Exercise 2 solution

    import os, struct, tempfile

    rec = struct.Struct("<HhI")
    records = [(1, -5, 1000), (2, 300, 70000), (3, -32768, 4294967295), (4, 0, 0)]
    path = os.path.join(tempfile.gettempdir(), "records.bin")

    with open(path, "wb") as f:
        for r in records:
            f.write(rec.pack(*r))

    with open(path, "rb") as f:
        raw = f.read()
    print(len(raw), rec.size * len(records))             # 32 32
    print(list(rec.iter_unpack(raw)) == records)         # True
    print(struct.unpack_from("<HhI", raw, rec.size * 2)) # (3, -32768, 4294967295)
    os.remove(path)

(Run as written; the output matched the comments.)

Explanation: "<HhI" is 2 + 2 + 4 = 8 bytes with no padding, so four records
are 32 bytes. iter_unpack walks the buffer in 8-byte steps and yields one tuple
per record; it raises struct.error if the buffer length is not a multiple of
the record size. Because every record has the same size, record number n starts
at offset n * 8, which is how unpack_from reaches the third record (index 2, at
offset 16) without reading the earlier ones. Note that h is signed and H, I are
unsigned, which is why -32768 and 4294967295 both round-trip.