Dont assume that unit attribute exists

[jenshnielsen]

Fixes #769 .

[AdriaanRol]

@jenshnielsen I tested your changes. They do not fix the problem.

Looking a bit more at the traceback I think it may have to do with having multiple subplots. I can see that the first plot get's created correctly (out of 2 I am trying to create) but the second one does not.

[jenshnielsen]

Do you get the same traceback? or a different one?

Can you inspect how the plot.traces dict looks normally there should be an element in the traces list for each subplot which in turn contains a config element which contains all the axis

[jenshnielsen]

I.e this is how it looks for me for a simple plot of a line graph and a 2d plot in

[{'config': {'figsize': (1200, 500), 'x': DataArray[6,6]: dac_ch2_set
   array([[ 0.,  1.,  2.,  3.,  4.,  5.],
          [ 0.,  1.,  2.,  3.,  4.,  5.],
          [ 0.,  1.,  2.,  3.,  4.,  5.],
          [ 0.,  1.,  2.,  3.,  4.,  5.],
          [ 0.,  1.,  2.,  3.,  4.,  5.],
          [ 0.,  1.,  2.,  3.,  4.,  5.]]), 'y': DataArray[6]: dac_ch1_set
   array([ 0.,  1.,  2.,  3.,  4.,  5.]), 'z': DataArray[6,6]: dmm_voltage
   array([[  25.,   52.,    2.,   17.,   73.,   68.],
          [  51.,   38.,   79.,   62.,   73.,  100.],
          [  60.,   72.,   20.,   37.,   56.,   51.],
          [   2.,   98.,   79.,   81.,   70.,   99.],
          [   4.,   71.,   10.,   39.,   11.,   30.],
          [  11.,    8.,   69.,   25.,   86.,   93.]])},
  'plot_object': {'cmap': 'hot',
   'hist': <ObjectProxy for process 24930, object 0x1db: <pyqtgraph.graphicsItems.HistogramLUTItem.HistogramLUTItem object at 0x117b7f948> >,
   'histlevels': (2.0, 100.0),
   'image': <ObjectProxy for process 24930, object 0x1d8: <pyqtgraph.graphicsItems.ImageItem.ImageItem object at 0x117b77a68> >,
   'scales': {'x': TransformState(translate=-0.5, scale=1.0, revisit=False),
    'y': TransformState(translate=-0.5, scale=1.0, revisit=False)}}},
 {'config': {'x': DataArray[6]: dac_ch1_set
   array([ 0.,  1.,  2.,  3.,  4.,  5.]), 'y': DataArray[6]: dac_ch1_set
   array([ 0.,  1.,  2.,  3.,  4.,  5.])},
  'plot_object': <ObjectProxy for process 24930, object 0x200: <pyqtgraph.graphicsItems.PlotDataItem.PlotDataItem object at 0x117b9c678> >}]

[AdriaanRol]

@jenshnielsen Below the traces and the plot.traces.
The problem occurs when in the same loop the second trace get's added.

I have to focus on my experiment now but I think this info should be useful.

Traceback :

d:\githubrepos\pycqed_py3\pycqed\measurement\measurement_control.py in measure(self, *kw)
    171     def measure(self, *kw):
    172         if self.live_plot_enabled():
--> 173             self.initialize_plot_monitor()
    174 
    175         for sweep_function in self.sweep_functions:

d:\githubrepos\pycqed_py3\pycqed\measurement\measurement_control.py in initialize_plot_monitor(self)
    570                                      subplot=j+1,
    571                                      color=color_cycle[j % len(color_cycle)],
--> 572                                      symbol='o', symbolSize=5)
    573                 self.curves.append(self.main_QtPlot.traces[-1])
    574                 j += 1

d:\githubrepos\qcodes\qcodes\plots\base.py in add(self, updater, *args, **kwargs)
     89         # TODO(giulioungaretti): replace with an explicit version, see expand trace
     90         self.expand_trace(args, kwargs)
---> 91         self.add_to_plot(**kwargs)
     92         self.add_updater(updater, kwargs)
     93 

d:\githubrepos\qcodes\qcodes\plots\pyqtgraph.py in add_to_plot(self, subplot, **kwargs)
    147         if prev_default_title == self.win.windowTitle():
    148             self.win.setWindowTitle(self.get_default_title())
--> 149         self.fixUnitScaling()
    150 
    151     def _draw_plot(self, subplot_object, y, x=None, color=None, width=None,

d:\githubrepos\qcodes\qcodes\plots\pyqtgraph.py in fixUnitScaling(self, startranges)
    547 
    548                     # set limits either from dataset or
--> 549                     setarr = self.traces[i]['config'][axis].ndarray
    550                     arrmin = None
    551                     arrmax = None

AttributeError: 'list' object has no attribute 'ndarray'

Inspecting traces:

[AdriaanRol]

More readable

[{'config': {'color': '#1f77b4',
   'symbol': 'o',
   'symbolSize': 5,
   'x': [0],
   'xlabel': 'Frequency',
   'xunit': 'Hz',
   'y': [0],
   'ylabel': 'Magn',
   'yunit': 'V'},
  'plot_object': <ObjectProxy for process 12880, object 0x1f: <pyqtgraph.graphicsItems.PlotDataItem.PlotDataItem object at 0x000000000564DE58> >}]

[jenshnielsen]

@AdriaanRol Thanks I will have a look

[jenshnielsen]

Your config object is quite different so the auto scaling will not work as is but I think this should be enough to keep the crash from happening

[jenshnielsen]

@AdriaanRol did you have a chance to test the latest version?

[AdriaanRol]

@jenshnielsen I didn't see you made any further changes. I'll have to check that out. Won't be today though, we are kind of deadline pressed at the moment.

On a related note, I think it would be good to add some unittests to the plotting to prevent these kind of issues in the future. I'd say they should only test if the code runs as testing for visual correctness is notoriously hard. If you are interested, I have a travis build script (based on the QCoDeS one) that also includes proper installation of pyqtgraph.

[jenshnielsen]

@AdriaanRol thanks thats fine whenever you have time. Unittesting is great but given that you are using it with a config object that is completely foreign to qcodes I don't see how we could have tested this without better visibility to what you are doing

[AdriaanRol]

@jenshnielsen . I'll try to create a unittest using only the default QCoDeS QtPlot API that reproduces the problem.

I think that should provide the visibility that you need and should prevent these kind of problems in the future.

[jenshnielsen]

Sounds good thanks

[AdriaanRol]

@jenshnielsen I've made a new PR containing a self contained test using only the default QCoDeS API that also contains all the work from your branch. Looking forward to your input.

[jenshnielsen]

Super-seeded by #805