Sharing axes labels goes wrong for twinned axes.

[cvanelteren]

First addresses in #232.

For non-twinned axes the labeling is shared; meaning for a grid we get an expected label on the left for sharing labels above or equal to one. For shared labels this does not occur.

For sharing level "all", the last plot get the ylabel but none are spawned.
For sharing level "labels" the label is copied from the axis it is shared from.

Expectation

The labels should behave similarly to how the non-twinned subplots behave, with the axes being shared. According to the docs, the properties are copied from the sharing axis. I can confirm that this is correctly done. Not sure how trivial this would be to implement since the sharing will make the determination more complex. We could use the _get_border_axes function to determine where to put the labels.

snippet

import ultraplot as uplt, numpy as np, pandas as pd

df = pd.DataFrame(np.random.rand(10, 2), columns=["a", "b"])
layout = [[1, 2], [3, 4]]
fig, axs = uplt.subplots(layout, share="all")
for axi in axs:
    alt = axi.alty()
    axi.plot(x="a", y="b", data=df)
    alt.plot(x="a", y="b", data=df)
uplt.show(block=1)

Problem plots

share labels

share all

[cvanelteren]

Need to confirm:

• How are y labels shared across subplots
• How are twinned axes spawned.
  They are held within _twinned_axes and the current yaxis is being shared.

[cvanelteren]

Small update. I think the labels on the twinned are not wrong, but rather the labels that are shared on y. That is, the label that should occur on the subplots should be on the left "shared" rather it now shows "twice" (once on the left and once on the right. Furthermore, the share="all" turns off the ticklabels on the left -- which is not intended.

[cvanelteren]

For example, this:

Looks great.

snippet

import ultraplot as uplt, numpy as np, pandas as pd

df = pd.DataFrame(np.random.rand(10, 2), columns=["a", "b"])
layout = [[1, 2], [3, 4]]
fig, ax = uplt.subplots(layout, share=1)

for axi in ax:
    alt = axi.alty()
    axi.plot(x="a", y="b", data=df)
    alt.plot(x="a", y="b", data=df)

ax.format(xlabel="Main x", ylabel="Main y")

fig.save("test.png")
uplt.show(block=1)

In this example on the labels are shared and I think this behaves accordingly as it is clear that the alty is depicting a different plot and as such can have a different value.

UltraPlot behaves strangely when the sharing level exceeds 2, e.g.:

Here the snippet will show the Main y in between the columns which is not as intended. The expectation for this plot would be that the ylabel for alty should be off and the labels are on the left. The docs state:

True or 3 (default): Same as 2, but axis tick labels are also shared. Tick labels will appear on the leftmost and bottommost subplots.

Which is clearly not happening. I have the suspicion that this is caused by alty truely generating a different CartesianAxes and the logic has some internal reference that looks for the outer axes that is not correct. Will have a further look.

[cvanelteren]

The issue is caused by the mapping between the sharer and sharee being inverted somehow for alternate axes. I am not sure yet when or where this happens. The function _add_alt seems to be free of this issue and it happens somewhere else in the chain.

[cvanelteren]

Found the core issue. See #244 for fix.

[lwq-star]

Found the core issue. See #244 for fix.找到核心问题。请参阅 #244 进行修复。

Can you release a new version so I can update the ultraplot package?

[cvanelteren]

Not yet, I have identified the issue but not yet a proper fix. Inverting the sharing will break the label sharing for top and right label.

[cvanelteren]

Partly fixed on main an on latest release (v.1.55). Keeping this open as a reminder for #246

[lwq-star]

How to merge shared axis labels?

import ultraplot as uplt, numpy as np, pandas as pd

df = pd.DataFrame(np.random.rand(10, 2), columns=["a", "b"])
layout = [[1, 2], [3, 4]]
fig, axs = uplt.subplots(layout, share="all")
for axi in axs:
    alt = axi.alty()
    axi.plot(x="a", y="b", data=df)
    alt.plot(x="a", y="b", data=df)
uplt.show(block=1)

[cvanelteren]

There is no functionality for this at this stage. When alternate y are placed it is not trivial to know which limits they are for all plots. Manual intervention is easier to do her than merging them.

[cvanelteren]

An alternative would be to show the ticks on both. And plot them in the same range. Both are valid but depend on your intention.

[lwq-star]

That's already good enough.

import ultraplot as uplt, numpy as np, pandas as pd

df = pd.DataFrame(np.random.rand(10, 2), columns=["a", "b"])
layout = [[1, 2], [3, 4]]
fig, axs = uplt.subplots(layout, span=False, share="all")
for i, axi in enumerate(axs):
    alt = axi.alty()
    axi.plot(x="a", y="b", data=df)
    alt.plot(x="a", y="b", data=df)
    
    if i % 2 == 1:
        alt.format(ylabel="b")
    else:
        alt.format(ylabel="")
    
uplt.show(block=1)