Changelog - 2026-09-29

inSql and ninSql

Feature

In one line. A where can now say "this column is among the rows a subquery returns" - { id: { inSql: sql`...` } } - so filtering by another table no longer needs a first query that collects ids.

What is new

import { sql } from 'drizzle-orm';

const taggedRed = sql`SELECT ${tagTable.orderId} FROM ${tagTable} WHERE ${tagTable.label} = ${label}`;

await orderRepository.find({ filter: { where: { id: { inSql: taggedRed } } } });
await orderRepository.count({ where: { id: { ninSql: taggedRed } } });

• inSql emits "col" IN (<subquery>), ninSql emits "col" NOT IN (<subquery>).
• Works in find, findOne, count, update and delete wheres, nested and/or, and together with the default filter and row scope.
• TWhereOperators gains inSql?: TSqlFragment and ninSql?: TSqlFragment. TSqlFragment is { getSQL(): unknown } - a Drizzle SQL or SQLWrapper - named structurally so @venizia/ignis-filter stays free of Drizzle.

Safety

The operand must be a SQL object built in server code. A string is refused with a 400, and a filter that arrives as JSON over HTTP cannot carry a SQL object, so the operators add no injection path. Values go into the subquery through Drizzle's ${} bindings.

Good to know

• An empty subquery matches nothing under inSql and everything under ninSql.
• If the subquery returns a NULL, NOT IN matches no row, as in SQL. Filter NULLs out inside the subquery when the column can hold them.
• Postgres and SQLite support both operators. Typesense and Meilisearch refuse them with a 400.

Who is affected

Nobody has to change anything. Code that collects ids in a first query and then filters with inq can move to inSql and drop the round trip and any cap on the id list.

Files: packages/connectors/src/relational/core/repositories/dialect/internal/subquery.ts