jac-sv-persistence

Query persisted nodes with list-comprehension traversals like [root -->][?:Post], adding bracket predicates such as [?published] to filter. Aggregate with len() since there is no count() form, and writes persist automatically once a node is reachable from root.

Find a node by id using either a loop over [root -->] comparing jid(n) to the target id, or jobj(id) with an isinstance check for O(1) lookup. Never use Python id(), which changes every restart and silently returns empty results.

Yes, Jac persists to SQLite in .jac/data/ by default, and setting MONGODB_URI switches to MongoDB with a Redis L2 cache. The same data model and schema migration rules run identically on both backends.

A renamed field without a declared alias looks like a remove-plus-add: old values move to the recoverable attic and the new field reads its default. Declare schema_alias("new", stored="old") inside __jac_schema__ so values flow into the renamed field.

Invalid anchor id 500 errors come from stale anchors persisted by a previous run under a different schema. In development, stop the server, delete .jac/data/, and restart; in production, use the alias and quarantine machinery via jac db commands instead.