work

work

A compact library for tracking and committing atomic changes to your entities.

What is it?

work does the heavy lifting of tracking changes that your application makes to entities within a particular operation. This is accomplished using what we refer to as a "work unit", which is essentially an implementation of the Unit Of Work pattern popularized by Martin Fowler. With work units, you no longer need to write any code to track, apply, or rollback changes atomically in your application. This lets you focus on just writing the code that handles changes when they happen.

Why use it?

There are a bundle of benefits you get by using work units:

• easier management of changes to your entities.
• rollback of changes when chaos ensues.
• centralization of save and rollback functionality.
• reduced overhead when applying changes.
• decoupling of code triggering changes from code that persists the changes.
• shorter transactions for SQL datastores.

How to use it?

The following assumes your application has types (fdm, bdm) that satisfy the SQLDataMapper and DataMapper interfaces, as well as *sql.DB (db).

Construction

Starting with entities Foo and Bar,

// type names.
fType, bType :=
	work.TypeNameOf(Foo{}), work.TypeNameOf(Bar{})

we can create SQL work units:

mappers := map[work.TypeName]work.SQLDataMapper {
	fType: fdm,
	bType: bdm,
}

unit, err := work.NewSQLUnit(mappers, db)
if err != nil {
	panic(err)
}

or we can create "best effort" units:

mappers := map[work.TypeName]work.DataMapper {
	fType: fdm,
	bType: bdm,
}

unit, err := work.NewBestEffortUnit(mappers)
if err != nil {
	panic(err)
}

Adding

When creating new entities, use Add:

additions := interface{}{Foo{}, Bar{}}
unit.Add(additions...)

Updating

When modifying existing entities, use Alter:

updates := interface{}{Foo{}, Bar{}}
unit.Alter(updates...)

Removing

When removing existing entities, use Remove:

removals := interface{}{Foo{}, Bar{}}
unit.Remove(removals...)

Registering

When retrieving existing entities, track their intial state using Register:

fetched := interface{}{Foo{}, Bar{}}
unit.Register(fetched...)

Saving

When you are ready to commit your work unit, use Save:

if err := unit.Save(); err != nil {
	panic(err)
}

Logging

We use zap as our logging library of choice. To leverage the logs emitted from the work units, simply pass in an instance of *zap.Logger upon creation:

l, _ := zap.NewDevelopment()

// create an SQL unit with logging.
unit, err := work.NewSQLUnit(mappers, db, work.UnitLogger(l))
if err != nil {
	panic(err)
}

Metrics

For emitting metrics, we use tally. To utilize the metrics emitted from the work units, pass in a Scope upon creation. Assuming we have an a scope s, it would look like so:

unit, err := work.NewBestEffortUnit(mappers, work.UnitScope(s))
if err != nil {
	panic(err)
}

Emitted Metrics

Uniters

In most circumstances, an application has many aspects that result in the creation of a work unit. To tackle that challenge, we recommend using Uniters to create instances of Unit, like so:

uniter := work.NewSQLUniter(mappers, db)

// create the unit.
unit, err := uniter.Unit()
if err != nil {
	panic(err)
}

Contribute

Want to lend us a hand? Check out our guidelines for contributing.

License

We are rocking an Apache 2.0 license for this project.

Code of Conduct

Please check out our code of conduct to get up to speed how we do things.

Artwork

Discovered via the interwebs, the artwork was created by Marcus Olsson and Jon Calhoun for Gophercises.