Bug report

Bug summary

In NeXpy, a PyQt application that subclasses NavigationToolbar2QT in backends/backend_qt5.py, a call to _init_toolbar fails because of a TypeError, which is generated by an internal conflict between positional and keyword arguments within the Matplotlib code. This affects Matplotlib v3.2.0, but not v3.1.2.

The issue is caused by a call to _icon in _init_toolbar, i.e., self._icon(image_file + '.png', icon_color), which has the color defined as a positional argument. However, the _icon function is defined as def _icon(self, name, color=None):

I don't understand why this is a problem. I thought that Python is tolerant of keyword arguments being defined as positional arguments when there is no ambiguity. I also have no idea why it doesn't appear to affect the toolbar initialization when opening a PyQt window in pyplot, e.g., within a Jupyter qtconsole client, which presumably calls the same function. I would be grateful if anyone else does understand it and can suggest a work-around. Nevertheless, I presume that the Matplotlib code should respect the use of keyword arguments.

Traceback

~/miniconda3/lib/python3.7/site-packages/nexpy/gui/plotview.py in _init_toolbar(self)
   3385             ('Add', 'Add plot data to tree', 'hand', 'add_data')
   3386                 )
-> 3387         super(NXNavigationToolbar, self)._init_toolbar()
   3388         self._actions['set_aspect'].setCheckable(True)
   3389         for action in self.findChildren(QtWidgets.QAction):

~/miniconda3/lib/python3.7/site-packages/matplotlib/backends/backend_qt5.py in _init_toolbar(self)
    698                 self.addSeparator()
    699             else:
--> 700                 a = self.addAction(self._icon(image_file + '.png', icon_color),
    701                                    text, getattr(self, callback))
    702                 self._actions[callback] = a

TypeError: _icon() takes 2 positional arguments but 3 were given

Matplotlib version

• Operating system: Mac OS X
• Matplotlib version: 3.2.0
• Matplotlib backend: Qt5Agg
• Python version: 3.7 and 3.8

I tested this with Matplotlib v3.2.0 installed by pip and conda.