Overview
A transaction is a sequence of data operations performed as a single logical unit of work. Many relational databases support transactions to help enforce data consistency and business logic requirements.
A repository can perform operations in a transaction when the backing datasource is attached to one of the following connectors:
- MySQL connector (IMPORTANT: Only with InnoDB as the storage engine).
- PostgreSQL connector
- SQL Server connector
- Oracle connector
- DashDB Connector
- IBM Db2 (for Linux, Unix, Windows) Connector
- IBM Db2 for i connector
- IBM Db2 for z/OS connector
- Informix connector
The repository class needs to extend from TransactionalRepository
repository
interface which exposes the beginTransaction()
method. Note that LoopBack only
supports database local transactions - only operations against the same
transaction-capable datasource can be grouped into a transaction.
Transaction APIs
The @loopback/repository
package includes TransactionalRepository
interface
based on EntityCrudRepository
interface. The TransactionalRepository
interface adds a beginTransaction()
API that, for connectors that allow it,
will start a new Transaction. The beginTransaction()
function gives access to
the lower-level transaction API, leaving it up to the user to create and manage
transaction objects, commit them on success or roll them back at the end of all
intended operations. See Handling Transactions below
for more details.
Alternatively, you can also begin a transaction by calling beginTransaction()
method of DataSource
class.
Handling Transactions
See the API reference for full transaction lower-level API documentation.
Performing operations in a transaction typically involves the following steps:
- Start a new transaction.
- Perform create, read, update, and delete operations in the transaction.
- Commit or rollback the transaction.
Start transaction
Use the beginTransaction()
method to start a new transaction from a repository
class using DefaultTransactionalRepository
as a base class.
Here is an example:
import {
Transaction,
DefaultTransactionalRepository,
IsolationLevel,
} from '@loopback/repository';
// assuming there is a Note model extending Entity class, and
// ds datasource which is backed by a transaction enabled
// connector
const repo = new DefaultTransactionalRepository(Note, ds);
// Now we have a transaction (tx)
const tx = await repo.beginTransaction(IsolationLevel.READ_COMMITTED);
You can also extend DefaultTransactionalRepository
for custom classes:
import {inject} from '@loopback/core';
import {
juggler,
Transaction,
DefaultTransactionalRepository,
IsolationLevel,
} from '@loopback/repository';
import {Note, NoteRelations} from '../models';
export class NoteRepository extends DefaultTransactionalRepository<
Note,
typeof Note.prototype.id,
NoteRelations
> {
constructor(@inject('datasources.ds') ds: juggler.DataSource) {
super(Note, ds);
}
}
Isolation levels
When you call beginTransaction()
, you can optionally specify a transaction
isolation level. LoopBack transactions support the following isolation levels:
Transaction.READ_UNCOMMITTED
Transaction.READ_COMMITTED
(default)Transaction.REPEATABLE_READ
Transaction.SERIALIZABLE
If you don’t specify an isolation level, the transaction uses READ_COMMITTED .
Important:
Oracle only supports READ_COMMITTED and SERIALIZABLE.
For more information about database-specific isolation levels, see:
- MySQL SET TRANSACTION Syntax
- Oracle Isolation Levels
- PostgreSQL Transaction Isolation
- SQL Server SET TRANSACTION ISOLATION LEVEL
Perform operations in a transaction
To perform create, retrieve, update, and delete operations in the transaction,
add the transaction object to the Options
parameter of the standard
create()
,
update()
,
deleteAll()
(and so on) methods.
For example, again assuming a Note
model, repo
transactional repository, and
transaction object tx
created as demonstrated in
Start transaction section:
const created = await repo.create({title: 'Groceries'}, {transaction: tx});
const updated = await repo.update(
{title: 'Errands', id: created.id},
{transaction: tx},
);
// commit the transaction to persist the changes
await tx.commit();
Propagating a transaction is explicit by passing the transaction object via the options argument for all create, retrieve, update, and delete and relation methods.
Commit or rollback
Transactions allow you either to commit the transaction and persist the CRUD behaviour onto the database or rollback the changes. The two methods available on transaction objects are as follows:
/**
* Commit the transaction
*/
commit(): Promise<void>;
/**
* Rollback the transaction
*/
rollback(): Promise<void>;
Checking Activeness
For SQL connectors, function isActive()
is exposed to return the activeness of
the transaction by checking the existence of field connection
. It is also
available on the transaction objects:
/**
* Check if the transaction has an active connection
*/
isActive(): boolean;
Suppose you have a transaction called tx
, you can call tx.isActive()
to
check whether it’s still active.
Note:
The postgresql connector doesn’t fully support this function due to its special way of creating the transaction instance. After committing or rollbacking a transaction, the reference of its connection is not deleted when the datasource is postgresql. Details are explained in issue #411.
Set up timeout
You can specify a timeout (in milliseconds) to begin a transaction. If a transaction is not finished (committed or rolled back) before the timeout, it will be automatically rolled back upon timeout by default.
For example, again assuming a Note
model and repo
transactional repository,
the timeout
can be specified as part of the Options
object passed into the
beginTransaction
method.
const tx: Transaction = await repo.beginTransaction({
isolationLevel: IsolationLevel.READ_COMMITTED,
timeout: 30000, // 30000ms = 30s
});
Avoid long waits or deadlocks
Please be aware that a transaction with certain isolation level will lock database objects. Performing multiple methods within a transaction asynchronously has the great potential to block other transactions (explicit or implicit). To avoid long waits or even deadlocks, you should:
- Keep the transaction as short-lived as possible
- Don’t serialize execution of methods across multiple transactions
Accessing multiple models inside one transaction
The transaction object created by beginTransaction
is not model specific. If
you have multiple models attached to the same datasource, you can pass the same
transaction object to different repository instances.
For example, assuming we have ProductRepository
and CategoryRepository
attached to the same datasource that’s backed by a SQL database:
// Obtain repository instances. In a typical application, instances are injected
// via dependency injection using `@repository` decorator.
const categoryRepo = await app.getRepository(CategoryRepository);
const productRepo = await app.getRepository(ProductRepository);
// Begin a new transaction.
// It's also possible to call `categoryRepo.beginTransaction()` instead.
const transaction = await categoryRepo.dataSource.beginTransaction(
IsolationLevel.SERIALIZABLE,
);
// Execute database commands in the transaction
const c = await categoryRepo.create({name: 'Stationery'}, {transaction});
const p = await productRepo.create({name: 'Pen'}, {transaction});
// Finally commit the changes
await transaction.commit();