make_chunks(file_name: str)

What it does

1. Memory-maps the entire file.
   • Calls mmap.mmap(...) to create a view of the file as a bytes object, living in virtual memory.
   • This means we can read slices like mm[start:end] without loading more data explicitly.
2. Divides the file into N chunks (where N = CPU cores).
   • Uses the file size and divides by cpu_count to get roughly equal byte sizes.
3. Adjusts each end to align with the next \\n (newline).
   • Ensures lines don’t get cut in half, so each chunk is safely processable on its own.

Why

• This allows parallel processing without overlap or data corruption.
• Each process can safely work on its own chunk.

Important note

• Sets a global mm object so that all forked child processes can read from it without copying memory.

agg_initializer()

What it does

• Returns the starting aggregation list: [inf, -inf, 0.0, 0]

  which means [min, max, sum, count].

Why

• defaultdict needs a way to automatically create a fresh aggregation state for each new city.

process_chunk(start, end, queue)

What it does

1. Takes a slice mm[start:end] of the file.