SDKMAN Database Migrations

This enables the users of SDKMAN to contribute new Installation Candidates and related Versions to be served by the API.

Tools used

This repo uses mongobee as database migration framework. In order to contribute a PR, it is required to have a local installation of MongoDB on your machine. Alternatively run it up with Docker (works on Linux):

    $ docker run -d --network=host --name=mongo mongo:3.2

The Build

The build is a standard Gradle build, with migration code written in Scala. To run migrations against your local database, simply run the following:

    $ ./gradlew clean run

The Model

The domain used for describing releases has two entities: candidates and versions. These are modelled as two distinct collections in our MongoDB datastore.

Candidates

This collection holds information about the SDK itself including the name, description, website url, distribution and default version. A typical entry would look something like this:

    { 
            "_id" : ObjectId("562beacb601daf84cec59999"),
            "candidate" : "scala", 
            "default" : "2.12.4", 
            "description" : "Scala is a programming language for general software applications. Scala has full support for functional programming and a very strong static type system...", 
            "websiteUrl" : "http://www.scala-lang.org/", 
            "name" : "Scala", 
            "distribution" : "UNIVERSAL"
    }

The distribution is usually set to UNIVERSAL unless platform specific binaries are to be served in which case it should be set to PLATFORM_SPECIFIC.

Versions

The versions collection will hold information about individual releases for a specific Candidate. It has fields representing the candidate, version, (absolute) URL to the binary, as well as platform.

    {
            "_id" : ObjectId("5a09d2dcffd8c740664b335b"), 
            "candidate" : "kotlin", 
            "version" : "1.1.60", 
            "platform" : "UNIVERSAL", 
            "url" : "https://github.com/JetBrains/kotlin/releases/download/v1.1.60/kotlin-compiler-1.1.60.zip"
    }

When serving up a simple universal zip binary, we always set the platform field to UNIVERSAL. Multi-platform SDKs (such as Java) usually come in several flavours ie. LINUX_64, MAC_OSX and WINDOWS_64. In these cases, multiple entries should be made for each platform binary:

    { "_id": ObjectId("xxx"), "candidate": "java", "version": "9.0.4", "platform": "LINUX_64", "url": "https://java_binary_linux.zip" },
    { "_id": ObjectId("xxx"), "candidate": "java", "version": "9.0.4", "platform": "MAC_OSX", "url": "https://java_binary_osx.zip" },
    { "_id": ObjectId("xxx"), "candidate": "java", "version": "9.0.4", "platform": "WINDOWS_64", "url": "https://java_binary_windows.zip" }

Workflow

Migration scripts can be found under changelogs and are divided by candidate. Various helper functions have been provided at package scope to perform simple tasks such as adding a Version / Candidate or setting a new Default version.

Adding a new Candidate migration class

Simply fork this repository and then add a db migration in the appropriate file (create a new class if your Candidate is not represented). Also ensure that the changelog order is set to the next value available among migration classes:

    @ChangeLog(order = "004")
    class JBakeMigrations {
        //TODO: migrations here
    }

Adding a new default Version (universal binary)

    @ChangeSet(order = "007", id = "007-add_scala_2_12_5", author = "marc0der")
      def migrate007(implicit db: MongoDatabase) = {
        Version("scala", "2.12.5", "https://downloads.lightbend.com/scala/2.12.5/scala-2.12.5.zip")
          .validate()
          .insert()
          .asCandidateDefault()
    }

Alternatively, a function is provided on package scope that allows the default version to be set explicitly:

    setCandidateDefault("groovy", "3.0.0")

Adding a historic Version (universal binary)

    @ChangeSet(order = "006", id = "006-add_scala_2_12_4", author = "marc0der")
      def migrate006(implicit db: MongoDatabase) =
        Version("scala", "2.12.4", "https://downloads.lightbend.com/scala/2.12.4/scala-2.12.4.zip")
          .validate()
          .insert()

Adding a new Version for multiple platforms

    @ChangeSet(order = "005", id = "005-add_oracle_jdk_10_0_0", author = "marc0der")
      def migrate005(implicit db: MongoDatabase) = {
        List(
          Version("java", "10.0.0-oracle", "http://download.oracle.com/java/jdk/10/7ea/jdk-10_osx-x64_bin.dmg", MacOSX),
          Version("java", "10.0.0-oracle", "http://download.oracle.com/java/jdk/10/7ea/jdk-10_linux-x64_bin.tar.gz", Linux),
          Version("java", "10.0.0-oracle", "http://download.oracle.com/java/jdk/10/7ea/jdk-10_windows-x64_bin.exe", Windows)
        ).validate().insert()
        setCandidateDefault("java", "10.0.0-oracle")
    }

Currently, four platforms identifiers are provided: Linux, Windows, MacOSX and Universal as the default.

Adding a new Candidate

    @ChangeSet(order = "001", id = "001_add_cxf_3_2_4", author = "r0b0")
    def migration001(implicit db: MongoDatabase) = {
        Candidate(
          candidate = "cxf",
          name = "CXF",
          description = "Apache CXF is an open source services framework...",
          websiteUrl = "https://cxf.apache.org/",
          distribution = "UNIVERSAL"
        ).insert()
    }

Removal of Candidates and Versions

    removeCandidate(candidate = "kobolt")
    removeAllVersions(candidate = "kobolt")
    removeVersion(candidate = "java", version = "10.0.0-oracle", platform = MacOSX)

Fast track

It is usually worth notifying us of the PR on the cli-dev Gitter chat in case we miss your PR.