plotcontainer instances not empty

[Michael Graber]

hi,

i'm using traits and chaco to interactively analyze neuroscience data  
for my phd. i love it!
now i bumped into an issue i dunno how to deal with it:

for better understanding see code attached.

i defined a TimeseriesPlot instance consisting of a ArrayPlotData and  
a Plotcontainer instance and some functions to easily add different  
plots to a container.

for example i can do the following:

---------------------------

In [125]: data = []

In [126]: data.append(numpy.random.rand(100))

In [127]: data.append(numpy.random.rand(100))

In [128]: data.append(numpy.random.rand(100))

In [129]: tsp = sc.TimeseriesPlot()

In [132]: tsp.add_single_trace_plot(data[0], 10., 'test')

In [134]: tsp.add_temp_distribution_plot(numpy.array(data), 10,  
'testdistr')

In [135]: tsp.configure_traits()

---------------------------

this nicely plots data into the plotcontainer.
but two things i do not understand:

-> if i create a second TimeseriesPlot instance, lets say

In [136]: tsa = sc.TimeseriesPlot()

In [137]: tsa.configure_traits()

the new plotcontainer already contains the previous plots?! why is that?
i would like to loop over loads of traces from different origins and  
save the plots as pdfs and pngs. but this really messes it up!
how do i get empty an empty plotcontainer and plotdata instance every  
time i create a TimeseriesPlot instance?

-> on my macbook when calling configure_traits() i first have to  
resize my window to see a plot, before that i only have an gray, empty  
window. why? what to do?
(this is not the case on my ubuntu workstation)

thanks a lot for hints!
michael







_______________________________________________
Enthought-Dev mailing list
[hidden email]
https://mail.enthought.com/mailman/listinfo/enthought-dev

[Peter Wang]

On Sun, Nov 14, 2010 at 5:04 PM, Michael Graber
<[hidden email]> wrote:
> the new plotcontainer already contains the previous plots?! why is that?
> i would like to loop over loads of traces from different origins and save
> the plots as pdfs and pngs. but this really messes it up!
> how do i get empty an empty plotcontainer and plotdata instance every time i
> create a TimeseriesPlot instance?

Your problem is that your plotdata and plotcontainer attributes are
*class* attributes, and not *traits*.  This means that all instances
of the class share those attributes.  What you really want to do is
use an Instance trait, thusly:

class TimeseriesPlot(HasTraits):
  plotdata = Instance(ArrayPlotData, ())
  plotcontainer = Instance(VPlotContainer, kw=dict(resizable="hv",
bgcolor="white", fill_padding=True, padding=PLOTPADDING))


> -> on my macbook when calling configure_traits() i first have to resize my
> window to see a plot, before that i only have an gray, empty window. why?
> what to do?
> (this is not the case on my ubuntu workstation)

I am not sure why this is happening to you... it doesn't happen to the
chaco demos that use the ComponentEditor and .configure_traits()...
Perhaps try fixing the class attribute/instance trait issue above, and
see if that helps matters?

-Peter
_______________________________________________
Enthought-Dev mailing list
[hidden email]
https://mail.enthought.com/mailman/listinfo/enthought-dev

[Michael Graber]

>> the new plotcontainer already contains the previous plots?! why is  
>> that?
>> i would like to loop over loads of traces from different origins  
>> and save
>> the plots as pdfs and pngs. but this really messes it up!
>> how do i get empty an empty plotcontainer and plotdata instance  
>> every time i
>> create a TimeseriesPlot instance?
>
> Your problem is that your plotdata and plotcontainer attributes are
> *class* attributes, and not *traits*.  This means that all instances
> of the class share those attributes.  What you really want to do is
> use an Instance trait, thusly:
>
> class TimeseriesPlot(HasTraits):
>  plotdata = Instance(ArrayPlotData, ())
>  plotcontainer = Instance(VPlotContainer, kw=dict(resizable="hv",
> bgcolor="white", fill_padding=True, padding=PLOTPADDING))

thanks peter, this solved the issue with non-empty containers!
but i must admit that do not fully understand the distinction between  
class attributes and traits and also the traits type Any. is there  
some documentation about that somewhere?

>
>> -> on my macbook when calling configure_traits() i first have to  
>> resize my
>> window to see a plot, before that i only have an gray, empty  
>> window. why?
>> what to do?
>> (this is not the case on my ubuntu workstation)
>
> I am not sure why this is happening to you... it doesn't happen to the
> chaco demos that use the ComponentEditor and .configure_traits()...
> Perhaps try fixing the class attribute/instance trait issue above, and
> see if that helps matters?

this issue was not fixed by your suggested change.

i could not find an example that uses the ComponentEditor  
and .configure_traits() directly. all the examples i checked use an  
enable.api Window somehow to display plots. this works. but these  
examples have the same problem if you use the .configure_traits()  
method on the plot class. for example bar_plot.py https://svn.enthought.com/enthought/browser/Chaco/trunk/examples/bar_plot.py 
  :

from bar_plot import Demo
demo = Demo()
demo.configure_traits()

the enable Window seems to go very deep down into wx handling ..

--------

now that the TimeseriesPlot works i want to loop over many traces,  
generate a plot and save it as a pdf or a bmp.
but if i do not use the configure_traits() method on the timeseries  
object the saved pdf is empty. however, i do really not want to have  
hundreds of windows being opened!

is there a way to 'render the plot' without calling configure_traits()

now, in the case of my MacBook i have to use .configure_traits() and  
then resize the window, otherwise the pdf is empty. this can't be the  
way ;)

btw. regarding the other pdf mailing list thread: saving pdf of a  
VPlotContainer like in my code works perfectly, apart from different  
line_styles ..

thanks for your help,
michael




_______________________________________________
Enthought-Dev mailing list
[hidden email]
https://mail.enthought.com/mailman/listinfo/enthought-dev