Add a GiST index

Ancestor, descendant, and pattern queries on ltree need a Generalized Search Tree (GiST) index. Prisma’s default B-tree index only helps comparisons (<, =, >). Declare GiST in your contract, then emit and migrate.

Why you need GiST

isAncestorOf, isDescendantOf, matchesLquery, and matchesLtxtquery use PostgreSQL operators that a B-tree doesn’t cover. Without GiST, those queries still run, and PostgreSQL reads every row.

Declare GiST on the path column

Add type: "gist" on your path column. Path below is ltree.Ltree() from Author ltree columns.

model Page {
  id   String @id @default(uuid())
  path Path

  @@index([path], type: "gist")
  @@map("page")
}

Emit the contract:

pnpm prisma contract emit

Apply the change with db update or migration plan. Prisma creates a GiST index on the column:

CREATE INDEX … ON "page" USING "gist" ("path")

Use type: "gist" on an ltree[] column too (ltree.LtreeArray() or ltreeArray()).

You can keep @unique on the column. That unique constraint is a B-tree, and GiST is a second index.

Prisma doesn’t support siglen yet

siglen is a PostgreSQL operator-class argument (gist_ltree_ops(siglen=N)). You can’t pass it in @@index or in TypeScript options.

Your index uses PostgreSQL defaults: 8 bytes for ltree, 28 bytes for ltree[]. Ancestor, descendant, and pattern queries still use the index.

If you need a different siglen, create the index in SQL yourself. Prisma won’t manage that index on later db update runs.