2
2011/8/3 Sergei Steshenko <[hidden email]>:

>
>
> --- On Wed, 8/3/11, Jordi Gutiérrez Hermoso <[hidden email]> wrote:
>
>> From: Jordi Gutiérrez Hermoso <[hidden email]>
>> Subject: Re: 'octave' and 'matio'
>> To: "Sergei Steshenko" <[hidden email]>
>> Cc: [hidden email]
>> Date: Wednesday, August 3, 2011, 12:08 AM
>> On 3 August 2011 01:10, Sergei
>> Steshenko <[hidden email]>
>> wrote:
>> > I've recently also started building 'scilab' and doing
>> it stumbled upon
>> > 'matio' library: http://sourceforge.net/projects/matio - 'scilab' uses
>> it.
>> >
>> > Is there a good reason for 'octave' not to use it ?
>>
>> Pardon my ignorance, but what are we missing in Octave
>> regarding mat files?
>>
>
> I have no idea - just asked. I feel some NIH syndrome smell though.

Hey, Octave was here first; if anyone it's Matio who has NIH, not
Octave. In fact, Octave was explicitly born without NIH, since it was,
and still is, primarily a glorified wrapper of numerical Fortran
libraries.

Anyways, my question was, we can already read and write mat files, are
we doing it incorrectly or insufficiently enough to warrant throwing
out out code or merging it with theirs? What's to be gained by this
development?

> Also, you could probably shed some light on
>
> https://mailman.cae.wisc.edu/pipermail/help-octave/2011-August/046850.html
>
> - if it is related to Matlab IO.

No, oct2mat has nothing to do with mat file IO. I have no further
insight on this matter.

- Jordi G. H.
_______________________________________________
Help-octave mailing list
[hidden email]
https://mailman.cae.wisc.edu/listinfo/help-octave
Reply | Threaded
Open this post in threaded view
|

Re: 'octave' and 'matio'

David Bateman
On 08/03/2011 09:43 AM, Jordi Gutiérrez Hermoso wrote:
> Anyways, my question was, we can already read and write mat files, are
> we doing it incorrectly or insufficiently enough to warrant throwing
> out out code or merging it with theirs? What's to be gained by this
> development?
>


Matio can read/write matlab V7.3 files, which are a matlab wrapper
around HDF5. Octave currently can't do that. So this is one major reason
to convert to use matio.

There's also a residual bug in Octave for zipped matfile's that I can't
seem to fix, see

http://savannah.gnu.org/bugs/?32167

that would probably be fixed by using matio.

However, matio can't treat matlab objects, such as anonymous functions
and so if we convert to matio we'd have to reimplement objects.

D.
_______________________________________________
Help-octave mailing list
[hidden email]
https://mailman.cae.wisc.edu/listinfo/help-octave
Reply | Threaded
Open this post in threaded view
|

Re: 'octave' and 'matio'

Sergei Steshenko


--- On Wed, 8/3/11, David Bateman <[hidden email]> wrote:

> From: David Bateman <[hidden email]>
> Subject: Re: 'octave' and 'matio'
> To: [hidden email]
> Date: Wednesday, August 3, 2011, 5:57 AM
> On 08/03/2011 09:43 AM, Jordi
> Gutiérrez Hermoso wrote:
> > Anyways, my question was, we can already read and
> write mat files, are
> > we doing it incorrectly or insufficiently enough to
> warrant throwing
> > out out code or merging it with theirs? What's to be
> gained by this
> > development?
> >
>
>
> Matio can read/write matlab V7.3 files, which are a matlab
> wrapper
> around HDF5. Octave currently can't do that. So this is one
> major reason
> to convert to use matio.
>
> There's also a residual bug in Octave for zipped matfile's
> that I can't
> seem to fix, see
>
> http://savannah.gnu.org/bugs/?32167
>
> that would probably be fixed by using matio.
>
> However, matio can't treat matlab objects, such as
> anonymous functions
> and so if we convert to matio we'd have to reimplement
> objects.
>
> D.
> _______________________________________________
> Help-octave mailing list
> [hidden email]
> https://mailman.cae.wisc.edu/listinfo/help-octave
>

Scilab team prefers meanwhile to use version 1.3.7 - see

http://bugzilla.scilab.org/show_bug.cgi?id=9792#c5
.

Regards,
  Sergei.
_______________________________________________
Help-octave mailing list
[hidden email]
https://mailman.cae.wisc.edu/listinfo/help-octave
Reply | Threaded
Open this post in threaded view
|

Re: 'octave' and 'matio'

Sergei Steshenko


--- On Wed, 8/3/11, Sergei Steshenko <[hidden email]> wrote:

[snip]

> Scilab team prefers meanwhile to use version 1.3.7 - see
>
> http://bugzilla.scilab.org/show_bug.cgi?id=9792#c5
> .
>
> Regards,
>   Sergei.
>

Sorry - matio-1.3.4, and not 1.3.7.

Regards,
  Sergei.
_______________________________________________
Help-octave mailing list
[hidden email]
https://mailman.cae.wisc.edu/listinfo/help-octave
Reply | Threaded
Open this post in threaded view
|

Re: 'octave' and 'matio'

Sergei Steshenko
In reply to this post by Sergei Steshenko
--- On Wed, 8/3/11, David Bateman <[hidden email]> wrote:

> From: David Bateman <[hidden email]>
> Subject: Re: 'octave' and 'matio'
> To: [hidden email]
> Date: Wednesday, August 3, 2011, 5:57 AM
> On 08/03/2011 09:43 AM, Jordi
> Gutiérrez Hermoso wrote:
> > Anyways, my question was, we can already read and
> write mat files, are
> > we doing it incorrectly or insufficiently enough to
> warrant throwing
> > out out code or merging it with theirs? What's to be
> gained by this
> > development?
> >
>
>
> Matio can read/write matlab V7.3 files, which are a matlab
> wrapper
> around HDF5. Octave currently can't do that. So this is one
> major reason
> to convert to use matio.
>
> There's also a residual bug in Octave for zipped matfile's
> that I can't
> seem to fix, see
>
> http://savannah.gnu.org/bugs/?32167
>
> that would probably be fixed by using matio.
>
> However, matio can't treat matlab objects, such as
> anonymous functions
> and so if we convert to matio we'd have to reimplement
> objects.
>
> D.
> _______________________________________________
> Help-octave mailing list
> [hidden email]
> https://mailman.cae.wisc.edu/listinfo/help-octave
>
By the way, to me HDF5 looks horrendously suspicious - one may have a look
at the attached file produced by

grep warning /mnt/sdb8/sergei/AFSWD_debug/build/hdf5-1.8.7/make.log | cut -d: -f5 | sort | uniq > hdf5-1.8.7_warnings.txt

command.

Regards,
  Sergei.
_______________________________________________
Help-octave mailing list
[hidden email]
https://mailman.cae.wisc.edu/listinfo/help-octave

hdf5-1.8.7_warnings.txt (57K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: 'octave' and 'matio'

John W. Eaton
Administrator
In reply to this post by David Bateman
On  3-Aug-2011, David Bateman wrote:

| On 08/03/2011 09:43 AM, Jordi Gutiérrez Hermoso wrote:
| > Anyways, my question was, we can already read and write mat files, are
| > we doing it incorrectly or insufficiently enough to warrant throwing
| > out out code or merging it with theirs? What's to be gained by this
| > development?
| >
|
|
| Matio can read/write matlab V7.3 files, which are a matlab wrapper
| around HDF5. Octave currently can't do that. So this is one major reason
| to convert to use matio.
|
| There's also a residual bug in Octave for zipped matfile's that I can't
| seem to fix, see
|
| http://savannah.gnu.org/bugs/?32167
|
| that would probably be fixed by using matio.

Alternatively, we could maybe look at how matio solves the problem and
adapt that to Octave.

Switching to matio would not be trivial.  Currently, our code works
directly with octave_value objects.  To use matio, we would have to
translate the matio representation of Matlab values to octave_value
objects.

| However, matio can't treat matlab objects, such as anonymous functions
| and so if we convert to matio we'd have to reimplement objects.

So maybe it would be better to just extract portions of matio that do
things that we don't currently do, and adapt the code for Octave.

jwe
_______________________________________________
Help-octave mailing list
[hidden email]
https://mailman.cae.wisc.edu/listinfo/help-octave
Reply | Threaded
Open this post in threaded view
|

Re: 'octave' and 'matio'

John W. Eaton
Administrator
In reply to this post by Jordi Gutiérrez Hermoso-2
On  3-Aug-2011, Jordi Gutiérrez Hermoso wrote:

| > Also, you could probably shed some light on
| >
| > https://mailman.cae.wisc.edu/pipermail/help-octave/2011-August/046850.html
| >
| > - if it is related to Matlab IO.
|
| No, oct2mat has nothing to do with mat file IO. I have no further
| insight on this matter.

My understanding is that oct2mat provides a set of functions that are
available in Octave but not Matlab, along with a way to convert
functions written for Octave so that they can run in Matlab.  There
are two parts to this.  First, a translation of some simple things
like comment and string quote styles, then some functions to provide
things that Octave has and Matlab doesn't.  As I understand it, you
aren't supposed to install those funtions with Octave, since Octave
already has them.  You are supposed to use them if you want to run
Octave code with some interpreter that doesn't provide those
functions.

jwe
_______________________________________________
Help-octave mailing list
[hidden email]
https://mailman.cae.wisc.edu/listinfo/help-octave
Reply | Threaded
Open this post in threaded view
|

Re: 'octave' and 'matio'

Sergei Steshenko


--- On Wed, 8/3/11, John W. Eaton <[hidden email]> wrote:

> From: John W. Eaton <[hidden email]>
> Subject: Re: 'octave' and 'matio'
> To: "Jordi Gutiérrez Hermoso" <[hidden email]>
> Cc: "Sergei Steshenko" <[hidden email]>, [hidden email]
> Date: Wednesday, August 3, 2011, 10:54 AM
> On  3-Aug-2011, Jordi Gutiérrez
> Hermoso wrote:
>
> | > Also, you could probably shed some light on
> | >
> | > https://mailman.cae.wisc.edu/pipermail/help-octave/2011-August/046850.html
> | >
> | > - if it is related to Matlab IO.
> |
> | No, oct2mat has nothing to do with mat file IO. I have no
> further
> | insight on this matter.
>
> My understanding is that oct2mat provides a set of
> functions that are
> available in Octave but not Matlab, along with a way to
> convert
> functions written for Octave so that they can run in
> Matlab.  There
> are two parts to this.  First, a translation of some
> simple things
> like comment and string quote styles, then some functions
> to provide
> things that Octave has and Matlab doesn't.  As I
> understand it, you
> aren't supposed to install those funtions with Octave,
> since Octave
> already has them.  You are supposed to use them if you
> want to run
> Octave code with some interpreter that doesn't provide
> those
> functions.
>
> jwe
>

I don't understand. If I install the _package_, some functions get
overridden. I.e. I do not control the functions, I just install the
package.

Is it OK in this case that some functions get overridden ?

Thanks,
  Sergei.
_______________________________________________
Help-octave mailing list
[hidden email]
https://mailman.cae.wisc.edu/listinfo/help-octave
Reply | Threaded
Open this post in threaded view
|

Re: 'octave' and 'matio'

John W. Eaton
Administrator
On  3-Aug-2011, Sergei Steshenko wrote:

| I don't understand. If I install the _package_, some functions get
| overridden. I.e. I do not control the functions, I just install the
| package.
|
| Is it OK in this case that some functions get overridden ?

The package is broken.  It should not install its functions with
Octave.

jwe
_______________________________________________
Help-octave mailing list
[hidden email]
https://mailman.cae.wisc.edu/listinfo/help-octave
Reply | Threaded
Open this post in threaded view
|

Re: 'octave' and 'matio'

Sergei Steshenko


--- On Wed, 8/3/11, John W. Eaton <[hidden email]> wrote:

> From: John W. Eaton <[hidden email]>
> Subject: Re: 'octave' and 'matio'
> To: "Sergei Steshenko" <[hidden email]>
> Cc: "Jordi Gutiérrez Hermoso" <[hidden email]>, "John W. Eaton" <[hidden email]>, [hidden email]
> Date: Wednesday, August 3, 2011, 11:22 AM
> On  3-Aug-2011, Sergei Steshenko
> wrote:
>
> | I don't understand. If I install the _package_, some
> functions get
> | overridden. I.e. I do not control the functions, I just
> install the
> | package.
> |
> | Is it OK in this case that some functions get overridden
> ?
>
> The package is broken.  It should not install its
> functions with
> Octave.
>
> jwe
>

Thanks for confirming the bug.

I think the package should _immediately_ be removed from sourceforge
(until it is fixed) because it breaks core Octave functionality.

I have already removed it from the list of packages I'm installing from
sourceforge - decided to be on the safe side.

Regards,
  Sergei.
_______________________________________________
Help-octave mailing list
[hidden email]
https://mailman.cae.wisc.edu/listinfo/help-octave
Reply | Threaded
Open this post in threaded view
|

Re: 'octave' and 'matio'

Jordi Gutiérrez Hermoso-2
On 3 August 2011 13:28, Sergei Steshenko <[hidden email]> wrote:

> I think the package should _immediately_ be removed from sourceforge
> (until it is fixed) because it breaks core Octave functionality.

Other packages do that, sometimes intentionally (e.g. the nan
package). Installing all of Octave-Forge seldom makes sense. This is
why it was broken down into individual packages, so you could install
the packages that were important to you.

- Jordi G. H.
_______________________________________________
Help-octave mailing list
[hidden email]
https://mailman.cae.wisc.edu/listinfo/help-octave