eon.process_id

Content-addressed process ids (xxHash).

Process table rows and procdata files are keyed by integer ids. A counter based on len(procs) freezes when duplicate ids collapse the dict; a counter based on max(id)+1 still couples identity to registration order and external table rewrites.

Instead, seed the id from a stable xxh64 of the process payload (saddle geometry bytes, barrier tag, direction). Geometric uniqueness remains find_repeat; the hash only assigns free keys that do not depend on table length. Rare collisions (or an already-occupied pure-content hash) are resolved by salting until free.

Module Contents

Functions

process_id_from_parts	xxh64 over parts (null-separated), optionally salted.
allocate_unique_process_id	Return a content-addressed id not already in occupied.

Data

BytesLike

API

eon.process_id.BytesLike

None

eon.process_id.process_id_from_parts(*parts: eon.process_id.BytesLike, salt: int = 0) → int

xxh64 over parts (null-separated), optionally salted.

Returns a non-negative 63-bit int so it is safe as a dict key and as the %d field in processtable / procdata/*_%d.* paths.

eon.process_id.allocate_unique_process_id(occupied: dict | set | None, *parts: eon.process_id.BytesLike, max_salts: int = 1024) → int

Return a content-addressed id not already in occupied.

Salt 0 is the pure content hash. If that id is already present (true re-registration of the same payload, or an xxh64 collision), salt until free. Raises if max_salts is exhausted.