Inter-module dependencies dependent on config possible in SBT? - sbt

I'm attempting to scope a dependency to a module in the same project using SBT's configurations.
In production, this dependency is satisfied by a jar on the classpath, but during dev it would be nice to do server/config-a:run or server/config-b:run to select the dependency manually.
Currently, I have something like this:
lazy val configA = config("config-a") extend Runtime
lazy val configB = config("config-b") extend Runtime
lazy val DevConfigA = Project(id = "dev-config-a", base = file("dev-config-a"))
lazy val DevConfigB = Project(id = "dev-config-b", base = file("dev-config-b"))
lazy val server = Project(id = "server",
base = file("server"),
dependencies = Seq(common))
.configs(configA, configB)
.dependsOn(DevConfigA % configA, DevConfigB % configB)
DevConfigA and DevConfigB bring in resources used for configuration. We want exactly one of them to be loaded. The goal is that server/config-a:run would depend on DevConfigA module, and not DevConfigB.
I had to move the configs and dependsOn out of the call to Project.apply to get it to compile. After that, the DevConfig* dependencies aren't showing up when I server/config-a:run or if I call show server/config-a:dependency-classpath.
Is there a way to make inter-module dependencies dependent on the config?

Yes, there's a way to make dependencies configuration-dependent - use libraryDependencies config-scoped.
I'm using the latest stable release of SBT.
[server]> show sbtVersion
[info] 0.13.1
Let's assume you need different versions of a library, e.g. scalaz, based upon what configuration you execute run with. As a matter of fact, you don't have to worry about the task, but the dependencies available in a given configuration, and since libraryDependencies drives it, I'm going to use it.
[server]> help libraryDependencies
Declares managed dependencies.
Here's the build.sbt that gives what you want.
build.sbt
lazy val configA = config("config-a") extend Runtime
lazy val configB = config("config-b") extend Runtime
lazy val server = project in file(".") configs(configA, configB)
val scalaz705 = "org.scalaz" %% "scalaz-core" % "7.0.5"
val scalaz710_M5 = "org.scalaz" %% "scalaz-core" % "7.1.0-M5"
libraryDependencies in configA += scalaz705
libraryDependencies in configB += scalaz710_M5
With the above build.sbt sbt lets us pick different versions of Scalaz based upon configuration.
[server]> show libraryDependencies
[info] List(org.scala-lang:scala-library:2.10.3)
[server]> show config-a:libraryDependencies
[info] List(org.scala-lang:scala-library:2.10.3, org.scalaz:scalaz-core:7.0.5)
[server]> show config-b:libraryDependencies
[info] List(org.scala-lang:scala-library:2.10.3, org.scalaz:scalaz-core:7.1.0-M5)

Related

How to set up a multi-project SBT project with inheritance and shared dependencies

I'd like to create an SBT project with inheritance and shared dependencies.
With Maven's POM files, there is the idea of Project Inheritance where you can set a parent project. I'd like to do the same thing with SBT.
The xchange-stream library uses Maven's Project Inheritance to resolve subproject dependencies when compiled from the parent project.
Here is my idea of what the file structure would look like:
sbt-project/
project/
dependencies.scala # Contains dependencies common to all projects
build.sbt # Contains definition of parent project with references
# to subprojects
subproject1/
build.sbt # Contains `subproject3` as a dependency
subproject2/
build.sbt # Contains `subproject3` as a dependency
subproject3/
build.sbt # Is a dependency for `subproject1` and `subproject2`
Where project1 and project2 can include project3 in their dependencies lists like this:
libraryDependencies ++= "tld.organization" % "project3" % "1.0.0"
Such that when subproject1 or subproject2 are compiled by invoking sbt compile from within their subdirectories, or when the parent: sbt-project is compiled from the main sbt-project directory, then subproject3 will be compiled and published locally with SBT, or otherwise be made available to the projects that need it.
Also, how would shared dependencies be specified in sbt-project/build.sbt or anywhere in the sbt-project/project directory, such that they are useable within subproject1 and subproject2, when invoking sbt compile within those subdirectories?
The following examples don't help answer either of the above points:
jbruggem/sbt-multiproject-example:
Uses recursive build.sbt files, but doesn't share dependencies among child projects.
Defining Multi-project Builds with sbt: pbassiner/sbt-multi-project-example:
Uses a single build.sbt file for the projects in their subdirectories.
sachabarber/SBT_MultiProject_Demo:
Uses a single build.sbt file.
Such that when subproject1 or subproject2 are compiled by invoking sbt compile from within their subdirectories...
Maybe Maven is meant to be used together with the shell environment and cd command, but that's not how sbt works at least as of sbt 1.x in 2019.
The sbt way is to use sbt as an interactive shell, and start it at the top level. You can then either invoke compilation as subproject1/compile, or switch into it using project subproject1, and call compile in there.
house plugin
A feature similar to parent POM would be achieved by creating a custom plugin.
package com.example
import sbt._
import Keys._
object FooProjectPlugin extends AutoPlugin {
override def requires = plugins.JvmPlugin
val commonsIo = "commons-io" % "commons-io" % "2.6"
override def buildSettings: Seq[Def.Setting[_]] = Seq(
organization := "com.example"
)
override def projectSettings: Seq[Def.Setting[_]] = Seq(
libraryDependencies += commonsIo
)
}
sbt-sriracha
It's not exactly what you are asking for, but I have an experimental plugin that allows you to switch between source dependency and binary dependency. See hot source dependencies using sbt-sriracha.
Using that you could create three individual sbt builds for project1, project2, and project3, all located inside $HOME/workspace directory.
ThisBuild / scalaVersion := "2.12.8"
ThisBuild / version := "0.1.1-SNAPSHOT"
lazy val project3Ref = ProjectRef(workspaceDirectory / "project3", "project3")
lazy val project3Lib = "tld.organization" %% "project3" % "0.1.0"
lazy val project1 = (project in file("."))
.enablePlugins(FooProjectPlugin)
.sourceDependency(project3Ref, project3Lib)
.settings(
name := "project1"
)
With this setup, you can launch sbt -Dsbt.sourcemode=true and it will pick up project3 as a subproject.
You can use Mecha super-repo concept. Take a look on the setup and docs here: https://github.com/storm-enroute/mecha
The basic idea is that you can combine dependent sbt projects (with their own build.sbt) under single root super-repo sbt project:
/root
/project/plugins.sbt
repos.conf
/project1
/src/..
/project/plugins.sbt
build.sbt
/project2
/src/..
/project/plugins.sbt
build.sbt
Please, note that there is no build.sbt in the root folder!
Instead there is repos.conf file. It contains definition of the sub-repos and looks like the folowing:
root {
dir = "."
origin = ""
mirrors = []
}
project1 {
dir = "project1"
origin = "git#github.com:some_user/project1.git"
mirrors = []
}
project2 {
dir = "project2"
origin = "git#github.com:some_user/project2.git"
mirrors = []
}
Then you can specify the Inter-Project, source-level Dependencies within individual projects.
There are two approaches:
dependencies.conf file
or in the build source code
For more details, please, see the docs

sbt aspectj with native packager

I'm attempting to use the sbt-aspectj plugin with the sbt native packager and am running into an issue where the associated -javaagent path to the aspectj load time weaver jar references an ivy cache location rather than something packaged.
That is, after running sbt stage, executing the staged application via bash -x target/universal/stage/bin/myapp/ results in this javaagent:
exec java -javaagent:/home/myuser/.ivy2/cache/org.aspectj/aspectjweaver/jars/aspectjweaver-1.8.10.jar -cp /home/myuser/myproject/target/universal/stage/lib/org.aspectj.aspectjweaver-1.8.10.jar:/home/myuser/myproject/target/universal/stage/lib/otherlibs.jar myorg.MyMainApp args
My target platform is Heroku where the artifacts are built before being effectively 'pushed' out to individual 'dynos' (very analogous to a docker setup). The issue here is that the resulting -javaagent path was valid on the machine in which the 'staged' deployable was built, but will not exist where it's ultimately run.
How can one configure the sbt-aspectj plugin to reference a packaged lib rather than one from the ivy cache?
Current configuration:
project/plugins.sbt:
addSbtPlugin("com.typesafe.sbt" % "sbt-aspectj" % "0.10.6")
addSbtPlugin("com.typesafe.sbt" % "sbt-native-packager" % "1.1.5")
build.sbt (selected parts):
import com.typesafe.sbt.SbtAspectj._
lazy val root = (project in file(".")).settings(
aspectjSettings,
javaOptions in Runtime ++= { AspectjKeys.weaverOptions in Aspectj }.value,
// see: https://github.com/sbt/sbt-native-packager/issues/598#issuecomment-111584866
javaOptions in Universal ++= { AspectjKeys.weaverOptions in Aspectj }.value
.map { "-J" + _ },
fork in run := true
)
Update
I've tried several approaches including pulling the relevant output for javaOptions from existing mappings, but the result is a cyclical dependency error thrown by sbt.
I have something that technically solves my problem but feels unsatisfactory. As of now, I'm including an aspectjweaver dependency directly and using the sbt-native-packager concept of bashScriptExtraDefines to append an appropriate javaagent:
updated build.sbt:
import com.typesafe.sbt.SbtAspectj._
lazy val root = (project in file(".")).settings(
aspectjSettings,
bashScriptExtraDefines += scriptClasspath.value
.filter(_.contains("aspectjweaver"))
.headOption
.map("addJava -javaagent:${lib_dir}/" + _)
.getOrElse(""),
fork in run := true
)
You can add the following settings in your sbt config:
.settings(
retrieveManaged := true,
libraryDependencies += "org.aspectj" % "aspectjweaver" % aspectJWeaverV)
AspectJ weaver JAR will be copied to ./lib_managed/jars/org.aspectj/aspectjweaver/aspectjweaver-[aspectJWeaverV].jar in your project root.
I actually solved this by using the sbt-javaagent plugin to adding agents to the runtime

Is there a way in sbt to only aggregate specific tasks?

My build consists of a root project, sub projects, and other meta projects. The sub projects produce a JAR while the meta projects are for other things like packaging etc.
See Cross platform build with SBT for the reason why I am doing this. For now assume this the right thing to do, but if not then feel free to comment on that question.
What I want to be able to do is to aggregate all my sub projects like so:
lazy val root = (project in file(".")).
aggregate(subProjects: _*)
val subProjects = Seq(projectA, projectB)
lazy val projectA = project
lazy val projectB = project
For the packaging I have something like:
lazy val packaging = (project in file("packaging").
enablePlugins(JavaServerAppPackaging). // Using sbt-native-packager
settings(
libraryDependencies ++= Seq(projectA.projectID, projectB.projectID)
)
Now before packaging I have to publish the sub projects:
sbt> publishLocal
And then to package from those published artifacts I do:
sbt> packaging/universal:packageBin
This all works great but now for the real question, how can I aggregate specific tasks? For example I want to be able to clean all projects including the packaging one.
I know that it is possible to do the opposite, see Disable aggregate for sbt custom task.
I've had a bit of a look at sbt.Aggregation but I can't make head or tail of it.
My ideal solution would look something like:
lazy val root = (project in file(".")).
aggregate(subProjects: _*).
settings(aggregatedTasks(metaTasks, metaProjects ): _*)
...
val metaProjects = Seq(packaging, someOtherProject)
val metaTasks = Seq(clean, someOtherTask)
def aggregatedTasks(tasks: Seq[TaskKey[_]], projects: Seq[Project]): Seq[Setting[_]] = {
??? // Is there a way to do this?
}

Include the contents of a dependency in (mappings in packageBin)

We assemble multiple projects into one jar:
val asmDep = "org.scala-lang.modules" % "scala-asm" % "5.0.3-scala-3"
lazy val compiler = (project in file(".") / "src" / "compiler")
.settings(
libraryDependencies ++= Seq(asmDep),
mappings in Compile in packageBin :=
(mappings in Compile in packageBin).value ++
(mappings in Compile in packageBin in LocalProject("interactive")).value ++
[...])
I'd like to also include the content of the asmDep dependency (it's a single JAR) into the project package. Is there an easy way to get that?
You want to use sbt-assembly.
I'm not sure if asmDep is the only dependency out of many in your project you want to add to your package or not, but there's a few different configuration options available in sbt-assembly that might suit your needs.

How do "transitive resolvers" work with SBT?

I have the following project build:
import sbt._
import Keys._
object ProjectBuild extends Build {
val buildVersion = "0.1-SNAPSHOT"
val delvingReleases = "Delving Releases Repository" at "http://development.delving.org:8081/nexus/content/repositories/releases"
val delvingSnapshots = "Delving Snapshot Repository" at "http://development.delving.org:8081/nexus/content/repositories/snapshots"
val delvingRepository = if (buildVersion.endsWith("SNAPSHOT")) delvingSnapshots else delvingReleases
lazy val root = Project(
id = "basex-scala-client",
base = file(".")
).settings(
organization := "eu.delving",
version := buildVersion,
resolvers += "BaseX Repository" at "http://files.basex.org/maven",
libraryDependencies += "org.basex" % "basex" % "7.2.1",
libraryDependencies += "org.specs2" %% "specs2" % "1.7.1" % "test",
publishTo := Some(delvingRepository),
credentials += Credentials(Path.userHome / ".ivy2" / ".credentials"),
publishMavenStyle := true
)
}
When I include the resulting library in another project, like so:
"eu.delving" %% "basex-scala-client" % "0.1-SNAPSHOT"
and I try to build that project, I get an error prompting me that the "org.basex % basex % 7.2.1" library referenced by that project cannot be found.
I have to go and add the resolver in the "client" project in order for the library to be found. Is there a way to avoid this?
There's no transitive resolvers, so the build user needs to know all the resolvers of all the transitive library dependencies. The benefit of this approach is that on the opensource projects, it encourages projects to publish to one of the known repositories connected to the known resolvers.
For corporate usage, you can prevent your traffic from going to unknown places introduced by some dependencies down the graph.
To share resolver settings within the organization, you can create an org-wide plugin.
Has this situation changed in the last 7 years, 10 months? I have a transitive library dependency at a custom repository. For its immediate client, I specify a resolver and the repository is written to the client's pom file when published. The client's client does not seem to use that information to find the transitive library. I have to "add the resolvers upstream by hand".

Resources