One markdown link of a record that resolves to nothing — the field the link was found in
against the virtual path it aimed at, kept apart rather than formatted into a sentence, so a
caller can group by field or retry the target without parsing text back out.
How often a handle catches up with foreign writes on its own, before serving a read — the port
of the TS FollowPolicy (packages/hayson/src/table/table.ts).
A validated, _seq-assigned, serialized log line produced by Table::prepare_txn_ops —
nothing on disk yet, but everything needed to journal it and, later, to append it.
The result of Table::put_with_warnings/Table::patch_with_warnings — the written record
plus diagnostics that do not block the write (currently only broken markdown links found while
Table::link_parsing is Eager; empty for a Lazy table, which surfaces them only on
request via Table::broken_links_of). Table::put/Table::patch stay untouched and
keep returning a bare HaysonRecord on a separate method, rather than changing their own
signature, so no existing caller has to start handling warnings it never asked for.
What one run of Table::rotate_cypher_keys did. Counts records, not fields: a record with
two cypher fields on a retired key is rotated once and counted once.
Ties TableRegistry (indexes), WriteQueue (validation, uniqueness, staging) and LogFile
(persistence) into one table API. Deliberately thin: put/patch/delete all delegate their
validation and unique-checking to WriteQueue::stage_put, they do not repeat it here.
How many parsed documents a table keeps in memory in front of its log — the read cache
crate::table::Table::get answers a repeated read from instead of re-reading and re-parsing
the line the registry points at.
What Table::refresh found on disk since this handle last looked. Modelled on
RestoreOutcome, which reports the same kind of thing about an open: not whether it worked,
but which of its two ways it had to work.