dhall-eta is a eta library that wraps the haskell implementation of dhall configuration language with a friendly java api.

Its main goal is to create a language binding to make possible configure java libraries and applications with dhall files.

We already can use eta to compile the dhall haskell implementation and etlas (a specific build tool for eta) to install dhall and its tools (dhall-to-json or dhall-text) as java cli applications. However, the classes generated by eta can't be used directly in java code in a easy way. Hence the need for this library.

Table of contents

• Api Overview
  • High level api
  • Low level api
• Building
  • etlas
  • gradle

API Overview

• The quick way to explore the java api is use jshell the java repl tool present since java 9.

  • To start a session with the library in the classpath you can do:

    > jshell --class-path /path/to/dhall-eta-all-${version}.jar

    NOTE: The version of java used is openjdk-9.0.4.

• To use the api in your project simply download from last release the dhall-eta-all-${version}.jar and put it in the project classpath.

• The project has an executable class with some more examples.

High level api

The main class to compile dhall expressions and get java objects is the class org.dhall.eta.Input.

• The class has methods to compile simple types and convert them to java:

jshell> import org.dhall.eta.Input

jshell> Input.bool("True")
$2 ==> true

jshell> Input.str("\"This is a dhall Text\"")
$3 ==> "This is a dhall Text"

jshell> Input.bigInt("-1234567890")
$4 ==> -1234567890

jshell> Input.natural("123456789 * 123456789")
$5 ==> Natural [15241578750190521]

• Those are shorthands for the main method Input.type. To use it we need to build a org.dhall.eta.Type<R> used by the api to extract a java value (of type R) and typecheck the dhall expression. Methods for basic types are already defined in org.dhall.eta.Types:

jshell> import org.dhall.eta.Type

jshell> import org.dhall.eta.Types

jshell> Type<Boolean> boolTy = Types.bool()
boolTy ==> eta.org.dhall.eta.Type$Eta@67bf0d91

jshell> Input.type(boolTy, "True && False")
$6 ==> false

• The class org.dhall.eta.Types has methods to build standard "container" types as List, Optional or Map.
  • For extract a java Map from a dhall record, we need to build a Map with the keys and types of the fields expected and then use Types.objMap to get a Type<Map<String, Object>> suitable to use with Input.type:

jshell> Map<String,Type<? extends Object>> fieldTypes=new HashMap<>();
fieldTypes ==> {}

jshell> fieldTypes.put("name",Types.str());
$9 ==> null

jshell> fieldTypes.put("nats",Types.list(Types.natural()))
$10 ==> null

jshell> Type<Map<String,Object>> mapTy=Types.objMap(FieldTypes.upcast(fieldTypes))
mapTy ==> eta.org.dhall.eta.Type$Eta@35884aff

jshell> Input.type(mapTy, "{ name = \"name\", nats=[1, 2, 3] }")
$11 ==> {name=name, nats=[Natural[1], Natural[2], Natural[3]]}

• Of course you can use the api to create an arbitrary java object from dhall code. We should use the generic Input.type method and create the appropiate type:
  • From a dhall record, using the method Types.record and implementing org.dhall.eta.RecordType
  • See an example implementation here.
  • From a dhall union, using the method Types.union and implementing org.dhall.eta.UnionType
  • See an example implementation here.
  • From arbitrary dhall code implementing the low-level org.dhall.eta.Type interface and using directly Input.type.

Low level api

• A more detailed overview of the low level api is coming soon.
• For now we can list the classes that can be used to handle each dhall compiler phase and a link to some example case uses:
  • Parsing: org.dhall.eta.Parser
  • Resolving imports: org.dhall.eta.Import
  • Normalizing and pretty printing: org.dhall.eta.Core
  • Typechecking: org.dhall.eta.TypeCheck
  • Binary encoding/decoding: org.dhall.eta.Binary

Building

Etlas

• Download etlas from https://eta-lang.org/docs/user-guides/eta-user-guide/installation/etlas

  • WARNING: Actually dhall-eta can only be built using eta and etlas built themselves from master. We are in the cutting edge!

• Run etlas build --enable-uberjar-mode to compile and generate an uberjar with all dependecies included in ./dist/build/eta-${version}/dhall-eta-${version}/x/dhall-eta-all/build/dhall-eta-all/dhall-eta-all.jar

  • The output jar is not minimized and include an example executable with etlas run dhall-eta-all

• Run etlas test to execute the test suite against dhall-lang acceptance tests

  • As a previous step you'll need to checkout dhall-lang project in the same parent directory as dhall-eta
  • Current dhall standard version supported is 4.0.0 and we have to use a fixed version of tests for 4.0.0 in https://github.com/jneira/dhall-lang
  • So the command to get them could be (being in the dhall-eta project dir):

    > git clone --branch v4.0.0 --depth=1 https://github.com/jneira/dhall-lang.git ../dhall-lang

  • The test suite does not check the validity of dhall-haskell (and it should not do it!), only that the dhall-haskell implementation compiled by eta and dhall-eta itself are consistent.

Gradle

• WARNING: For now, the gradle build is configured to use the system wide versions of eta and etlas so you will must have them built from source and available in PATH. We are in the cutting edge here too!
• This is the preferred method to build the project and the used one to generate the artifacts needed for a release of the library.
• It uses the gradle plugin for eta and the proguard and shadow ones to generate a minimized jar with all the dependencies included.
• It is recommended to use the gradle wrapper cause the eta gradle plugin will likely work with it.
• So ./gradlew build will generate the following artifacts:
  • A minimized uberjar ready to be used standalone in ./lib/dhall-eta-${version}-all.jar
  • An artefact with the objects representing the core business model of dhall. In this one will not be any code for actually parse, normalize o compile dhall code.
  • The package root name has not reference to eta: it is simply org.dhall.
  • The code is in a gradle subproject named dhall-model in the folder ./java
  • The artifact generated is ./java/build/libs/dhall-model-${version}.jar
  • An artefact with the java executable example org.dhall.eta.examples.Client.
  • The code is in a gradle subproject named dhall-eta-examples in the folder examples
  • You can run it with ./gradlew run.