Repository navigation
post-1.0: design and implement database branching/forking #201
Description
Roadmap source
ROADMAP.md: Phase 9.5 Database Branching / Forking.
Work item
Design and implement database branching/forking.
Concept
db.branch() produces a new independent .graph file pre-populated with facts from the parent. db.branch_as_of(path, tx_count) forks from a specific transaction count.
Acceptance criteria
- Produce an independent single-file database.
- Ensure WAL is fully flushed/checkpointed before fork.
- Support current-state fork and explicit transaction-count fork if design confirms API.
- Add tests proving child writes do not affect the parent.
- Document semantics and limitations.
Activity
adityamukho commentedon May 15, 2026
Prerequisites
Database branching/forking interacts deeply with WAL state and checkpoint boundaries — a fork of an uncheckpointed database must flush/checkpoint first, and the semantics of a fork mid-WAL-replay need to be well-defined. The WAL reliability work below should land first so that branching is built on a well-tested foundation and doesn't duplicate fault-case reasoning.
adityamukho commentedon Oct 27, 2026
Non-trivial data model to be designed first. No known use case at present. Deferred indefinitely.