Re: bounds on position for text()

[bpabbott]

On Feb 09, 2011, at 03:52 PM, Michael D Godfrey <address@hidden> wrote:

On 02/09/2011 12:19 PM, logari81 wrote:

On Wed, 2011-02-09 at 12:07 -0800, Michael D Godfrey wrote:

> On 02/09/2011 10:24 AM, bpabbott wrote: 

> > On Feb 09, 2011, at 01:14 PM, Konstantinos Poulios
> > <address@hidden> wrote:
> > 
> > 

> > > On Mon, Feb 7, 2011 at 12:37 AM, Ben Abbott <address@hidden>
> > > wrote:

> > > > On Feb 6, 2011, at 6:35 PM, Michael D Godfrey wrote:
> > > >

> > > >> On 02/06/2011 03:21 PM, Konstantinos Poulios wrote:

> > > >>> On Mon, Feb 7, 2011 at 12:16 AM, Michael D Godfrey
> > > >>>
> > > >>> <address@hidden>
> > > >>>  wrote:
> > > >>>

> > > >>>> >

> > > >>>> On 02/06/2011 02:49 PM, logari81 wrote:
> > > >>>>

> > > >>>> >
> > > >>>> >

> > > >>>> On Sun, 2011-02-06 at 17:21 -0500, Ben Abbott wrote:
> > > >>>>

> > > >>>> >

> > > >>>>> >>

> > > >>>>> On Feb 6, 2011, at 4:02 PM, Michael D Godfrey wrote:
> > > >>>>>

> > > >>>>> >>

> > > >>>> >

> > > >>>>>> >> >

> > > >>>>>> A short while ago I created a bug report
> > > >>>>>>

> > > >>>>>> >> >

> > > >>>>>> (32311) which pointed out the fact that
> > > >>>>>>

> > > >>>>>> >> >

> > > >>>>>> text(x,y,"xxx') only works for x,y within the
> > > >>>>>>

> > > >>>>>> >> >

> > > >>>>>> axes. Matlab and legend allow writing the
> > > >>>>>>

> > > >>>>>> >> >

> > > >>>>>> string anywhere in the window.
> > > >>>>>>

> > > >>>>>> >> >
> > > >>>>>> >> >

> > > >>>>>> Since you just fixed legend, any chance
> > > >>>>>>

> > > >>>>>> >> >

> > > >>>>>> you could look at this?  Quite a lot of Matlab
> > > >>>>>>

> > > >>>>>> >> >

> > > >>>>>> code write annotations on plots using text
> > > >>>>>>

> > > >>>>>> >> >

> > > >>>>>> and positions this outside the axes.
> > > >>>>>>

> > > >>>>>> >> >
> > > >>>>>> >> >

> > > >>>>>> Michael
> > > >>>>>>

> > > >>>> >

> > > >>>>> >>
> > > >>>>> >>

> > > >>>>> I notice this work for the gnuplot backend, but not for the

> > > FLTK backend.

> > > >>>>>

> > > >>>>> >>
> > > >>>>> >>

> > > >>>>> It looks to me like something is checking the x/ydata values

> > > and not

> > > >>>>>

> > > >>>>> >>

> > > >>>>> rendering the result if it is outsize the axis limits.
> > > >>>>>

> > > >>>>> >>
> > > >>>>> >>

> > > >>>>> I'd expect this to be done in gl_render.cc, but don't see

> > > where it is.

> > > >>>>>

> > > >>>>> >>

> > > >>>>> I've cc'd Kostas, in case he knows where this is done.
> > > >>>>>

> > > >>>>> >>
> > > >>>>> >>

> > > >>>>> Ben
> > > >>>>>

> > > >>>>> >>

> > > >>>> >
> > > >>>> >

> > > >>>> printing of text objects is done in draw_axes_children in

> > > gl-render.cc

> > > >>>>

> > > >>>> >

> > > >>>> after the clipping property is checked.
> > > >>>>

> > > >>>> >
> > > >>>> >

> > > >>>> So it works if you just set clipping off, like:
> > > >>>>

> > > >>>> >
> > > >>>> >

> > > >>>> set(ht,'clipping','off')
> > > >>>>

> > > >>>> >
> > > >>>> >

> > > >>>> If it would make sense we could change the default clipping

> > > to on.

> > > >>>>

> > > >>>> >
> > > >>>> >

> > > >>>> Kostas
> > > >>>>

> > > >>>> >
> > > >>>> >

> > > >>>> I just tried this.  It does not work.
> > > >>>>

> > > >>>> >

> > > >>>> octave:1> ht=plot(1:200)
> > > >>>>

> > > >>>> >

> > > >>>> ht = -19.142
> > > >>>>

> > > >>>> >

> > > >>>> octave:2> set(ht,'clipping','off');
> > > >>>>

> > > >>>> >

> > > >>>> octave:3> text(200,-2,'this is it')
> > > >>>>

> > > >>>> >

> > > >>>> octave:4> set(ht,'clipping','on');
> > > >>>>

> > > >>>> >

> > > >>>> octave:5> text(200,-2,'this is it')
> > > >>>>

> > > >>>> >

> > > >>>> octave:6> text(200,2,'this is it')
> > > >>>>

> > > >>>> >

> > > >>>> octave:7> text(200,20,'this is it')
> > > >>>>

> > > >>>> >

> > > >>>> octave:8> text(201,20,'this is it')
> > > >>>>

> > > >>>> >

> > > >>>> octave:9>
> > > >>>>

> > > >>>> >

> > > >>>> =========================
> > > >>>>

> > > >>>> >

> > > >>>> None of the text commands with x,y outside axes prints.
> > > >>>>

> > > >>>> >
> > > >>>> >

> > > >>> the clipping property refers to the text object, so just try:
> > > >>>
> > > >>> ht=text(200,-2,'this is it');
> > > >>> set(ht,'clipping','on');
> > > >>>
> > > >>> Kostas
> > > >>>
> > > >>>

> > > >> Right. Now it "works."  Except, where you say above 'on'
> > > >> you really need 'off'.   But, it is not like Matlab and it is a

> > > bit of

> > > >> work to obtain the text handle and set 'off' for each text use.
> > > >> So, how about default  'off.'?    Someday it might be useful
> > > >> to have a global on/off, but maybe after the release?
> > > >>
> > > >> Michael

> > > >
> > > > Matlab R2010b has the default set to "off".
> > > >
> > > > I recommend we do the same.
> > > >
> > > > Ben
> > > >
> > > >

> > > 
> > > I have taken a look at this, but I couldn't find a very easy way
> > > to
> > > change the default clipping value in the backend. This is because
> > > the
> > > clipping property is shared among many different classes and the
> > > text
> > > class doesn't have an init function which would let us override
> > > any
> > > default values of the base class.
> > > 
> > > Here we have to take a decision, should we add a init function for
> > > the
> > > text class or should we set the default value for clipping to
> > > "off" in
> > > the frontend (e.g. text.m)? Maybe John could give us some advice
> > > on
> > > that.
> > > 
> > > Kostas
> > > 

> >  
> > It looks to me like the "clipping" property needs to be removed from
> > the BASE_PROPERTIES list and added to each object.
> > 
> > 
> > This would be done in graphics.h.in. I haven't tried this myself,
> > so ... ?
> > 
> > 
> > Ben
> > 
> > 

> You might take a look at bug 32311.  There is a patch attached to it
> which
> sets the default clipping 'off'.  This comes close to matching Matlab
> behavior.
> (It appears to match with gnuplot in any case, the differences are for
> fltk.)
> The only known difference  under fltk is that if clipping is 'off' and
> you shrink
> the axes data that is then outside the axes is still visible.  I am
> not sure that
> this is a bad thing.  It appears to me that clipping is already a
> property of objects,
> like text (you have to turn it off with respect to each ht = text() in
> order for
> it to be 'off' (or 'on') for that text string.
> 
> In any case, a good first step would be to commit the patch that
> changes the
> default clipping 'off'.  It may be, in any case, that making it more
> easily selectable
> will be useful.  It might be good to allow resetting the default.
> 
> Michael
> 

the patch that you have pointed to sets clipping to off for all classes
based on base_properties. I don't think that we want that.

Who is we?  This needs more review.  It would be reasonable to allow
users choices.  And, the "defaults" should be configurable.  Now, a
set(ht,...) is required after each ht=text() command  And, this is not
explained in the Manual...   

I don't think having the clipping set to off for all objects is a good idea.

For example ...

x = -10:10;

plot (x, x)

axis ([-1, 1, -1, 1])

What is the result if the line object isn't cropped?

My understanding is that Matlab crops all graphics objects, with the exception of text objects. Is there a reason why we should deviate from Matlab?

In any event, I like the idea of configurable defaults, and improvements to the manual. Should the former be added to the list of projects on the wiki?

http://wiki.octave.org/wiki.pl?Projects

Ben