x axis labeling disappears when using twinx()

[bradyrx]

Code sample, a copy-pastable example if possible

A "Minimal, Complete and Verifiable Example" will make it much easier for maintainers to help you.

import proplot as plot
import numpy as np
import xarray as xr

times = np.arange(1990, 2000)
data = xr.DataArray(np.random.rand(len(times)), dims=['time'], coords=[times])

f, ax = plot.subplots(aspect=2)
ax.plot(data.time, data)

twinax = ax.twinx()

# twinax.format(xlocator=1)
ax.format(xticks=np.arange(1990, 2000, 2))

Actual result vs. expected result

Using twinx(), the tick labeling disappears on the twin axis. This happens for datetime and normal numbers. Using xticks/xlocator on the main axis or twinax instance changes the ticks but does not produce labels.

[bradyrx]

You can actually see this in the RC settings demo on the docs on the right subplot.

https://proplot.readthedocs.io/en/latest/basics.html#Changing-rc-settings

[lukelbd]

Thanks for the report -- which version are you using? Is this the "dev" release?

[bradyrx]

I'm using the 0.0.5 that's on pip now.

[lukelbd]

Actually this is intended behavior. My guess is you have confused twinx with twiny? Counterintuitively, the matplotlib twinx function generates two y axes with identical "twin" x axes, and ProPlot ensures the shared axes ticks, tick labels, and axis labels on the secondary x axis are hidden so you don't accidentally draw duplicate labels from both "twins" on top of each other. If you want to set the x axis properties, you should do so on the original axes.

Since this naming is really confusing for users, proplot introduces the more intutively-named altx function for making an "alternate x axes" and alty for making an "alternate y axes". I recommend using these instead.

Anyway maybe you already knew all this; since the whole purpose of "twin" axes is that their properties should be identical, I thought it made sense to forbid specially modifying the properties of the "twin". I'm open to arguments against this though.

[zxdawn]

@lukelbd I met the same problem today.

If I use the same example as @bradyrx showed above without twin, the xaxis tick labels are correct:

But, if I add the the twinx(), the tick labels disappeared:

This is the actual case I met: I want to plot two kinds of profiles with pressure, but I couldn't let the pressure ticks labels show up ...

[bradyrx]

Counterintuitively, the matplotlib twinx function generates two y axes with identical "twin" x axes, and ProPlot ensures the shared axes ticks, tick labels, and axis labels on the secondary x axis are hidden so you don't accidentally draw duplicate labels from both "twins" on top of each other. If you want to set the x axis properties, you should do so on the original axes.

Yes I understand this and agree with you that the naming is bad. I also agree that you shouldn't be able to modify the dual axis so you don't get out-of-phase ticks, clutter, etc.

The issue, as @zxdawn also mentioned, is that the original axis completely disappears. And in my example, you see that modifying the original axis does nothing to fix this (below with alty):

import proplot as plot
import numpy as np
import xarray as xr

times = np.arange(1990, 2000)
data = xr.DataArray(np.random.rand(len(times)), dims=['time'], coords=[times])

f, ax = plot.subplots(aspect=2)
ax.plot(data.time, data)

twinax = ax.alty()

# twinax.format(xlocator=1)
ax.format(xticks=np.arange(1990, 2000, 2))

I agree you shouldn't be able to modify the twin axis, but it shouldn't render the original axis unreadable. It's still important to see the ticks, time series labeling, etc.

[lukelbd]

My bad! I see now, this definitely is a major problem. I will address this before the end of the weekend. Thanks for the additional examples.

[lukelbd]

Sorry for the wait guys; I had to step back from proplot for a while and knew this one would be tricky to fix. But it's finally done.

[zxdawn]

@lukelbd Thanks for the fix! Will test it for the O3 and Qv profile tomorrow.

[bradyrx]

This is great, thanks Luke!

[zxdawn]

Works well. Thank Luke!