Lifecycle and I/O#
A MemoryMap owns a native mapping until it is closed. Context-manager
use is preferred because cleanup remains deterministic when an exception is
raised.
read(size, offset) and write(data, offset) operate on byte ranges.
Validate bounds before treating a mapping as a structured record store; this
module does not know your file format, element type, endianness, or record
layout.
File-backed mappings#
A file-backed mapping exposes bytes through the operating-system page cache. Changes and visibility depend on mapping flags and synchronization choices. Do not assume that Python object lifetime alone publishes data durably to disk.
Anonymous mappings#
Anonymous mappings are not associated with a file. They are useful for native scratch space and shared-memory-like patterns where the process controls both the layout and lifetime.