Incompatible with matplotlib >= 3.6.0

[hassec]

The current code doesn't seem to work with more recent matplotlib versions.

Reproducer:

from omas import ODS

o = ODS().sample(ntimes=2)
o.plot_equilibrium_summary()

Error:

AttributeError                            Traceback (most recent call last)
Cell In[4], line 1
----> 1 e.plot_equilibrium_summary()

File ~/software/pyenv/versions/3.10.12/envs/mosaic/lib/python3.10/site-packages/omas/omas_plot.py:986, in equilibrium_summary(ods, time_index, time, fig, ggd_points_triangles, **kw)
    984         time_index = time_index[0]
    985     else:
--> 986         return ods_time_plot(
    987             equilibrium_summary, ods, time_index, time, fig=fig, ggd_points_triangles=ggd_points_triangles, ax=axs, **kw
    988         )
    990 ax = cached_add_subplot(fig, axs, 1, 3, 1)
    991 contour_quantity = kw.pop('contour_quantity', 'rho_tor_norm')

File ~/software/pyenv/versions/3.10.12/envs/mosaic/lib/python3.10/site-packages/omas/omas_plot.py:518, in ods_time_plot(ods_plot_function, ods, time_index, time, **kw)
    516 stime.on_changed(update)
    517 for time0 in time:
--> 518     axtime.axvline(time0, color=['r', 'y', 'c', 'm'][stime.cnt - 2])
    519 return {'stime': (stime, axtime), 'ax': axs}

AttributeError: 'Slider' object has no attribute 'cnt'

Additional info

The code in question is: https://github.com/gafusion/omas/blob/0cd6e21bb135ae6f1cb64d57e252c5a3c4e480ca/omas/omas_plot.py#L517-L519

which accesses Slider.cnt which was deprecated in version 3.4.0 and then removed in 3.6.0.

[orso82]

Thank you for pointing that out @hassec ! Do you know what it was substituted with?

[hassec]

@orso82 AFAIU it was removed and didn't get an equivalent replacement.

[orso82]

Perhaps slider.val ? can you please try it out?

>>> import matplotlib.widgets
>>> from matplotlib import pyplot
>>> ax=pyplot.ax()
>>> ax=pyplot.gca()
>>> slider = matplotlib.widgets.Slider(ax, "bla", 1, 10)
>>> slider.<TAB>
slider.active               slider.connect_event(       slider.eventson             slider.orientation          slider.slidermax            slider.valinit              slider.vline
slider.ax                   slider.disconnect(          slider.get_active()         slider.poly                 slider.slidermin            slider.valmax
slider.canvas               slider.disconnect_events()  slider.ignore(              slider.reset()              slider.track                slider.valmin
slider.closedmax            slider.drag_active          slider.label                slider.set_active(          slider.val                  slider.valstep
slider.closedmin            slider.drawon               slider.on_changed(          slider.set_val(             slider.valfmt               slider.valtext
>>> slider.val
1
>>> slider.set_val(2)
>>> slider.val
2

[hassec]

slider.val is the current value, and set_val sets that value.

Slider.cnt was a property that gave the quantity of registered callbacks as far as I understand. And while deprecated, it was implemented like:

https://github.com/matplotlib/matplotlib/blob/d04c8de4a89ab756a137ef146e8542e81041909f/lib/matplotlib/widgets.py#L156-L157

in 3.4 and 3.5 versions before it was removed.

I can try and have a look later to see if I an find a workaround that still works with both versions.

[orso82]

Yes, you're right. That would be great. Worse case we can always simplify with:

for time0 in time: 
     axtime.axvline(time0, color='r')

[hassec]

Hi @orso82,

sorry for the slow action, I've not had any time to check into this more until today. I'm not sure what the best fix would be because I'm not sure I really understand what

     axtime.axvline(time0, color=['r', 'y', 'c', 'm'][stime.cnt - 2]) 

is used for. I checked the tests but couldn't find a use case where stime.cnt has any value other than 2. When would that value be higher?

Do you have an example?

[orso82]

That would mark with different colors the times at which the data of multiple time dependent quantities are available (eg. equilibrium and ECH)

[hassec]

@orso82 do you have an example minimal reproducer or some code in OMFIT that I could turn into a test?

[orso82]

Not really, not now. I'll simply set it to axtime.axvline(time0, color='r'), which will be sufficient for most use cases.

[orso82]

running on matplotlib >= 3.6.0 is definitely more important

[hassec]

Thanks @orso82 for implementing the workaround :+1:

Is there already a rough ETA for when a new release OMAS will be made that includes this and the fixes since the last release?

[orso82]

We can release any time, but perhaps I'll let @smithsp @bechtt @jmcclena comment if we should squeeze in some of their latest PRs

[hassec]

That would be fantastic, thanks @orso82! If we could also get #262 into that release, it would mean we can remove all omas patches we currently have in place in our projects :)

[torrinba]

My PR is ready (just waiting for Joey's approval) and also requires updating the OMAS version used in OMFIT. It can come later if this is more pressing though

[jmcclena]

@orso82 I finished squeezing all my PRs in.

[orso82]

https://pypi.org/project/omas/0.92.0