Re: Autoloading of toolboxes?

[David Bateman]

John W. Eaton wrote:
> On 29-Jan-2007, David Bateman wrote:
>
> | One major difference between toolboxes installed with pkg and similar
> | toolboxes under matlab is the need to explicitly load the toolboxes with
> | the "pkg load <package>" or "pkg load all" commands. In matlab an
> | installed toolboxes is always available if a license is available.
>
> If the packages are installed, then I think Octave should load them,
> and users should not have to do anything more to enable the packages.
> This is separate from the issue of whether the packages should all be
> bundled and installed together.
>   

Ok, then a "pkg load all" on startup is all that is needed.

> | This
> | might confuse some users, and so I wonder whether we should include the line
> | 
> | ## PKG_ADD: pkg("load","all");
> | 
> | in pkg.m... Thoughts?
>
> Is this the best place for it, or should this statement (or something
> similar) be added to the main octaverc file?
>
> Should the diretories for add-on packages be placed at the end of the
> load path by default?  I think that it is confusing for users when
> they install a package that replaces Octave functions and things break
> because the replacement functions work differently.  Remember the code
> in FIXES from Octave Forge?  I think we want to avoid that kind of
> thing in the future.  Maybe we need to have some kind of warning
> so that users know when a package shadows existing functions.  If we
> opt for a warning it can be optional but I think it should be on by
> default.
>   
I think shadowed functions should be first in the load path, otherwise
they will never be used. However, I think we have eliminated all of
these shadowed functions like FIXES/ and I don't think they should ever
come back. I kept the toggle_graceplot function in the graceplot package
and so simply loading graceplot won't shadow the plotting functions...

Things that might be an issue are TSA and NaN packages that replace
existing functions and behave differently. If these are installed and we
do a "pkg load all" at startup then they will always be used. I suppose
the current state of affairs is similar with monolithic octave-forge.
However, shouldn't we supply a means that the functionality can be
disabled? I suppose then that if they are installed they are used until
the user emits a "pkg unload tsa" or "pkg unload nan" command. I can
live with that rather than the "auto", "manual" load function proposed
by Bill..

Regards
David



-- 
David Bateman                                address@hidden
Motorola Labs - Paris                        +33 1 69 35 48 04 (Ph) 
Parc Les Algorithmes, Commune de St Aubin    +33 6 72 01 06 33 (Mob) 
91193 Gif-Sur-Yvette FRANCE                  +33 1 69 35 77 01 (Fax) 

The information contained in this communication has been classified as: 

[x] General Business Information 
[ ] Motorola Internal Use Only 
[ ] Motorola Confidential Proprietary