dualx not correctly shown

[miniufo]

There is a dualx funcitonality that could be very useful in labelling the wavelength and wavenumber simultaneously. Here is an example from a paper:

The x-axis at the bottom shows the wavenumber (cycle per km) while the one at the top shows the wavelength (km), as wavenumber = 1 / wavelength.

I tried this with the following simple codes:

import ultraplot as uplt
import numpy as np
import xarray as xr

data = xr.DataArray(np.random.randn(100)+2, dims=['wn'], coords={'wn':np.logspace(-2, 2, 100)})

fig, axes = uplt.subplots(figsize=(5,4))

ax = axes[0]
ax.plot(data)
ax.set_xscale('log')
ax.set_yscale('log')
ax.format(xformatter='log')
ax.set_xlabel('wavenumber (cycle per km)')
ax.dualx(
    lambda x: 1 / x,
    label="length scale (km)",
    formatter="log",
    grid=True,
    color='k',
    gridcolor='k',
)

which shows

It is a bit strange to me that the two x-axes do not align exactly. Just want to know how to do this correctly. Thanks very much.

[cvanelteren]

Strange that should not happen.

[cvanelteren]

Found the issue. The issue is the order to which the transform is applied.

The issue was that dualx applied your conversion after the parent log transform, so the secondary axis was effectively using 1/log10(x) instead of 1/x, which broke the limits and hid the axis. I switched the transform order so the conversion runs in data space first and then the parent scale (log) is applied; that keeps the limits consistent and makes the secondary ticks line up correctly on log axes.

[miniufo]

That's quick! I see the reason, and it is now fixed by changing the order of the transforms. Thank you very much.

[cvanelteren]

I am changing it in a PR now so it will be fixed on next release @miniufo