Pass `locator` argument to matplotlib when calling `plot.contourf`

[iuryt]

Is your feature request related to a problem? Please describe.
Everytime I have to do a contourf I need to call it from matplotlib directly because locator argument is not passed from xarray.plot.contourf to matplotlib.

Describe the solution you'd like
Being ds a xarray.DataArray, I want to have the behaviour described here when passing locator = ticker.LogLocator() to ds.plot.contourf.

Describe alternatives you've considered
I usually have to do plt.contourf(ds.dim_0,ds.dim_1,ds.values,locator = ticker.LogLocator()).

[dcherian]

ah this is to decide what the levels are

We'll have to pass it down to here and make appropriate modifications:

xarray/xarray/plot/utils.py

Lines 272 to 281 in 8ff8113

	if is_scalar(levels):
	if user_minmax:
	levels = np.linspace(vmin, vmax, levels)
	elif levels == 1:
	levels = np.asarray([(vmin + vmax) / 2])
	else:
	# N in MaxNLocator refers to bins, not ticks
	ticker = mpl.ticker.MaxNLocator(levels - 1)
	levels = ticker.tick_values(vmin, vmax)
	vmin, vmax = levels[0], levels[-1]

PR welcome!

[iuryt]

I can tackle this one, if no one does that before, I just have to separate a time in my week schedule for open-source projects.

[dcherian]

@iuryt you could also create levels manually using the locator you like and pass that. That seems better than adding yet another kwarg to our plotting functions.