Persistent objects (base behavior)
PyCauset objects are typically backed by either RAM or a memory-mapped file, and they can be persisted as .pycauset snapshots.
There is no stable public Python class named pycauset.PersistentObject in Release 1.
Instead, these behaviors surface through:
Storage concepts
- Backed by disk or RAM: the API is the same either way.
- Snapshots are immutable by default:
load()does not implicitly overwrite the file you loaded. - Metadata-first: shape, dtype, view-state, and semantic properties are stored/propagated without scanning payload.
The canonical Release 1 persistence semantics and container format are documented in:
Semantic properties
Matrices and vectors expose obj.properties, a typed mapping used for:
- gospel semantic assertions (e.g.
is_upper_triangular=True), and - cached-derived values (e.g.
trace,determinant,norm) with strict validity.
See R1 Properties for the user-facing contract.