rls-policy-optimizer

Optimize RLS policies by replacing direct auth.uid() calls with the (select auth.uid()) pattern in USING clauses. This subquery approach evaluates auth.uid() once per policy instead of per row, significantly improving query performance across tables like leads, cotizaciones, and items_cotizacion.

The (select auth.uid()) pattern forces the database to execute auth.uid() once and cache the result, rather than re-evaluating it for every row during policy filtering. This reduces function call overhead and accelerates row-level security checks on large datasets.

This optimization applies to leads, cotizaciones, items_cotizacion, and email_logs tables. It addresses auth.uid() reevaluation issues and aligns with PLAN_MEJORAS tasks 1.3 and 2.1, ensuring consistent performance improvements across your security policies.

Create migration files at database/migrations/YYYYMMDD_optimize_rls_<tabla>.sql containing DROP POLICY and CREATE POLICY blocks that replace auth.uid() with (select auth.uid()). Apply migrations using mcp_supabase_apply_migration and verify results via mcp_supabase_get_advisors type="performance".

Yes. The optimization maintains the same security logic while improving execution. Use mcp_supabase_get_advisors to retrieve current policies, then generate migrations that replace only the auth.uid() evaluation pattern. Drop and recreate policies atomically to ensure no security gaps.

After applying migrations, run mcp_supabase_get_advisors with type="performance" to retrieve performance metrics and confirm the optimization reduced auth.uid() evaluation costs. Document results and performance gains in docs/PLAN_MEJORAS.md for tracking and auditing.