db-indexing (M11)

Indexing is the single heaviest Performance & Scale (axis performance) lever (relational weight 20, Indexación). This module checks whether the indexes that exist match how the data is queried — missing, mis-ordered, or wrong-type indexes turn a sub-millisecond lookup into a sequential scan.

What it checks

1. Composite order (ESR) — multi-column indexes should order columns Equality → Sort → Range. A leading range/low-selectivity column wastes the index.
2. Covering / partial — frequent SELECT a,b WHERE c benefits from an INCLUDE covering index; queries always filtered on a predicate (WHERE deleted_at IS NULL) benefit from a partial index.
3. Specialized types — GIN for jsonb/array/FTS containment, GiST/SP-GiST for geometry/range, BRIN for naturally-ordered append-only columns (timestamps). Flag full-text/geo/JSONB filters served by a plain B-tree (or no index).
4. FK-no-index — every foreign key column should have a covering index; an unindexed FK forces a sequential scan on the referenced side during ON DELETE/ON UPDATE and on joins. This is the highest-yield, lowest-risk index finding.