Haskell PostgreSQL-typed

A Haskell PostgreSQL interface that provides type-safety through compile-time (template Haskell) database access. See the Haddock documentation in Database.PostgreSQL.Typed or the test cases for simple examples.

Getting started


Use your preferred package manager to install or add to your package dependencies:

  • stack install postgresql-typed or
  • cabal install postgresql-typed

You’ll also likely need to add network as a dependency.

Enable ghc extensions

Make sure you enable TemplateHaskell, QuasiQuotes, and DataKinds language extensions, either in your cabal default-extensions or in a {-# LANGUAGE TemplateHaskell, QuasiQuotes, DataKinds #-} pragma in your source.

Setup compile-time database connection

Either set the following environment variables:

  • TPG_DB the database name to use (default: same as user)
  • TPG_USER the username to connect as (default: $USER or postgres)
  • TPG_PASS the password to use (default: empty)
  • TPG_HOST the host to connect to (default: localhost)
  • TPG_PORT or TPG_SOCK the port number or local socket path to connect on (default port: 5432)

Or in your code call Database.PostgreSQL.Typed.useTPGDatabase with a database config as a top-level quote in each code file where you have SQL queries. It’s often helpful to make your own utility function to do this:

-- |Call this at top-level at the beginning of every file (rather than 'useTPGDatabase')
useMyTPGConfig :: Language.Haskell.TH.DecsQ
useMyTPGConfig = useTPGDatabase PGDatabase{ ... } -- or load config from file

Setup your database schema

Your tables and other schema need to be created in your development (compile-time) database before you compile your code. No queries will actually be executed, so there does not need to be any data, but it will do query parsing with the database (prepare queries) so any referenced objects must exist.

Setup run-time database connection

Use pgConnect to connect to your database using a PGDatabase configuration. The run-time database does not need to be the same as the build-time database (though it can be), but it must have the same schema. It’s recommended to use bracket (pgConnect PGDatabase{..}) pgDisconnect. If you need a pool of connections, consider resource-pool (while PGConnections are mostly thread-safe, they can’t be used for multiple queries simultaneously).

Complete example




{-# LANGUAGE OverloadedStrings #-}
module DBConfig where

import qualified Database.PostgreSQL.Typed as PG
import           Network.Socket (SockAddr(SockAddrUnix))

myPGDatabase :: PG.PGDatabase
myPGDatabase = PG.defaultPGDatabase
  { PG.pgDBAddr = if tcp then Left ("localhost", "5432") else Right (SockAddrUnix "/run/postgresql/.s.PGSQL.5432")
  , PG.pgDBUser = "user"
  , PG.pgDBName = "db"
  } where tcp = False


{-# LANGUAGE DataKinds #-}
{-# LANGUAGE QuasiQuotes #-}
{-# LANGUAGE TemplateHaskell #-}

import           Control.Exception (bracket)
import           Control.Monad (void, unless)
import           Data.Int (Int32)
import           Data.Maybe (listToMaybe)
import qualified Database.PostgreSQL.Typed as PG

import DBConfig

PG.useTPGDatabase myPGDatabase

data Thing = Thing Int32 String
  deriving (Eq)

createThing :: PG.PGConnection -> Thing -> IO ()
createThing pg (Thing tid tname) =
  void $ PG.pgExecute pg [PG.pgSQL|INSERT INTO thing (id, name) VALUES (${tid}, ${tname})|]

lookupThing :: PG.PGConnection -> Int32 -> IO (Maybe Thing)
lookupThing pg tid = fmap (uncurry Thing) . listToMaybe <$>
  PG.pgQuery pg [PG.pgSQL|SELECT id, name FROM thing WHERE id = ${tid}|]

main = bracket (PG.pgConnect myPGDatabase) PG.pgDisconnect $ \pg -> do
  let myt = Thing 1 "cat"
  createThing pg myt
  t <- lookupThing pg 1
  unless (t == Just myt) $ fail "wrong thing!"