Jupyterlab shows tables in black on black

[horsto]

In Jupyterlab table previews are suddenly unreadable in dark and dark-high-contrast themes. I am not sure what happened here, but I seem not to be able to revert it to showing the font in white. This might be a jupyter lab issue and not a datajoint issue.

OS: Mac OS
Jupyter lab: Version 4.2.4
Datajoint python: Version 0.14.1

[CBroz1]

The html table style is defined here, hardcoded for light themes. Not ideal, but running an editable install would allow you to adjust these values

[dimitri-yatsenko]

@CBroz1 what would you suggest for a better solution?

[CBroz1]

I can see a few possible paths...

1. Determine theme and adjust html accordingly. Discussion

• Pros: maintains status quo for light-mode users
• Cons: maintenance burden of front-end variance

2. Defer rendering to pandas html, which already adjusts for theme.

• Pros: offloads maintenance to another package, no additional dependencies
• Cons:
  • mismatch across versions
  • may require deprecation timeline for html in case someone uses this for other tools
  • minor edits to show primary keys and table header

3. Add a config option that determines the html displayed

• Pros: maintains status quo, no interface changes, minimal work
• Cons: Additional config line item

Example Pandas adjustment for primary keys

import pandas as pd
from IPython.display import display

data = {
    'id': [1, 2, 3],
    'name': ['Alice', 'Bob', 'Charlie'],
    'age': [24, 27, 22],
    'email': ['[email protected]', '[email protected]', '[email protected]']
}
df = pd.DataFrame(data)
primary_key_cols = ['id', 'name']

def style_primary_keys(df, primary_key_cols):
    styled_cols = {col: f"*{col}" if col in primary_key_cols else col for col in df.columns}
    df = df.rename(columns=styled_cols)
    return df

styled_df = style_primary_keys(df, primary_key_cols)
display(styled_df.style)

I don't think 1 is worth the additional maintenance burden. 2 may be the most long-term solution, but 3 is an easier short-term solution. I lean toward 2.