1D line plot with data on the x axis

[rabernat]

Consider the following Dataset, representing a function f = cos(z)

z = np.arange(10)
ds = xray.Dataset( {'f': ('z', np.cos(z))}, coords={'z': z})

If I call

ds.f.plot()

xray naturally puts "z" on the x-axis.

However, since z represents the vertical dimension, it would be more natural do put it on the y-axis, i.e.

plt.plot(ds.f, ds.z)

This is conventional in atmospheric science and oceanography for buoy data or balloon data.

Is there an easy way to do this with xray's plotting functions? I scanned the code and didn't see an obvious solution, but maybe I missed it.

[shoyer]

I don't think we have an easy way to do this now.

Seaborn uses the keyword argument orient='h' or orient='v' to control orientation for categorical plots like violinplot. That might be a reasonable API to emulate here.

[clarkfitzg]

Not possible now. Wouldn't be much trouble to add it though. @drewokane has been working on generalizing the 1d plotting in #566. This might be a nice thing to include.

[drewokane]

I agree orient would be a natural keyword to use for this situation.

[rabernat]

Yes, 👍 orient.

[jhamman]

I'm fine with orient={h, v} but I will offer a few other alternatives.

• Explicit is better than implicit. I know what h and v mean but it is more explicit to say horizontal or vertical. These keywords also require less in terms of documentation.
• Can we only imagine two options for the orientation? If so, maybe a True/False keyword is a better way to go (orient_horizontal=True)? Maybe not?

I'm not married to either of the opinions I just mentioned, just a few thoughts.

[kuchaale]

Would it make sense to have such minimalistic line function?

[shoyer]

I wonder if it makes more sense to make .plot() a Dataset method as well, so we could write things like ds.plot.line(x='f', y='z') and ds.plot.scatter(x='f', y='z'). Writing ds.f.plot(orient='h') is much less self-explanatory.

[fmaussion]

In a first step at least, we could add the x and y arguments to the line plot method to make the API consistent with the 2d plot methods (http://xarray.pydata.org/en/latest/api.html#plotting). But now that I think of it, they should be kwargs, not args

I wonder if it makes more sense to make .plot() a Dataset method as well

Yes. This might be a bit of an undertaking as there will be several possible cases to handle but this is a good idea!

[shoyer]

In a first step at least, we could add the x and y arguments to the line plot method to make the API consistent with the 2d plot methods

If an argument is a 1D DataArray, how do we indicate this array's data? The best I can think of is to use some sort of sentinel value, maybe None: df.f.plot(x=None, y='z')

[kuchaale]

I think plotting two 1D variables from one Dataset against each other may be useful. However, I was not able to figure out yet how to proceed with it (see my amateur PR).

[dcherian]

@kuchaale I see you have an open PR. Are you planning to revisit this soon? If not, I'd like to build off what you started and tackle this bit (not the Dataset method)

In a first step at least, we could add the x and y arguments to the line plot method to make the API consistent with the 2d plot methods (http://xarray.pydata.org/en/latest/api.html#plotting).

[kuchaale]

@dcherian go ahead. I planned to revisit it, but it would not be definitely soon.

[dcherian]

@kuchaale Thanks!