SQL Connectors

Overview

The SQL schema connector (sql-schema-connector) provides a unified implementation for all SQL databases with database-specific “flavours” for PostgreSQL, MySQL, SQLite, SQL Server (MSSQL), and CockroachDB. Location: schema-engine/connectors/sql-schema-connector/

Architecture

SqlSchemaConnector

The top-level SQL connector wraps database-specific flavour implementations:

pub struct SqlSchemaConnector {
    inner: Box<dyn SqlConnector + Send + Sync + 'static>,
    host: Arc<dyn ConnectorHost>,
}

SQL Dialect

Each database has a dialect implementing database-specific behavior:

pub trait SqlDialect: Send + Sync + 'static {
    fn renderer(&self) -> Box<dyn SqlRenderer>;
    fn schema_differ(&self) -> Box<dyn SqlSchemaDifferFlavour>;
    fn schema_calculator(&self) -> Box<dyn SqlSchemaCalculatorFlavour>;
    fn destructive_change_checker(&self) -> Box<dyn DestructiveChangeCheckerFlavour>;
    fn datamodel_connector(&self) -> &'static dyn Connector;
    fn migrations_table(&self) -> Table<'static>;
    // ...
}

SQL Connector Trait

Internal trait for SQL-specific operations:

pub trait SqlConnector: Send + Sync + Debug {
    fn dialect(&self) -> Box<dyn SqlDialect>;
    fn describe_schema(&mut self, namespaces: Option<Namespaces>) -> BoxFuture<'_, ConnectorResult<SqlSchema>>;
    fn introspect(&mut self, namespaces: Option<Namespaces>, ctx: &IntrospectionContext) -> BoxFuture<'_, ConnectorResult<SqlSchema>>;
    fn create_database(&mut self) -> BoxFuture<'_, ConnectorResult<String>>;
    fn drop_database(&mut self) -> BoxFuture<'_, ConnectorResult<()>>;
    fn reset(&mut self, namespaces: Option<Namespaces>) -> BoxFuture<'_, ConnectorResult<()>>;
    // ...
}

Supported Databases

PostgreSQL
MySQL
SQLite
SQL Server
CockroachDB

PostgreSQL Connector

Provider: postgresqlSupported Versions: 9.6+, recommended 12+Features:

Full ACID transactions with all isolation levels
Native enums
JSON/JSONB with filtering support
Array types (scalar lists)
Full-text search
Multi-schema support
Extensions (PostGIS, pg_trgm, etc.)
Lateral joins
Partial indexes
Row-level security (RLS) aware

Capabilities:

CAPABILITIES = AdvancedJsonNullability | AnyId | AutoIncrement |
    AutoIncrementAllowedOnNonId | AutoIncrementMultipleAllowed |
    CompoundIds | CreateMany | CreateSkipDuplicates | Enums |
    NativeFullTextSearch | InsensitiveFilters | Json |
    JsonFiltering | ScalarLists | MultiSchema | NamedForeignKeys |
    NamedPrimaryKeys | UpdateableId | DecimalType |
    OrderByNullsFirstLast | NativeUpsert | InsertReturning |
    UpdateReturning | DeleteReturning | LateralJoin | PartialIndex

Connection String Format:

postgresql://USER:PASSWORD@HOST:PORT/DATABASE?schema=SCHEMA
postgresql://USER:PASSWORD@HOST:PORT/DATABASE?schema=SCHEMA&connection_limit=10
postgresql://USER:PASSWORD@HOST/DATABASE?host=/var/run/postgresql/

Parameters:

schema: Default schema (default: public)
connection_limit: Max connections (default: unlimited)
connect_timeout: Connection timeout in seconds
pool_timeout: Pool acquisition timeout
sslmode: SSL mode (prefer, disable, require)
sslcert, sslidentity, sslpassword: SSL certificate options
host: Unix socket path (alternative to hostname)

Native Types:

SmallInt, Integer, BigInt
Decimal(p,s), Money
Real, DoublePrecision
VarChar(n), Char(n), Text
ByteA
Timestamp(p), Timestamptz(p), Date, Time(p), Timetz(p)
Boolean
Uuid
Json, JsonB
Xml, Inet, Citext, Bit(n), VarBit(n)

Initialization:

#[cfg(feature = "postgresql-native")]
SqlSchemaConnector::new_postgres(params)?;

// Or auto-detect PostgreSQL vs CockroachDB
SqlSchemaConnector::new_postgres_like(params)?;

MySQL Connector

Provider: mysqlSupported Versions: 5.7+, 8.0+Features:

Transactions with isolation levels
Native enums
JSON columns (MySQL 5.7.8+)
Full-text indexes
Multiple full-text attributes per model
Index column length prefixing
Correlated subqueries

Capabilities:

CAPABILITIES = Enums | Json | AutoIncrementAllowedOnNonId |
    CreateMany | CreateSkipDuplicates | UpdateableId |
    JsonFiltering | CreateManyWriteableAutoIncId | AutoIncrement |
    CompoundIds | AnyId | NamedForeignKeys | IndexColumnLengthPrefixing |
    FullTextIndex | NativeFullTextSearch | DecimalType |
    OrderByNullsFirstLast | RowIn | CorrelatedSubqueries

Connection String Format:

mysql://USER:PASSWORD@HOST:PORT/DATABASE
mysql://USER:PASSWORD@HOST:PORT/DATABASE?socket=/tmp/mysql.sock
mysql://USER:PASSWORD@HOST:PORT/DATABASE?connection_limit=10&ssl_mode=REQUIRED

Parameters:

connection_limit: Max connections
connect_timeout: Connection timeout in seconds
pool_timeout: Pool acquisition timeout
socket: Unix socket path
sslcert, sslidentity, sslpassword, ssl_accept: SSL options

Native Types:

TinyInt, SmallInt, MediumInt, Int, BigInt
Decimal(p,s)
Float, Double
Bit(n)
Char(n), VarChar(n)
Binary(n), VarBinary(n)
TinyBlob, Blob, MediumBlob, LongBlob
TinyText, Text, MediumText, LongText
Date, Time(p), DateTime(p), Timestamp(p)
Year
Json

Initialization:

#[cfg(feature = "mysql-native")]
SqlSchemaConnector::new_mysql(params)?;

SQLite Connector

Provider: sqliteSupported Versions: 3.8.0+, recommended 3.35+Features:

File-based and in-memory databases
Transactions (serializable only)
Foreign key constraints (requires explicit enable)
JSON functions (3.38.0+)
Flexible type system

Capabilities:

CAPABILITIES = CompoundIds | InsensitiveFilters | JsonFiltering |
    CreateMany | ScalarLists | JsonLists | AutoIncrement | AnyId |
    NamedForeignKeys | CreateSkipDuplicates | UpdateableId |
    ImplicitManyToManyRelation | DecimalType

Connection String Format:

file:./dev.db
file:./dev.db?connection_limit=1
file:/absolute/path/to/dev.db
file::memory:

Parameters:

connection_limit: Max connections (default: 1, recommended for SQLite)
socket_timeout: Query timeout in seconds

Native Types: SQLite uses dynamic typing with type affinities:

Integer
Real
Text
Blob

Prisma maps to storage classes:

Int → INTEGER
BigInt → INTEGER
Float → REAL
Decimal → TEXT (stored as string)
String → TEXT
Boolean → INTEGER (0/1)
DateTime → TEXT (ISO8601)
Bytes → BLOB
Json → TEXT

Initialization:

#[cfg(feature = "sqlite-native")]
SqlSchemaConnector::new_sqlite(params)?;

// In-memory database
SqlSchemaConnector::new_sqlite_inmem(preview_features)?;

SQL Server (MSSQL) Connector

Provider: sqlserverSupported Versions: SQL Server 2017+, Azure SQL DatabaseFeatures:

Transactions with isolation levels
Named constraints
Computed columns
Filtered indexes
Multiple result sets

Capabilities:

CAPABILITIES = AnyId | CompoundIds | AutoIncrement |
    InsensitiveFilters | CreateMany | ScalarLists | Enums |
    NamedForeignKeys | NamedPrimaryKeys | AutoIncrementAllowedOnNonId |
    CreateSkipDuplicates | UpdateableId | ImplicitManyToManyRelation |
    DecimalType | OrderByNullsFirstLast | FilteredInlineChildNestedToOneDisconnect

Connection String Format:

sqlserver://HOST:PORT;database=DATABASE;user=USER;password=PASSWORD
sqlserver://HOST:PORT;database=DATABASE;integratedSecurity=true
sqlserver://HOST;database=DATABASE;user=USER;password=PASSWORD;encrypt=true

Parameters:

database: Database name (required)
schema: Default schema (default: dbo)
user, password: Authentication
integratedSecurity: Windows authentication
encrypt: Enable TLS encryption
trustServerCertificate: Trust server certificate
connection_limit: Max connections
connectTimeout: Connection timeout (ms)

Native Types:

TinyInt, SmallInt, Int, BigInt
Decimal(p,s), Numeric(p,s), Money, SmallMoney
Real, Float(n)
Bit
Char(n), NChar(n), VarChar(n), NVarChar(n), Text, NText
Binary(n), VarBinary(n), Image
Date, Time(p), DateTime, DateTime2(p), SmallDateTime, DateTimeOffset(p)
UniqueIdentifier
Xml

Initialization:

#[cfg(feature = "mssql-native")]
SqlSchemaConnector::new_mssql(params)?;

CockroachDB Connector

Provider: cockroachdbSupported Versions: 21.2+, recommended 22.1+Features:

PostgreSQL wire protocol compatible
Distributed SQL with strong consistency
Serializable isolation (default)
Same capabilities as PostgreSQL with some differences

Differences from PostgreSQL:

Default schema is public (cannot be changed)
Serial types use unique_rowid() instead of sequences
Some PostgreSQL extensions not available
Optimistic concurrency control

Connection String Format:

postgresql://USER:PASSWORD@HOST:PORT/DATABASE?schema=public

Same parameters as PostgreSQL.Initialization:

#[cfg(feature = "cockroachdb-native")]
SqlSchemaConnector::new_cockroach(params)?;

Migration Workflow

1. Schema Diffing

let migration = dialect.diff(from_schema, to_schema, &filter);

The differ generates migration steps:

CreateTable, DropTable, AlterTable
CreateIndex, DropIndex
CreateEnum, DropEnum, AlterEnum
AddColumn, DropColumn, AlterColumn
AddForeignKey, DropForeignKey
Database-specific steps (e.g., CreateExtension for PostgreSQL)

2. Script Rendering

let script = dialect.render_script(&migration, &diagnostics)?;

Renderers generate database-specific SQL:

PostgreSQL: DDL with IF EXISTS, CASCADE support
MySQL: DDL with backtick identifiers
SQLite: DDL with table recreations (no ALTER COLUMN)
SQL Server: DDL with GO batch separators

3. Destructive Change Checking

let diagnostics = connector.destructive_change_checker()
    .check(&migration)
    .await?;

Warns about:

Data loss (dropping tables, columns)
Type changes that may truncate data
Constraint changes that may fail on existing data

4. Migration Application

let steps_applied = connector.apply_migration(&migration).await?;

Steps:

Acquire advisory lock (if supported)
Start transaction (if supported for DDL)
Execute migration steps
Record in _prisma_migrations table
Commit and release lock

Introspection

Introspection scans the database and generates a Prisma schema:

let result = connector.introspect(&ctx, extension_types).await?;

Process:

Describe schema: Query system tables/catalogs
Build internal schema: Create SqlSchema representation
Calculate datamodel: Generate PSL from schema
Add warnings: Flag unsupported features

Generated elements:

Models from tables
Fields from columns with native types
Relations from foreign keys
Enums from database enums
Indexes and unique constraints
Default values

Shadow Database

For migration validation, SQL connectors can use a shadow database:

let schema = dialect.schema_from_migrations_with_target(
    migrations,
    namespaces,
    filter,
    ExternalShadowDatabase::ConnectionString {
        connection_string: shadow_url,
        preview_features,
    },
).await?;

The shadow database:

Is a temporary database for applying migrations
Validates migration scripts work correctly
Is automatically created and dropped
Can be user-provided or auto-generated

Multi-Schema Support

PostgreSQL and SQL Server support multiple schemas/namespaces:

datasource db {
  provider = "postgresql"
  url      = env("DATABASE_URL")
  schemas  = ["public", "auth", "billing"]
}

Connector handles:

Schema-qualified table names
Cross-schema relations
Search path configuration
Migration targeting specific schemas

Advisory Locks

PostgreSQL and MySQL use advisory locks to prevent concurrent migrations:

connector.acquire_lock().await?;

PostgreSQL: pg_advisory_lock()
MySQL: GET_LOCK()
Other databases: No-op or file-based locking

Disable with:

PRISMA_SCHEMA_DISABLE_ADVISORY_LOCK=1

Performance Considerations

Connection pooling: Configure connection_limit appropriately
Batch operations: Use create_records instead of multiple create_record
Relation loading: JoinStrategy for N+1 query prevention (database version dependent)
Indexes: Introspection captures all indexes for query optimization

Next Steps

MongoDB Connector

Learn about the MongoDB connector implementation

Libraries

Binaries

Connectors

Overview

Architecture

SqlSchemaConnector

SQL Dialect

SQL Connector Trait

Supported Databases

PostgreSQL Connector

MySQL Connector

SQLite Connector

SQL Server (MSSQL) Connector

CockroachDB Connector

Migration Workflow

1. Schema Diffing

2. Script Rendering

3. Destructive Change Checking

4. Migration Application

Introspection

Shadow Database

Multi-Schema Support

Advisory Locks

Performance Considerations

Next Steps

MongoDB Connector

Build docs developers (and LLMs) love

Libraries

Binaries

Connectors

​Overview

​Architecture

​SqlSchemaConnector

​SQL Dialect

​SQL Connector Trait

​Supported Databases

​PostgreSQL Connector

​MySQL Connector

​SQLite Connector

​SQL Server (MSSQL) Connector

​CockroachDB Connector

​Migration Workflow

​1. Schema Diffing

​2. Script Rendering

​3. Destructive Change Checking

​4. Migration Application

​Introspection

​Shadow Database

​Multi-Schema Support

​Advisory Locks

​Performance Considerations

​Next Steps

MongoDB Connector

Build docs developers (and LLMs) love

Overview

Architecture

SqlSchemaConnector

SQL Dialect

SQL Connector Trait

Supported Databases

PostgreSQL Connector

MySQL Connector

SQLite Connector

SQL Server (MSSQL) Connector

CockroachDB Connector

Migration Workflow

1. Schema Diffing

2. Script Rendering

3. Destructive Change Checking

4. Migration Application

Introspection

Shadow Database

Multi-Schema Support

Advisory Locks

Performance Considerations

Next Steps