fix(python): Apply thousands_separator to count/null_count in describe() for non-numeric columns

[veeceey]

Summary

• Fixes describe doesn't abide thousands separator for string columns #25946
• When using pl.Config(thousands_separator=...), the describe() method now correctly applies the thousands separator to count and null_count statistics for non-numeric columns (string, categorical, temporal, etc.)
• Previously, these values were converted to plain strings via str(v) without formatting, while numeric columns received proper thousands formatting through their float representation

Changes

• In LazyFrame.describe(), when casting non-numeric column statistics to strings, count and null_count values are now formatted with the configured thousands separator using f"{int(v):,}".replace(",", thousands_sep)
• The separator is retrieved via a local import of get_thousands_separator() from polars._plr, avoiding any module-level import changes
• When no thousands separator is configured (empty string), behavior is unchanged

Test plan

• Added test_df_describe_thousands_separator_string_columns parametrized for both DataFrame and LazyFrame
• All 13 existing describe() tests pass
• Manually verified with the exact reproducer from the issue using both _ and ç separators

Before:

│ count      ┆ 2ç000.0  ┆ 2000 │

After:

│ count      ┆ 2ç000.0  ┆ 2ç000 │

[veeceey]

Manual Testing Results

Verified the fix for issue #25946 where describe() was not applying the configured thousands separator to count/null_count for non-numeric columns.

Test 1: Original Issue Scenario

import polars as pl

pl.Config.set_thousands_separator(",")
df = pl.DataFrame({"a": ["x"] * 2000, "b": [1.0, 2.0] * 1000})
print(df.describe())

Before fix:

• String column 'a' count: 2000 (no separator)
• Numeric column 'b' count: 2,000.0 (with separator)

After fix:

• String column 'a' count: 2,000 (with separator)
• Numeric column 'b' count: 2,000.0 (with separator)

Both columns now consistently use the configured separator.

Test 2: Different Separators

Tested with space (" "), underscore ("_"), and unicode ("ç") separators - all work correctly for both string and numeric columns.

Test 3: Null Count Formatting

pl.Config.set_thousands_separator(",")
df = pl.DataFrame({
    "with_nulls": [None] * 1500 + ["value"] * 500,
    "numeric": [1.0] * 1000 + [None] * 1000
})
print(df.describe())

Both count and null_count rows now correctly show separators for string columns (e.g., "1,500").

Test 4: LazyFrame

Verified that LazyFrame.describe() has the same behavior as DataFrame.describe() - thousands separators are applied consistently to string column statistics.

Test 5: Edge Cases

• Empty separator (""): Works correctly, no formatting applied (plain integers)
• Large numbers (>10,000): Separators applied at all appropriate positions

All tests pass. The fix successfully applies the configured thousands separator to count and null_count statistics for non-numeric columns, providing a consistent user experience across all column types.

[veeceey]

Friendly ping - any chance someone could take a look at this when they get a chance? Happy to make any changes if needed.