prisma-normalize

Rewrite snake_case identifiers in Prisma schema files to PascalCase / camelCase while preserving the original database names via @map and @@map.

Useful after prisma db pull against a snake_case database when you want idiomatic camelCase TypeScript without touching the schema by hand.

Install

npm install -D prisma-normalize

Use

# walk a directory for .prisma files and rewrite in place
prisma-normalize ./prisma

# or a specific file
prisma-normalize ./prisma/schema.prisma

# CI mode: don't write, exit 1 if anything would change
prisma-normalize --check ./prisma

# print what would change without writing
prisma-normalize --dry-run ./prisma

Wire into your workflow:

{
  "scripts": {
    "prisma:pull": "prisma db pull && prisma-normalize ./prisma && prisma format",
    "lint:prisma": "prisma-normalize --check ./prisma"
  }
}

What it does

Given:

model blog_post {
  id          Int      @id @default(autoincrement())
  author_id   Int
  created_at  DateTime @default(now())
  title       String   @db.VarChar(255)
  author      user     @relation(fields: [author_id], references: [id])

  @@index([author_id])
}

It produces:

model BlogPost {
  id        Int      @id @default(autoincrement())
  authorId  Int      @map("author_id")
  createdAt DateTime @default(now()) @map("created_at")
  title     String   @db.VarChar(255)
  author    User     @relation(fields: [authorId], references: [id])

  @@index([authorId])
  @@map("blog_post")
}

The database schema is untouched — Prisma translates camelCase ↔ snake_case at query time.

Rules

• model / view / enum names in snake_case → PascalCase, with @@map("original_name") appended.
• Scalar field names in snake_case → camelCase, with @map("original_name") appended.
• Relation field names → camelCase, no @map (relations are not DB columns).
• Field type references update to the renamed type.
• @relation(fields: [...]), @@index([...]), @@unique([...]), @@id([...]) field lists update to the renamed field names.
• Already-normalized blocks are left alone — the transform is idempotent.
• Constraint names inside attributes (@id(map: "pk_…"), @@index([…], map: "idx_…")) are preserved untouched.

Multiple files

When you pass a directory, the CLI does a two-pass walk: first it scans every .prisma file to learn all model/enum/view names, then it transforms each file using that shared map. This is what makes cross-file type references rename correctly — e.g. a user relation field in post.prisma becomes User because model user was discovered in user.prisma.

After running, you'll usually want to chain prisma format to tidy column alignment:

prisma-normalize ./prisma && prisma format

Library API

import { normalizeSchema, buildTypeMap } from 'prisma-normalize'

// Single file — local typeMap built automatically:
const { output, changed } = normalizeSchema(schemaText)

// Multiple files sharing a typeMap (recommended for projects split across files):
const typeMap = buildTypeMap(allFileContents)
for (const content of allFileContents) {
  const { output } = normalizeSchema(content, typeMap)
  // write output back...
}

License

MIT