tranhieutt/prisma-expert

Prisma Expert

Critical rules (non-obvious)

• Never use migrate dev in production — use migrate deploy; migrate dev can reset data
• Singleton client in serverless — new PrismaClient() per request exhausts connections; use global singleton
• include vs select: include: { posts: true } fetches ALL post fields; use select to limit
• $queryRaw returns unknown[] — you must cast with as or validate; Prisma can't infer raw query types
• Missing @relation on both sides causes "The relation is not defined on both sides" runtime error

Schema: canonical model structure

model User {
  id        String   @id @default(cuid())
  email     String   @unique
  role      Role     @default(USER)
  posts     Post[]   @relation("UserPosts")
  profile   Profile? @relation("UserProfile")
  createdAt DateTime @default(now())
  updatedAt DateTime @updatedAt

  @@index([email])
  @@map("users")
}

model Post {
  id       String  @id @default(cuid())
  title    String
  content  String?
  published Boolean @default(false)
  author   User    @relation("UserPosts", fields: [authorId], references: [id], onDelete: Cascade)
  authorId String

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

enum Role { USER ADMIN MODERATOR }

Query optimization: N+1 fix

// ❌ N+1
const users = await prisma.user.findMany();
for (const user of users) {
  const posts = await prisma.post.findMany({ where: { authorId: user.id } });
}

// ✅ Include (fetches all post fields)
const users = await prisma.user.findMany({ include: { posts: true } });

// ✅✅ Select (fetch only needed fields — best for performance)
const users = await prisma.user.findMany({
  select: {
    id: true, email: true,
    posts: { select: { id: true, title: true } },
  },
});

// ✅ Complex aggregations → use raw
const result = await prisma.$queryRaw<{ id: string; count: number }[]>`
  SELECT u.id, COUNT(p.id)::int as count
  FROM users u LEFT JOIN posts p ON p.author_id = u.id
  GROUP BY u.id
`;

Transactions

// Sequential (auto-atomic, no rollback control)
const [user, profile] = await prisma.$transaction([
  prisma.user.create({ data: userData }),
  prisma.profile.create({ data: profileData }),
]);

// Interactive (full control, rollback on throw)
const result = await prisma.$transaction(async (tx) => {
  const user = await tx.user.create({ data: userData });
  if (user.email.endsWith("@blocked.com")) throw new Error("Blocked domain");
  return tx.profile.create({ data: { ...profileData, userId: user.id } });
}, {
  maxWait: 5000,
  timeout: 10000,
  isolationLevel: "Serializable",  // use ReadCommitted for most cases
});

Singleton client (Next.js / serverless)

// lib/prisma.ts
import { PrismaClient } from "@prisma/client";

const globalForPrisma = globalThis as unknown as { prisma: PrismaClient };

export const prisma = globalForPrisma.prisma ?? new PrismaClient({
  log: process.env.NODE_ENV === "development" ? ["query"] : [],
});

if (process.env.NODE_ENV !== "production") globalForPrisma.prisma = prisma;

Migration commands

# Development: creates migration file + applies
npx prisma migrate dev --name add_role_to_users

# Production: apply pending migrations only (safe)
npx prisma migrate deploy

# Check migration status
npx prisma migrate status

# Fix stuck migration
npx prisma migrate resolve --applied "20240115_migration_name"

# Validate schema (no DB connection needed)
npx prisma validate

# Format schema
npx prisma format

Common pitfalls

| Pitfall | Fix | |---|---| | migrate dev in production | Use migrate deploy | | New PrismaClient() per request | Use global singleton | | include: { all: true } on large models | Use select to fetch only needed fields | | P2034 (transaction conflict) | Retry with exponential backoff | | Shadow database error in dev | Set shadowDatabaseUrl or use Neon branching | | Enum not syncing | Run migrate dev after enum changes |