[Rajpal14]

Problem

Fixes #494.

When tabular data contains rows of varying lengths (ragged rows) and the first row has fewer columns than subsequent rows, _normalize_tabular_data computed headers_pad using only len(rows[0]):

headers_pad = max(0, len(rows[0]) - len(headers))

If headers was shorter than the maximum column count across all data rows (e.g. tabulate([["a"], ["b", "KEEP"]], headers=["tail"])), headers remained shorter than the total number of columns. Later, zip(cols, aligns_copy, minwidths) truncated the columns to the length of minwidths (derived from headers), silently dropping data columns like "KEEP".

Solution

• Updated _normalize_tabular_data to compute headers_pad using the maximum column count across all non-separating data rows (max(len(r) for r in rows if not _is_separating_line(r))).
• Added a regression test test_ragged_rows_with_short_headers in test/test_regression.py to verify that columns beyond the first row's width are preserved when short headers are supplied.

Verification

• Tested with the minimal reproducible example from Short headers silently drop columns from ragged tables #494.
• Ran full test suite (pytest test/): 307 passed, 60 skipped, 0 failed.

[Rajpal14]

Hi maintainers, when you have time, could you please review this fix for #494? It includes a regression test for ragged rows with shorter headers, and the full test suite passes locally. Thank you!