Re: Merging in the gui branch

[Richard Crozier]

On 19/01/2012 15:15, John W. Eaton wrote:
> On 18-Jan-2012, Jordi Gutiérrez Hermoso wrote:
> 
> | The gui branch was supposed to be a temporary fix to not interfere
> | with the release of 3.6. Now that 3.6.0 is pretty much out and it
> | doesn't need the default branch anymore, I propose to make the bold
> | move to merge the gui branch onto default and close it.
> | 
> | This would mean our next stable release will probably have to have
> | some sort of GUI, perhaps still not compiled by default, but
> | available, sort of like what we have been doing with OpenGL plotting.
> | 
> | Objections?
> 
> At first I was hesitant to do this now, but I'm warming up to the
> idea.
> 
> I built Octave with the current GUI sources.  It looks promising, but
> I did not do a lot of testing.  From my brief experience with it, I
> see the following things that I would like to see done:
> 
>   * Remove the IRC client.  There are plenty of IRC clients around and
>     I see no need to have one built in to Octave.  We don't need to
>     add this extra code that we have to maintain.
> 
>   * Using the terminal and reading the documentation should be
>     possible at the same time.  The current tabbed interface does not
>     seem to allow this way of working.
> 
>   * Reading the documentation should not require a network
>     connection.  I also don't think we need to build a web browser
>     into Octave itself.  Similar to the IRC client issue, there are
>     plenty of web browsers to choose from.  We don't need to have
>     another one and embed it in Octave.  There's no need for us to add
>     this extra code that we have to maintain.
> 
>   * I noticed some redisplay problems when trying to use keyboard
>     commands for navigating the history.  For example, using C-p to
>     retrieve previous history items garbled the display.
> 
>   * If we are going with Qt for the GUI, then we need to use Qt for
>     the graphics windows instead of FLTK.
> 
>   * The code must be integrated with the current sources and build
>     system.  The GUI will be part of Octave, not a separate program,
>     and it must be possible to build it in a separate directory tree
>     as is usual with the GNU build system.
> 
> Is it likely that these things can be done before the next release?
> 
> I can see the value of the GUI for some things, but Octave should stay
> focused on numerical programming, not web browsing and IRC.  We have
> limited resources.  I think it will be enough work making the GUI work
> well without including these extra features.
> 
> jwe
> 

my two cents.

I worked on QtOctave, and think I made some worthwhile contributions
before it was superseded by the official GUI. I agree with everything
John says, except one thing which is regarding the documentation. It
should be possible to highlight a function in the GUI editor, hit F1 or
some equivalent and have the relevant documentation display in a help
browser of some kind, and certainly without requiring a network
connection. The help browser should also have a searchable index, with
the capability to add further functionality like this with time.
Another major improvement to the Matlab help browser would be the
ability to add your own notes to this documentation somehow. Can this
kind of thing be achieved with an external browser?

The help should also be able to display proper math symbols, ideally
described using latex markup or similar, something lacking in the online
docs at the moment incidentally.

In QtOctave you could have viewed the terminal and help at the same
time, just as you could effectively have viewed the editor or terminal
at the same time, or pop it out into an external window if you liked.
Clearly this functionality could and should be introduced to the
official GUI as well.

Richard

-- 
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.