The user can't revert to linear scale after setting the scale to "Log" or "Square root" on an unsupported dataset

[ckuenzle]

Describe the bug

After switching to "Log" or "Square root" scale on an unsupported dataset, the scale selector disappears, making it impossible to reset the scale.

To Reproduce

1. Create a dataset that only contains negative values (e. g. a 2D matrix).
   Example Python code:

import h5py

with h5py.File("test.h5", "w") as hdf5_file:
    hdf5_file.create_dataset("positive_matrix", data=[[1, 2], [3, 4]])
    hdf5_file.create_dataset("mixed_matrix", data=[[-1, 2], [3, -4]])
    hdf5_file.create_dataset("negative_matrix", data=[[-1, -2], [-3, -4]])

2. Open that dataset as a line plot or heatmap.

3. Change the scale to "Log" or "Square root"

4. Now we get an error indicating our dataset cannot be displayed: "Expected domain compatible with log scale". This is expected behaviour, as the logarithm and square root of a negative number are undefined (at least in the real number space). However, the dropdown selector that is needed to revert to the linear scale is gone. The only way I found to get the linear scale back is to open another dataset that contains positive values.

Expected behaviour

Keep displaying the scale selector even if a dataset contains unsupported data.

Context

• OS: Windows 11 24H2
• H5Web context: H5Web VS Code extension v0.1.7
• I also tried H5Wasm, so it's not a VS Code related issue

[loichuder]

👍

I actually encountered this myself a few times but forgot to write it down...