Composite is a group of libraries focusing on practical uses of composite records, in particular Vinyl, such as querying records from a database and converting them to JSON. These libraries are based on the excellent Frames style use of Vinyl records, though composite implements its own derived from Frames to make for a smaller dependency graph, as Frames is a full CSV parsing/printing and data manipulation library.
Core library is located at https://github.com/composite-hs/composite-base.
composite-aeson
provides JSON formatting facilities for records. JSON formats can be derived automatically when default formats are available, explicitly assembled, combined, or a mix. Aeson's use of FromJSON
/ToJSON
type classes is mostly avoided to make using JSON formats first-class while still convenient.
Example:
{-# LANGUAGE DataKinds, OverloadedStrings, PatternSynonyms, TypeOperators #-}
import qualified Data.Aeson as Aeson
import Composite.Aeson (JsonFormat, defaultJsonFormatRecord, recordJsonFormat, toJsonWithFormat)
import Composite.Record (Record, Rec(RNil), (:->), pattern (:*:))
import Data.Text (Text)
type FId = "id" :-> Int
type FName = "name" :-> Text
type User = '[FId, FName]
userFormat :: JsonFormat e (Record User)
userFormat = recordJsonFormat defaultJsonFormatRecord
alice :: Record User
alice = 1 :*: "Alice" :*: RNil
aliceJson :: Aeson.Value
aliceJson = toJsonWithFormat userFormat alice
composite-aeson
support for the path library.
composite-aeson
support for the refined library.
Instance of Binary
from the binary library for composite records.
Autoconfiguration of EKG from a record of EKG metrics.
Instance of Hashable
from the hashable library for composite records.
composite-opaleye
provides the necessary instances to use a Vinyl record with the opaleye library, letting you use records for query expressions as well as result rows.
Example:
{-# LANGUAGE Arrows, DataKinds, FlexibleContexts, OverloadedStrings, PatternSynonyms, TemplateHaskell, TypeOperators #-}
import Control.Arrow (returnA)
import Composite.Opaleye (defaultRecTable)
import Composite.Record (Record, (:->))
import Composite.TH (withLensesAndProxies)
import Control.Lens (view)
import Data.Int (Int64)
import Data.Proxy (Proxy(Proxy))
import Data.Text (Text)
import Opaleye (Column, PGInt8, PGText, Query, Table(Table), (./=), asc, constant, orderBy, queryTable, restrict)
-- For each field type defined with, withLensesAndProxies will expand to the type, a record lens for the type,
-- and a proxy for the type, so for example FId is the type, fId is a lens which accesses the "id" field of any
-- record which contains that field, and fId_ is a proxy for the field type in case it's needed.
withLensesAndProxies [d|
type FId = "id" :-> Int64
type CId = "id" :-> Column PGInt8
type FName = "name" :-> Text
type CName = "name" :-> Column PGText
|]
type User = '[FId, FName]
type UserCols = '[CId, CName]
userTable :: Table (Record UserCols) (Record UserCols)
userTable = Table "users" defaultRecTable
userQuery :: Query (Record UserCols)
userQuery =
orderBy (asc $ view cName) $ proc () -> do
user <- queryTable userTable -< ()
let recId = view cId user
restrict -< recId ./= constant (1 :: Int64)
returnA -< user
Automatic derivation of Swagger 2 (ala swagger2) definitions for composite records.
compdoc
provides functionality for reading a Pandoc into a record.composite-dhall
providesToDhall
andFromDhall
instances for composite records.composite-tuple
provides utility functions for treating composite records as tuples, alaRelude.Extra.Tuple
from relude.fcf-composite
provides integration with first-class-families for type-level computation of records.polysemy-methodology-composite
provides functions for using polysemy-methodology with composite.
A small servant based server which uses composite-opaleye
to pull records from the database, reshape
the record to an API type, and send the records out to the client as JSON via composite-aeson
.
As of writing, we use these libraries in all our Haskell projects internally and have had no major issues. There are spots using either composite or vinyl where the compiler error messages could use improvement. There are certain use cases that can cause the simplifier to crash, though we have not observed any runtime errors as yet. They have not been proven out for performance at larger scale. We'd appreciate any fixes, improvements, or experience reports.
Contributions and feedback welcome! File an issue or make a PR.