NumPy Memory Mapping

How it works

This local-only example uses an owned temporary file and Python mmap as the explicit resource owner. NumPy reads its buffer while the mapping is open; all array views must be released before the mapping closes. Operations that allocate a result still need memory, and real I/O behaviour depends on the operating system and workload. The browser playground is not offered as a demonstration of local disk mapping.

Run the code

This memory-mapping example is local-only. It uses an owned temporary file and an explicit mapping lifetime; the browser editor is not a test of local disk mapping.

For local work, use a separate virtual environment and run python memory-maps.py. Download the complete source. The reviewed local environment uses Python 3.12.14 and NumPy 2.5.4; the browser uses Pyodide 0.27.7 with NumPy 2.0.2. Compatible examples are checked in both. See the environment setup. No Apache or database server is required.

Example

import numpy as np

import mmap
from pathlib import Path
from tempfile import TemporaryDirectory
with TemporaryDirectory(prefix="numpy-map-") as folder:
    file = Path(folder) / "readings.bin"
    file.write_bytes(np.array([12, 18, 24], dtype="<i8").tobytes())
    with file.open("r+b") as stream:
        with mmap.mmap(stream.fileno(), 0) as mapping:
            values = np.frombuffer(mapping, dtype="<i8")
            print(int(values.sum()))
            del values

Expected result

54

The explicit little-endian dtype describes how bytes become integers. Releasing the array before leaving the mapping context avoids retaining a view into a closed resource. The private file disappears with its temporary directory.