Blaireau

Blaireau is a minimalistic type safe SQL DSL for the Skunk library

⚠️ Blaireau is a personal project in addition to being a work in progress, hence the lack of documentation, code and functionalities.

If this project interests you, feel free to drop a ⭐ to encourage me working on this.

Table of Content

• Quickstart with sbt
• Metas
  • Datatype integrations
• Configuration
  • Field formatter
  • String Codec
  • Default Json Type
  • Using the configuration
• Table
  • Overriding sql names
• Field selection
• Filtering
• Queries and Commands builders
  • Select
    • Where
    • Querying
    • Example
  • Update
    • Commands
  • Delete
    • Commands
  • Insert
    • Commands

Quickstart with sbt

In your build.sbt add the following line:

resolvers += Resolver.sonatypeRepo("public")

// For the SQL DSL:
libraryDependencies += "fr.valentin-henry" %% "blaireau-dsl" % Version
// For Timepit's Refined Metas
libraryDependencies += "fr.valentin-henry" %% "blaireau-refined" % Version
// For Estatico's Newtype Metas
libraryDependencies += "fr.valentin-henry" %% "blaireau-newtype" % Version
// For Circe's Json Metas
libraryDependencies += "fr.valentin-henry" %% "blaireau-circe" % Version

Metas

A meta is a representation of the class in the database. It is derived using Shapeless.

At the moment, only simple types are supported.

A MetaS[T] is a meta instance of a type T. You can create any MetaS you want using the Meta.of(...) function. It requires only a skunk.Codec[T].

⚠️ It sould only be types, not case class or it will mess with the derivation!

Metas must be implicitly available in the scope where the Table and Meta generation take places.

Datatype integrations

Blaireau supports datatypes from the libraries below:

• Refined
• NewType
• Circe

Configuration

The configuration is used for meta derivation and field formatting.

Field formatter

In order to translate from the scala camelCase to the one used in postgresql, a field formatter should be given in the configuration, currently there are 6 available options:

• camelCase (same as Scala)
• PascalCase
• ALLCAPS
• lowercase
• snake_case
• UPPER_SNAKE_CASE

By default, Blaireau will use snake_case.

String Codec

The default string codec is configurable, by default it is set as text but it is possible to change it as varchar or any one required as long as it is a skunk.Codec[String].

This codec is used everywhere there is a String field.

Default Json Type

When using blaireau-circe, an implicit encoder is required for Json objects. The field jsonTypeAsJsonb is used to chose the format of the Json objects in the DB (json or jsonb).

By default it is set to jsonb

Using the configuration

The default configuration sets the following fields:

• field formatter: camel_case
• string codec: text
• default json type: jsonb

In case these configurations are not the one required, an implicit instance of a Configuration must be provided in the derivation scope like below:

In the following examples, the default configuration will be used.

Table

Blaireau maps a case class to a database object. This mapping is done through a Meta-object derivation.

To create a table, you just have to use the apply method of the Table object like below:

import blaireau.dsl._

import java.util.UUID

final case class Address(street: String, info: Option[String], postalCode: String, city: String, country: String)
final case class User(id: UUID, email: String, firstName: String, lastName: String, age: Int, address: Option[Address])

val users = Table[User]("users")

We will be using the table above for all following examples.

Overriding sql names

Sometimes it is necessary to override sql fields name in case an object has fields which does not correspond to the formatted one. For example:

CREATE TABLE childs (
  id            TEXT PRIMARY KEY,
  name          TEXT NOT EMPTY
  mail          TEXT NOT EMPTY,
  parents_name  TEXT NOT EMPTY,
  parents_email TEXT NOT EMPTY,
)

A class mapping directly the fields would look like this:

case class Child(id: UUID, name: String, mail: String, parentsName: String, parentsEmail: String)

However, we might want the information name and email contained in am information class, like below:

case class Information(name: String, email: String)
case class Child2(id: UUID, info: Information, parentsInfo: Information)

Using the classes above without overriding the fields names would not work as expected since there will be two "name" and two "email" columns.

val nonViableTable =  Table[Child2]("childs")
nonViableTable.select.where(_.id === UUID.randomUUID())
// SELECT id, name, email, name, email FROM childs WHERE id = $1

To fix this issue, we can tell blaireau to use the given name instead of the derived one.

val childs = Table[Child2]("childs")
  .columns(
    _.parentsInfo.name -> "parents_name",
    _.parentsInfo.email -> "parents_email"
  )

childs.select.where(_.id === UUID.randomUUID())
// SELECT id name, email, parent_name, parent_email FROM childs where id = $1

Field selection

This is used in the select function from the Select Query Builder and in the update function from the Update Command Builder.

The fields are accessed using their scala names (regardless of the format used in the db).

It is possible to select multiple columns using the ~ operator just like in Tapir.

Examples:

table.select(e => e.id)...
table.update(e => e.email ~ e.lastName)

Embedded objects and embedded fields are supported.

Examples:

table.update(_.address)
table.select(e => e.id ~ e.address.country)

Filtering

This is used in the where, whereAnd and whereOr functions of the builders.

Just like the select, the classes fields are accessed using their scala names (regardless of the format used in the db).

All boolean operations and boolean fields can be composed with those operators:

Type Constraint	Scala	Postgresql	Blaireau
Boolean	&&	AND	&&
Boolean	||	OR	||
Boolean	!	NOT	!

Equality operator are available for all fields

Type Constraint	Scala	Postgresql	Blaireau
Any	==	=	===
Any			=~=
Any	!=	<>	<>
Any			=!=

The difference between <> and =!=: the first one returns true if one of the field is not the same, the later checks for a full inequality.

The difference between =~= and ===: the first one returns true if one of the field is the same, the latter checks for full equality.

Examples:

// e is considered as the parameter of a where function of a Table[User]

e.email === "[email protected]" 
// email = $1

e.firstName <> "Patrick"
// first_name <> $1

e.address <> someAddress
// street <> $1 OR info <> $2 OR postal_code <> $3 OR city <> $4 OR country <> $5

e.address =!= someAddress
// street <> $1 AND info <> $2 AND postal_code <> $3 AND city <> $4 AND country <> $5

e.address === someAddress
// street = $1 AND info = $2 AND postal_code = $3 AND city = $4 AND country = $5

e.address =~= someAddress
// street = $1 OR info = $2 OR postal_code = $3 OR city = $4 OR country = $5

Comparison operators are available for the Numeric (Int, Float etc.), Strings and Temporal (Dates etc.)

Type Constraint	Scala	Postgresql	Blaireau
Numeric | String | Temporal	<	<	<
Numeric | String | Temporal	<=	<=	<=
Numeric | String | Temporal	>	>	>
Numeric | String | Temporal	>=	>=	>=

Example:

e.age < 5
// age < $1

String fields has the like function which pattern match unsing PostgreSQL syntax

Type Constraint	Scala	Postgresql	Blaireau
String		like	like

Example:

e.firstName.like("P%")
// first_name LIKE $1

Optional fields and objects have their own functions

Type Constraint	Scala	Postgresql	Blaireau
Optional	isEmpty	IS NULL	isEmpty
Optional	isDefined	IS NOT NULL	isDefined

Example:

e.address.isDefined
// street IS NOT NULL OR info IS NOT NULL OR postal_code IS NOT NULL OR city IS NOT NULL OR country IS NOT NULL

isEmpty is the negation of isDefined.

⚠️ Keep in mind that this function is not completely safe since it is looking at fields instead of the mapped object itself.

The function below allows you to interact with the object as if it was non-empty just like in scala.

Type Constraint	Scala	Postgresql	Blaireau
Optional	contains		contains
Optional	exists		exists
Optional	forall		forall

Example:

//val dummyAddress: Address = ???

e.address.contains(dummyAddress)
// street = $1 AND info = $2 AND postal_code = $3 AND city = $4 AND country = $5

e.address.exists(_.info.isDefined)
// ${Same as e.address.isDefined} AND info IS NOT NULL

e.address.forall(_.city === "Paris")
// ${Same as e.address.isEmpty} OR city = $1

TODO: creating own operators

Queries and Commands builders

Select

Only simple select queries can be generated using Blaireau, for more advance ones, it is better to use Skunk sql interpolator.

There are options regarding the selection of fields:

• Whole object selection
• Specific fields selection

For the whole object selection, the select function should not have parameters

val users: Query[Void, User] = users.select...

For a more specific selection of fields, it can be done using the ~ operator as explained in the Field selection section.

import skunk.~
val namesAndAge: Query[Void, String ~ Int] = users.select(e => e.firstName ~ e.age)...

Where

A select query has three functions which can be used for composing the where part (no where is considered as TRUE).

• .where(...) is the first function to use.
• .whereAnd(...) is appending the previous where and the one specified as parameter with an AND
• .whereOr(...) is like the previous one but with OR

Querying

Once the select query fits your needs, you can chose the function which fits your needs the best.

The toQuery function compiles the query into a Skunk Query.

import java.util.UUID
val findById: Query[UUID, User] = users.select.where(_.id === UUID.randomUUID()).toQuery

The queryIn returns the input parameter given to the query builder

val findChristopheOrRetired = users.select(_.id).where(e => e.firstName === "Christophe" || e.age >= 60)
// val findCORQuery: Query[String ~ Int, UUID] = findChristopheOrRetired.toQuery
val findCORQueryInputParams: String ~ Int = findChristopheOrRetired.queryIn
// Eq to: ("Christophe", 60)

Or you can use wrappers on Skunk which returns a F[_].

Example

import cats.effect.MonadCancelThrow
import blaireau.dsl._
import skunk.Session

final class UsersSql[F[_]: MonadCancelThrow](s: Resource[F, Session[F]]){
  val users = Table[User]("users")
  
  def findById(id: UUID): F[User] =
    s.use(users.select.where(_.id === id).unique(_))
    
  def findByCountry(country: String): F[fs2.Stream[F, User]] =
    s.use(users.select.where(_.address.country === country).stream(_))
    
  def findByEmail(email: String): F[Option[User]] =
    s.use(users.select.where(_.email === email).option(_))
    
  def findAllEmailsOfPeopleOverTheAgeOf(age: Int): F[fs2.Stream[F, String]] =
    s.use(users.select(_.email).where(_.age > age).stream(_))
}

Update

The Table's update function can be either empty or with a combination of assignments:

An assignment is done using the operators below:

Type	Scala	Postgresql	Blaireau
Any	=	=	:=
Numeric	-=	field = field - _	-=
Numeric	+=	field = field + _	+=

These assignments are combined using the <+> operator.

The pseudo-variadic update function supports up to 15 different assignation params.

def updateFullName(id: UUID, firstName: String, lastName: String) =
  users.update(u => (u.firstName := firstName) <+> (u.lastName := lastName)).where(_.id === id).command

def updateFullName2(id: UUID, firstName: String, lastName: String) =
  users
    .update(
      _.firstName := firstName,
      _.lastName := lastName
    ).where(_.id === id).command

Full class / embedded class is also supported

def updateAddress(id: UUID, address: Address) =
  users.update(_.address := address).where(_.id === id)
  
def updateFullUser1(user: User) = 
  users.update(_ := user).where(_.id === user.id)
  
def updateFullUser2(user: User) =
  users.update(user).where(_.id === user)

As you can see above, when the full class is updates, you can omit the := operator.

If you want to know more about the where function, it has the same dsl as the Select's Where.

Commands

Once your update command fits your needs, you have three functions which you can use.

The toCommand function returns a Skunk Command[...]

def updateUserAddressAndAgeCommand(id: UUID, address: Address, age: Int): Command[Address ~ Int ~ UUID] =
  users
    .update(u => (u.address := address) <+> (u.age := age))
    .where(_.id === id)
    .toCommand

The commandIn function returns the input parameters of the Command

val in: String ~ (String ~ String) = users
  .update(_.address.street := "Teerts Street")
  .where(e => e.firstName === "Chloe" && e.lastName === "Fontvi")
  .commandIn
// eq: ("Teerts Street", ("Chloe", "Fontvi"))

The execute function prepares and executes the command with the given Session

def updateUser(u: User): F[Completion] =
  users.update(u).where(_.id === u.id).execute(s)

Delete

Delete uses the same where dsl as Select or Update.

Commands

Once your delete command fits your needs, you have three functions which you can use.

The toCommand function returns a Skunk Command[...]

def deleteSpecificUser(id: UUID): Command[UUID] =
  users.delete.where(_.id === id).toCommand

The commandIn function returns the input parameters of the Command

val in: String = users.delete.where(_.firstName === "Valentin").commandIn
// eq: "Valentin"

The execute function prepares and executes the command with the given Session

def deleteUser(u: UUID): F[Completion] =
  users.delete.where(_.id === u.id).execute(s)

Insert

There are options regarding the insertion of fields:

• Whole object insertion
• Specific fields insertion

For the whole object insertion, the insert function should not have parameters

val insertUser: Command[User] = users.insert...

For a more specific insertion, it can be done using the ~ operator seen in the Field selection section.

import skunk.~
val createUserWithTheAddress: Command[UUID ~ Address] = users.insert(e => e.id ~ e.address)...
// INSERT INTO users(id, street, postalCode, city, country) VALUES ($1, $2, $3, $4, $5)...

You can insert either a single value or a list of values:

val dummyUser: User = ???
val insertUser = users.insert.value(dummyUser)

val dummyUserList: List[User] = ???
val insertAllUsers = users.insert.values(dummyUserList)

⚠️ The values insertion keeps in memory the size of the list given. The command created by this cannot be used with a list of a different size.

Commands

Once your insert command fits your needs, you have three functions which you can use.

The toCommand function returns a Skunk Command[...]

def insertUser(user: User): Command[User] =
  users.insert.value(user).toCommand

The commandIn function returns the input parameters of the Command

val dummyId = UUID("db373385-29cd-4380-9cf5-5501c94b91a1")
val dummyAddress = Address("Str", "Pc", "City", "Ctr")
val in: UUID ~ Address = users.insert(u => u.id ~ u.address).value(dummyId ~ dummyAddress).commandIn
// eq: (UUID("db373385-29cd-4380-9cf5-5501c94b91a1"), Address("Str", "Pc", "City", "Ctr"))

The execute function prepares and executes the command with the given Session

def insertUser(users: List[User]): F[Completion] =
  users.insert.values(users).execute(s)