Re: demo syntax and using dump_demos to compare to ML

[Ben Abbott]

On Jan 21, 2012, at 7:36 PM, Rik wrote:

> On 01/21/2012 09:20 AM, Ben Abbott wrote:
>> Rik,
>> 
>> I noticed you replaced the "..." line continuation with "\" in the demos.
> Actually, as far as I remember I was going the other way and trying to use
> "..." in place of "\".  There were a lot of changes though and I might have
> reversed one the other way.
>> Unfortunately, this breaks the "dump_demos" script Soren wrote to compare 
>> the gnuplot and fltk toolkits with Matlab (breaks the ability to run the 
>> demos using Matlab).
>> 
>> There may be some other changes the cause problems for using dump_demos to 
>> produce plots, via Matlab, as well.
>> 
>> Maybe it is a good idea to commit dump_demos.m and its companion 
>> html_compare.m and add the necessary parts to configure ?
> I certainly hadn't heard of this so I think other coders may also be
> unaware that you are trying to preserve Matlab syntax for plot demos. 
> Maybe, we could create something in to the build environment that tests
> whether rundemos() passes without warnings when Octave is run under
> --traditional?
>> That way more of the developers will be aware these scripts are used to look 
>> over the demo results, and that the plot demos need to avoid some Octave 
>> only syntax to allow plots to be produces using Matlab.
> How far do you want to extend this compatibility?  Do scripts under the
> image directory also need to be written for Matlab compatibility?  Should
> all demos that use plotting, such as treeplot in sparse/, be included?
> 
> Besides adjusting the spacing, I was also trying to get the demos to
> reliably reproduce their figures.  I added "clf;" to a lot of demos that
> use plotting, but that weren't in the plot/ directory where you had already
> made that change.  I also discovered, using rundemos, that a lot of the
> plots assume the default colormap which isn't a reliable assumption.  Thus,
> I'm going to have to build another big changeset to add "colormap
> ('default');" to those functions where it makes a difference.  I will bring
> back as much Matlab compatability as I can in that changeset.  But to which
> directories, besides plot/, should it apply?
> 
> --Rik

I'm only been comparing demos for the plot directory. So that should be 
sufficient for now. I'll try to put together a changeset to add dump_demo.m and 
its friends to the default branch. When I'm done I'll post it to the list for 
comments.

Ben