tomland: Bidirectional TOML serialization

[ configuration, library, mpl, program, text, toml ] [ Propose Tags ]

Implementation of bidirectional TOML serialization. Simple codecs look like this:

data User = User
    { userName :: Text
    , userAge  :: Int
    }

userCodec :: TomlCodec User
userCodec = User
    <$> Toml.text "name" .= userName
    <*> Toml.int  "age"  .= userAge

The following blog post has more details about library design:


[Skip to Readme]
Versions [faq] 0.0.0, 0.1.0, 0.2.0, 0.2.1, 0.3, 0.3.1, 0.4.0, 0.5.0, 1.0.0, 1.0.1.0, 1.1.0.0, 1.1.0.1, 1.2.0.0, 1.2.1.0, 1.3.0.0, 1.3.1.0
Change log CHANGELOG.md
Dependencies base (>=4.11 && <4.15), bytestring (==0.10.*), containers (>=0.5.7 && <0.7), deepseq (==1.4.*), hashable (>=1.2 && <1.4), megaparsec (>=7.0.5 && <9.1), mtl (==2.2.*), parser-combinators (>=1.1.0 && <1.3), text (==1.2.*), time (>=1.8 && <1.11), tomland, transformers (==0.5.*), unordered-containers (>=0.2.7 && <0.3), validation-selective (>=0.1.0.0 && <0.2) [details]
License MPL-2.0
Copyright 2018-2020 Kowainik
Author Dmitrii Kovanikov, Veronika Romashkina
Maintainer Kowainik <xrom.xkov@gmail.com>
Category TOML, Text, Configuration
Home page https://github.com/kowainik/tomland
Bug tracker https://github.com/kowainik/tomland/issues
Source repo head: git clone https://github.com/kowainik/tomland.git
Uploaded by vrom911 at 2020-09-21T17:03:25Z
Distributions Arch:1.3.1.0, LTSHaskell:0.5.0, NixOS:1.3.1.0, Stackage:1.0.0
Executables play-tomland, readme
Downloads 7576 total (142 in the last 30 days)
Rating 2.5 (votes: 3) [estimated by Bayesian average]
Your Rating
  • λ
  • λ
  • λ
Status Hackage Matrix CI
Docs uploaded by user
Build status unknown [no reports yet]

Modules

[Index] [Quick Jump]

Downloads

Maintainer's Corner

For package maintainers and hackage trustees


Readme for tomland-1.3.1.0

[back to package description]

tomland

palm

GitHub CI AppVeyor CI Hackage Stackage LTS Stackage Nightly MPL-2.0 license

“A library is like an island in the middle of a vast sea of ignorance, particularly if the library is very tall and the surrounding area has been flooded.”

― Lemony Snicket, Horseradish

tomland is a Haskell library for Bidirectional TOML Serialization. It provides the composable interface for implementing TOML codecs. If you want to use TOML as a configuration for your tool or application, you can use tomland to easily convert in both ways between textual TOML representation and Haskell types.

✍️ tomland supports TOML spec version 0.5.0.

The following blog post has more details about the library design and internal implementation details:

This README contains a basic usage example of the tomland library. All code below can be compiled and run with the following command:

cabal run readme

Preamble: imports and language extensions

Since this is a literate haskell file, we need to specify all our language extensions and imports up front.

{-# OPTIONS -Wno-unused-top-binds #-}

{-# LANGUAGE LambdaCase        #-}
{-# LANGUAGE OverloadedStrings #-}

import Control.Applicative ((<|>))
import Data.Text (Text)
import Data.Time (Day)
import Toml (TomlCodec, (.=))

import qualified Data.Text.IO as TIO
import qualified Toml

tomland is designed for qualified imports and intended to be imported as follows:

import Toml (TomlCodec, (.=))  -- add 'TomlBiMap' and 'Key' here optionally
import qualified Toml

Data type: parsing and printing

We're going to parse TOML configuration from examples/readme.toml file. The configuration contains the following description of our data:

server.port        = 8080
server.codes       = [ 5, 10, 42 ]
server.description = """
This is production server.
Don't touch it!
"""

[mail]
    host = "smtp.gmail.com"
    send-if-inactive = false

[[user]]
   guestId = 42

[[user]]
   guestId = 114

[[user]]
    login = "Foo Bar"
    createdAt = 2020-05-19

The above static configuration describes Settings for some server. It has several top-level fields, a table with the name mail and an array of tables with the name user that stores list of different types of users.

We can model such TOML using the following Haskell data types:

data Settings = Settings
    { settingsPort        :: !Port
    , settingsDescription :: !Text
    , settingsCodes       :: [Int]
    , settingsMail        :: !Mail
    , settingsUsers       :: ![User]
    }

data Mail = Mail
    { mailHost           :: !Host
    , mailSendIfInactive :: !Bool
    }

data User
    = Guest !Integer  -- id of guest
    | Registered !RegisteredUser  -- login and createdAt of registered user

data RegisteredUser = RegisteredUser
    { registeredUserLogin     :: !Text
    , registeredUserCreatedAt :: !Day
    }

newtype Port = Port Int
newtype Host = Host Text

Using the tomland library, you can write bidirectional converters for these types with the following guidelines and helper functions:

  1. If your fields are some simple primitive types like Int or Text you can just use standard codecs like Toml.int and Toml.text.
  2. If you want to parse newtypes, use Toml.diwrap to wrap parsers for underlying newtype representation.
  3. For parsing nested data types, use Toml.table. But it requires to specify this data type as TOML table in the .toml file.
  4. If you have lists of custom data types, use Toml.list. Such lists are represented as array of tables in TOML. If you have lists of the primitive types like Int, Bool, Double, Text or time types, that you can use Toml.arrayOf and parse arrays of values.
  5. If you have sets of custom data types, use Toml.set or Toml.HashSet. Such sets are represented as array of tables in TOML.
  6. For parsing sum types, use Toml.dimatch. This requires writing matching functions for the constructors of the sum type.
  7. tomland separates conversion between Haskell types and TOML values from matching values by keys. Converters between types and values have type TomlBiMap and are named with capital letter started with underscore. Main type for TOML codecs is called TomlCodec. To lift TomlBiMap to TomlCodec you need to use Toml.match function.
settingsCodec :: TomlCodec Settings
settingsCodec = Settings
    <$> Toml.diwrap (Toml.int  "server.port")       .= settingsPort
    <*> Toml.text              "server.description" .= settingsDescription
    <*> Toml.arrayOf Toml._Int "server.codes"       .= settingsCodes
    <*> Toml.table mailCodec   "mail"               .= settingsMail
    <*> Toml.list  userCodec   "user"               .= settingsUsers

mailCodec :: TomlCodec Mail
mailCodec = Mail
    <$> Toml.diwrap (Toml.text "host") .= mailHost
    <*> Toml.bool "send-if-inactive"   .= mailSendIfInactive

matchGuest :: User -> Maybe Integer
matchGuest = \case
   Guest i -> Just i
   _ -> Nothing

matchRegistered :: User -> Maybe RegisteredUser
matchRegistered = \case
   Registered u -> Just u
   _ -> Nothing

userCodec :: TomlCodec User
userCodec =
        Toml.dimatch matchGuest      Guest      (Toml.integer "guestId")
    <|> Toml.dimatch matchRegistered Registered registeredUserCodec

registeredUserCodec :: TomlCodec RegisteredUser
registeredUserCodec = RegisteredUser
    <$> Toml.text "login"     .= registeredUserLogin
    <*> Toml.day  "createdAt" .= registeredUserCreatedAt

And now we are ready to parse our TOML and print the result back to see whether everything is okay.

main :: IO ()
main = do
    tomlRes <- Toml.decodeFileEither settingsCodec "examples/readme.toml"
    case tomlRes of
        Left errs      -> TIO.putStrLn $ Toml.prettyTomlDecodeErrors errs
        Right settings -> TIO.putStrLn $ Toml.encode settingsCodec settings

Benchmarks and comparison with other libraries

You can find benchmarks of the tomland library in the following repository:

Since tomland uses 2-step approach with converting text to intermediate AST and only then decoding Haskell type from this AST, benchmarks are also implemented in a way to reflect this difference.

| Library | parse :: Text -> AST | transform :: AST -> Haskell | |--------------------|----------------------|-----------------------------| | tomland | 305.5 μs | 1.280 μs | | htoml | 852.8 μs | 33.37 μs | | htoml-megaparsec | 295.0 μs | 33.62 μs | | toml-parser | 164.6 μs | 1.101 μs |

In addition to the above numbers, tomland has several features that make it unique:

  1. tomland is the only Haskell library that has pretty-printing.
  2. tomland is compatible with the latest TOML spec while other libraries are not.
  3. tomland is bidirectional, which means that your encoding and decoding are consistent with each other by construction.
  4. tomland provides abilities for Generic and DerivingVia deriving out-of-the-box.
  5. Despite being the fastest, toml-parser doesn’t support the array of tables and because of that it’s hardly possible to specify the list of custom data types in TOML with this library. In addition, toml-parser doesn’t have ways to convert TOML AST to custom Haskell types and htoml* libraries use typeclasses-based approach via aeson library.

Acknowledgement

Icons made by Freepik from www.flaticon.com is licensed by CC 3.0 BY.