Skip to content

Backends

Beam is backend-agnostic and doesn't provide any means to connect to a database. Beam backend libraries usually use well-used Haskell libraries to provide database connectivity. For example, the beam-sqlite backend uses the sqlite-simple backend.

Beam distinguishes each backend via type indexes. Each backend defines a type that is used to enable backend-specific behavior. For example, the beam-sqlite backend ships with the Sqlite type that is used to distinguish sqlite specific constructs with generic or other backend-specific ones.

Each backend can have one or more 'syntaxes', which are particular ways to query the database. While the beam-core library ships with a standard ANSI SQL builder, few real-world database implementations fully follow the standard. Most backends use their own custom syntax type. Internally, beam uses a finally-tagless representation for syntax trees that allow straightforward construction against any backend.

Beam offers backend-generic functions for the most common operations against databases. These functions are meant to fit the lowest common denominator. For example, no control is offered over streaming results from SELECT statements. While these backend-generic functions are useful for ad-hoc querying and development, it is wisest to use backend-specific functions in production for maximum control. Refer to backend-specific documentation for more information.

For our examples, we will use the beam-sqlite backend and demonstrate usage of the beam standard query functions.

Connecting to a database

Okay, so we can print out a SQL statement, but how do we execute it against a database? Beam provides a convenient MonadBeam type class that allows us to write queries in a backend agnostic manner. This is good-enough for most applications and preserves portability across databases. However, MonadBeam does not support features specific to each backend, nor does it guarantee the highest-performance. Most backends provide additional methods to query a database, and you should prefer these if you've committed to a particular backend. For tutorial purposes, we will use the beam-sqlite backend.

First, install beam-sqlite with cabal or stack:

$ cabal install beam-sqlite
# or
$ stack install beam-sqlite

Now, load beam-sqlite in GHCi.

Prelude> import Database.Beam.Sqlite
Prelude Database.Beam.Sqlite> 

Now, in another terminal, load the example database provided.

$ sqlite3 basics.db < beam-sqlite/examples/basics.sql

Now, back in GHCi, we can create a connection to this database.

Prelude Database.Beam.Sqlite> basics <- open "basics.db"
Prelude Database.Beam.Sqlite> runBeamSqlite basics $ runSelectReturningList (select (all_ (persons exampleDb)))
[ .. ]

The runSelectReturningList function takes a SqlSelect for the given syntax and returns the results via a list.

VoilĂ ! We've successfully created our first query and run it against an example database. We have now seen the major functionalities of the beam library. In the next section we'll explore more advanced querying and using relationships between tables.

Inserting data

First, let's connect to a sqlite database, and create our schema. The beam-core does not offer any support for the SQL DDL language. There is a separate core library beam-migrate that offers complete support for ANSI-standard SQL DDL operations, as well as tools to manipulate database schemas. See the section on migrations for more information.

For our example, we will simply issue a CREATE TABLE command directly against the database using sqlite-simple functionality:

Prelude Schema> execute_ conn "CREATE TABLE persons ( first_name TEXT NOT NULL, last_name TEXT NOT NULL, age INT NOT NULL, PRIMARY KEY(first_name, last_name) )"

Now we can insert some data into our database. beam-sqlite ships with a function runBeamSqlite, with the following signature:

runBeamSqlite :: Connection -> SqliteM a -> IO a

beam-sqlite uses the sqlite-simple library, so its handle type is Connection from Database.SQLite.Simple.

SqliteM is a monad implementing MonadBeam which we can use to construct database actions from individual SQL commands (select, insert, update, delete). MonadBeam is a type class that relates a particular SQL syntax (syntax) to a backend (be), and a command monad (m). Inside the m monad, we can execute data query and manipulation commands.

Let's insert some data into our database. We are going to use the runInsert function from MonadBeam. INSERTs are discussed in more detail in the data manipulation guide.

Prelude Schema> :{
Prelude Schema| runBeamSqlite conn $ do
Prelude Schema|   runInsert $ insert (persons exampleDb) $
Prelude Schema|               insertValues [ Person "Bob" "Smith" 50
Prelude Schema|                            , Person "Alice" "Wong" 55
Prelude Schema|                            , Person "John" "Quincy" 30 ]
Prelude Schema| :}

The runInsert function has the type signature

runInsert :: MonadBeam syntax be m => SqlInsert syntax -> m ()

SqlInsert syntax represents a SQL INSERT command in the given syntax. We construct this value using the insert function from Database.Beam.Query.

insert :: IsSql92InsertSyntax syntax =>
          DatabaseEntity be db (TableEntity table)
       -> Sql92InsertValuesSyntax syntax
       -> SqlInsert syntax

Intuitively, insert takes a database table descriptor and some values (particular to the given syntax) and returns a statement to insert these values. Sql92InsertValuesSyntax syntax always implements the IsSql92InsertValuesSyntax typeclass, which is where we get the insertValues function from. IsSql92InsertValuesSyntax also defines the insertSelect function for inserting values from the result of a SELECT statement. Other backends may provide other ways of specifying the source of values.

Now, we can query the database, using the runSelect function. Like runInsert and insert, we use the select function to construct a value of type SqlSelect syntax, which can be run inside MonadBeam.

We can use the runBeamSqliteDebug function to install a hook that beam will call with every SQL command it is about to run. In the following example, beam will print its query to stdout via putStrLn. You can use this functionality to hook beam in to a logging framework.

Prelude Schema> runBeamSqliteDebug putStrLn conn $ runSelect (select (all_ (persons exampleDb)))
[ Person { personFirstName = "Bob", personLastName="Smith", personAge=50 }, ... ]