Four Bug Reports for One Short Function

A museum's visitor report is produced by this function, which receives the lines of an exported file (often straight from the file, so possibly a one-pass iterator):

def busiest_days(lines, k=3):
    rows = (line.strip().split(",") for line in lines if line.strip())
    total = sum(int(v) for _, v in rows)
    ranked = sorted(rows, key=lambda r: r[1], reverse=True)
    return total, [date for date, _ in ranked[:k]]

The intended behaviour: the first non-empty line is a header (date,visitors) and is skipped; every other non-empty line is YYYY-MM-DD,count. Return (total visitors, the dates of the k busiest days), busiest first, and among days with the same count the earlier date first. Blank lines (possibly with spaces) are ignored, and a file with only a header, or nothing at all, gives (0, []). Four reports came in:

1. "With the real export it crashes: ValueError: invalid literal for int() with base 10: 'visitors'."
2. "After working around that, the list of days is always empty, though the total is right."
3. "A day with 900 visitors is ranked above a day with 1200."
4. "Two days with the same count come out in file order; the earlier date must come first."

Write the corrected busiest_days(lines, k=3). The setup helper export(n, seed) generates a file's lines one at a time (a generator, header first), as the real file reader does.

Examples

Input:  busiest_days(iter(["date,visitors", "2024-05-02,900", "", "2024-05-01,1200", "2024-05-03,900"]), k=2)
Output: (3000, ["2024-05-01", "2024-05-02"])

Input:  busiest_days(["date,visitors", "  "])
Output: (0, [])

Constraints

• Up to 10**5 lines; counts are whole numbers from 0 to 10**6; each date appears at most once.
• 0 <= k; if there are fewer than k days, all are returned.

Goals

• Reproduce each bug report with a small input before changing anything
• Recognise an iterator that is used up by a first pass
• Fix sort keys that compare text instead of numbers, and tie-breaks that follow the wrong order