fix(python): When setting write_excel column totals, don't forget to include any row-total cols

[alexander-beedie]

When setting column_totals=True we should account for any columns created via row_totals (we used to¹, and it seems this went missing somewhere in a refactor).

Also:

• The row_totals column(s) dtype (int/float) should be determined by the constant columns; if all are int, the total should also be int, not float.
• Default numeric value formats now better-match explicit table style colours.

Example

import polars as pl

df = pl.DataFrame({
    "id": ["aaa", "bbb", "ccc", "ddd", "eee"],
    "q1": [100.0, 55.5, -20.0, 0.5, 35.0],
    "q2": [30.5, -10.25, 15.0, 60.0, 20.5],
    "q3": [-50, 0, 40, 80, 80],
    "q4": [75, 55, 25, -10, -55],
})

df.write_excel(
    "testing.xlsx",
    table_style = "Table Style Light 2",
    sparklines = {"trend": ["q1", "q2", "q3", "q4"]},
    row_totals = {
        # add semiannual row total columns
        "h1": ("q1","q2"),
        "h2": ("q3","q4"),
    },
    column_totals = True,
    autofit = True,
)

Output:

• Column totals now exist for "h1" and "h2".
• The "h1" totals are float, "h2" totals are integer.
• Numeric values match the table style colour (and string value colour).

Footnotes

1. https://github.com/pola-rs/polars/pull/7751 ↩

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 80.39%. Comparing base (ca6d46c) to head (a9e4cec).

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #18042      +/-   ##
==========================================
- Coverage   80.39%   80.39%   -0.01%     
==========================================
  Files        1496     1496              
  Lines      197480   197489       +9     
  Branches     2820     2821       +1     
==========================================
- Hits       158764   158763       -1     
- Misses      38194    38205      +11     
+ Partials      522      521       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.