JSON path queries

Read into nested JSON columns directly from where. Same scalar comparison vocabulary as regular fields — eq / ne / gt / gte / lt / lte / contains / in / has.

const Doc = model('docs', {
  id: f.id(),
  meta: f.json(),
});

// Dotted-path navigation.
await db.doc.findMany({
  where: { meta: { path: 'profile.age', gte: 18 } },
});

// Array indexing with [N] syntax.
await db.doc.findMany({
  where: { meta: { path: 'addresses[0].city', eq: 'Lagos' } },
});

// Substring search on the extracted value.
await db.doc.findMany({
  where: { meta: { path: 'bio', contains: 'engineer' } },
});

// Explicit array form (skips the dotted-path parser).
await db.doc.findMany({
  where: { meta: { path: ['tags', '0'], eq: 'urgent' } },
});

// Works on embedded objects too.
await db.user.findMany({
  where: { address: { path: 'city', eq: 'SF' } },
});

Works on f.json() / f.embed() / f.embedMany() / f.stringArray() / f.intArray() fields. Non-JSON fields raise a clear error.

Per dialect:

Dialect	Compiles to
Postgres	(col->'a'->>'b')::numeric (cast by operand type — string / numeric / boolean)
MySQL	JSON_UNQUOTE(JSON_EXTRACT(col, '$.a.b'))
SQLite	json_extract(col, '$.a.b') (JSON1 — built-in)
DuckDB	json_extract(col, '$.a.b')
MSSQL	JSON_VALUE(col, '$.a.b')
Mongo	dotted key: { 'meta.a.b': … }

See also: docs/JSON-PATH.md — per-dialect SQL emit, GIN/multi-valued/expression indexes, column ↔ JSON migration, full operator-by-dialect matrix, null-marker semantics, audit-log/webhook patterns, common bugs.

---