slick-jdbc-extension CI Maven Central

An extension to slick.jdbc, Slick's plain SQL queries, including the following features.

These features can be selectively enabled.

Getting started

Add dependency in your build.sbt as the following.

    libraryDependencies ++= Seq(
      ...
      "com.github.tarao" %% "slick-jdbc-extension" % "0.1.0"
    )

The library is available on Maven Central. Currently, supported Scala versions are 2.12, and 2.13.

Overview

Slick supports plain SQL queries but it lacks some essential features. For example, there is no way to bind a list parameter to a prepared statement.

This extension provides additional features to Slick-style plain SQL queries. The extension consists of a family of traits and each of them can be enabled selectively. The easiest way to use the extension is to define a class with the traits.

Let's see an example of the usage of the extension. To start with, we assume to have an abstract class to run a DB query something like this.

abstract class DBHandler {
  import scala.concurrent.{Future, Await}
  import scala.concurrent.duration.Duration
  import slick.dbio.{DBIOAction, NoStream, Effect}
  import slick.backend.DatabaseComponent

  type Database = DatabaseComponent#DatabaseDef

  def run[R](a: DBIOAction[R, NoStream, Nothing])(implicit db: Database): R =
    Await.result(db.run(a), Duration.Inf)
}

Then a repository class to use plain SQL queries with the extension will look like this. (We are not interested in how the concrete DB object is supplied.)

import com.github.tarao.slickjdbc._
import interpolation.{SQLInterpolation, CompoundParameter, TableName}
import getresult.{GetResult, AutoUnwrapOption}

class SampleRepository extends DBHandler
    with SQLInterpolation with CompoundParameter
    with GetResult with AutoUnwrapOption {
  implicit def db: Database = ???

  ...
}

SQLInterpolation, CompoundParameter, GetResult and AutoUnwrapOption are feature selectors of the extension.

For the rest of the example, we use Entry model class defined as the following.

case class Entry(id: Long, url: String)

To define a mapping from a column name to a field of the model in the repository class, use getResult method defined in GetResult trait.

  implicit val getEntryResult = getResult { Entry(
    column("entry_id"),
    column("url")
  ) }

This will map entry_id column of a query result to the first field of Entry and url column to the second. This is very similar to Slick's GetResult.apply. The type of the column is inferred by the field type of Entry.id (it is Long in this case). Note that this example also uses AutoUnwrapOption to convert from Option[Long] to Long. The default column type is Option[_]ed since the resulting column value may be NULL.

Let's see an example of SQL interpolation. In addition to Slick's SQL interpolation, we can specify a structured value directly as a parameter.

  val table = TableName("entry")

  def add(entry: Entry) =
    run { sqlu"""
    | INSERT INTO ${table} (entry_id, url)
    | VALUES ${entry}
    """
  }

This actually illustrates three things.

  • You can specify a table name by a value of TableName. This value will be embedded literally in the raw query.
  • A value of case class (entry in this case) is expanded to placeholders like (?, ?) and the fields of the value are bound to them.
  • Characters before | is stripped in the raw query.

You can also specify a list as a parameter. But in this case, you must say that it is not empty.

  import eu.timepit.refined.api.Refined
  import eu.timepit.refined.collection.NonEmpty

  def findAll(entryIds: Seq[Long] Refined NonEmpty): Seq[Entry] = run {
    sql"""
    | SELECT * FROM ${table}
    | WHERE entry_id IN $ids
    """.as[Entry]
  }

$ids must be refined by NonEmpty and it expands to (?, ?, ..., ?). You can use refineV to refine values.

  import eu.timepit.refined.collection.NonEmpty
  import eu.timepit.refined.refineV

  val idsOrError = refineV[NonEmpty](Seq(101L, 102L, 103L))
  val entries = ids match {
    case Right(ids) => repository.findAll(ids)
    case Left(_)    => sys.error("Never happen in this case")
  }

See the documentation of refined for the detail about refined types.

If you were using nonempty from older versions, it still works as expected because nonempty.NonEmpty is a kind of refined collection type.

  import com.github.tarao.nonempty.NonEmpty

  def findAll(entryIds: Option[NonEmpty[Long]]): Seq[Entry] = entryIds match {
    case Some(ids) => run {
      sql"""
      | SELECT * FROM ${table}
      | WHERE entry_id IN $ids
      """.as[Entry]
    }
    case None => Seq.empty
  }

In this case, you call findAll method like the following since there is an implicit conversion from Seq[Long] to Option[NonEmpty[Long]].

repository.findAll(Seq(101L, 102L, 103L))

See the documentation of nonempty for the detail.

Additional types in SQL interpolation

To use the extended SQL interpolation, you have to first enable it. To enable the interpolation inside a particular trait or class, add SQLInterpolation trait to the trait or class.

import com.github.tarao.slickjdbc._
class YourRepositoryClass extends interpolation.SQLInterpolation

If you want to explicitly import the feature, then import methods in SQLInterpolation object.

import com.github.tarao.slickjdbc._
import interpolation.SQLInterpolation._

If you are importing a Slick driver API, doing that by import slick.driver.SomeDriver.api._ causes a conflict with the extended SQL interpolation since the API also defines a SQL interpolation of Slick. In this case, import only desired features to avoid the conflict. For example, if you are using Database from the API, import it as import slick.driver.SomeDriver.api.Database.

Literal parameters

When a value with Literal is passed as a parameter of the interpolation, a string returned by toString on the value is embedded literally in the raw query as if it was embedded by #${}. There are two predefined Literals.

  • SimpleString
    • takes a string as a parameter of the constructor and embed that string.
  • TableName
    • takes a name as a parameter of the constructor and embed that name.

Note that a Literal value is not expanded to a placeholder ?. If the value needs to be escaped as a part of a SQL string, then you should do it by yourself.

Non-empty lists

This feature requires interpolation.ComboundParameter trait or import interpolation.ComboundParameter._.

When a non-empty list is passed as a parameter of the interpolation, the value expands to ?s in the prepared statement, and items in the value are bound to the statement.

For example, if you have val Right(ne) = refineV[NonEmpty](Seq(1, 2, 3)) and embed it as sql"... ($ne) ...", the raw query will be ... (?, ?, ?) ... and 1, 2, 3 are bound to the placeholders. The resulting raw query will be the same if you omit parentheses like sql"... $ne ...".

If you pass a non-empty list of tuples, for example pairs, it expands to (?, ?), (?, ?), ..., (?, ?). This is useful to INSERT multiple values by INSERT INTO table (col1, col2) VALUES $aListOfPairs.

Case classes (products)

This feature requires interpolation.ComboundParameter trait or import interpolation.ComboundParameter._.

When a product especially an instance of a case class is passed as a parameter of the interpolation, the value expands to ?s in the prepared statement, and fields in the value are bound to the statement.

For example, if you have val p = SomeCaseClass(1, "foo") and embed it as sql"... ($p) ...", the raw query will be ... (?, ?) ... and 1, "foo" are bound to the placeholders. The resulting raw query will be the same if you omit parentheses like sql"... $p ...".

If you have another product value in a field of the passed value, it also be expanded and the placeholders are flattened.

You can also put a product value in a non-empty list and it has the same effect as putting tuples except when the arity of the product is 1. When the product arity is exactly 1, the value is treated as if it is a primitive value. In other words, a value of Seq[[Single] Refined NonEmpty (where Single is a Product of arity 1) expands to (?, ?, ..., ?) not to (?), (?), ..., (?). In contrast, a value of Seq[Tuple1[_]] Refined NonEmpty expands to (?), (?), ..., (?) not to (?, ?, ..., ?).

Note that passing Option[_] as an interpolation parameter is statically rejected and passing a product of arity zero (this can be made by for example case class Empty()) causes java.sql.SQLException at runtime.

Raw query translator

To use this feature, you have to enable the extended SQL interpolation. Follow the instruction above to enable it.

The query translator translates a raw query generated by the interpolation. A translation is simply from string to string and you can stack multiple translators.

The default translators

  • MarginStripper
    • Strip characters before | from each line of a query string. This makes a query log a bit more readable.
  • CallerCommenter
    • Record the code position that the interpolation is invoked as a comment of SQL query. You can trace a line number, a file name and a method name in a class from an ordinary query log on the DB server.

Custom translators

You can use a custom translator stack by defining an implicit value of Iterable[query.Translator]. For example, if you want to add a translator after the default translators, define an implicit value in a static scope where you invoke the SQL interpolation.

import com.github.tarao.slickjdbc._

implicit val translators: Iterable[query.Translator] = Seq(
  query.MarginStripper,
  query.CallerCommenter,
  new query.Translator {
    def apply(query: String, context: query.Context) =
      ??? /* return a translated string */
  }
)

Result mapper by column name

This feature requires getresult.GetResult trait or object.

With the trait, you can use getResult method and can use column method in the block of getResult without a receiver like in the example in Overview section.

If you prefer the object, you just need to import it.

import com.github.tarao.slickjdbc._
import getresult.GetResult

The usage of GetResult is quite similar to that of Slick's GetResult. You will receive an object to extract a result value in a block passed to GetResult.apply.

implicit val getEntryResult = GetResult { r => Entry(
  r.column("entry_id"),
  r.column("url")
) }

It also requires getresult.AutoUnwrapOption to extract a column value of non-Option[_] type. This also available both as a trait and an object.

You can use this feature without the extended SQL interpolation.

Mapping columns by name

As you have already seen, a column value can be extracted by column method. If you want to explicitly specify a return type or the return type somehow cannot be inferred, you can pass a type parameter to column to specify a return type.

implicit val getEntryResult = getResult { Entry(
  column[Long]("entry_id"),
  column[String]("url")
) }

Mapping columns by position

You can also extract the result by the index of a column.

implicit val getEntryResult = getResult { Entry(
  column(1),
  column(2)
) }

If you prefer the way of << method in Slick's GetResult, it is also available in our GetResult.

implicit val getEntryResult = getResult { Entry(<<, <<) }

<<? and skip are also available.

Option[_] values

getresult.AutoUnwrapOption makes it possible to get non-Option[_] value directly. Note that this may cause NoSuchElementException thrown if the value in the DB record is NULL.

If you want to specify a default value instead of throwing an exception, then retrieve the result as an Option[_] and give a default value.

implicit val getEntryResult = getResult { Entry(
  column[Option[Long]]("entry_id").getOrElse(0),
  column[Option[String]]("url").getOrElse("http://example.com/default/page")
) }

User-defined type binders

The way of extracting a typed value from a raw query result is different from that of Slick's GetResult. We provide a TypeBinder[_] which is similar to TypeBinder[_] in ScalikeJDBC to specify extractor methods for both indexed and named results.

Let's take url field of Entry as an example. Assume that you want to use your custom case class URL as a url field instead of String.

case class URL(url: String)
case class Entry(id: Long, url: URL)

Then you need to define a custom TypeBinder[_] for URL. A good way to define a custom type binder for a user-defined type is to extend a type binder for a primitive type.

implicit def urlBinder(implicit
  binder: TypeBinder[Option[String]]
): TypeBinder[Option[URL]] = binder.map(_.map(URL(_)))

This binder uses a TypeBinder[Option[String]] as a parent binder, which is defined by default, and convert the string retrieved by the parent to a URL. (The first map is to convert retrieved value, and the second one is to take care of the Option[_] type.)

Note that you should define a TypeBinder[_] for Option[URL] not for URL since unwrapping Option[_] will be done by general unwrapping conversion by AutoUnwrapOption.

With this binder, you can define a result mapper for Entry.

implicit val getEntryResult = getResult { Entry(
  column("entry_id"),
  column("url")
) }

Notice that there is no difference in the code compared to the one defined before for case class Entry(id: Long, url: String) but the (inferred) return type of column("url") is now URL.

License

  • Copyright (C) INA Lintaro
  • MIT License