Libraries and tools for Polylith component based architecture.
Clone or download

readme.md

A leiningen plugin used to develop Polylith based architectures.

CircleCI

Welcome to the wonderful world of Polylith!

Polylith is a new way of thinking around system architecture, that puts the developer in the driving seat and the code in the center.

Polylith is a way of organising code into reusable building blocks that are used to create systems. To better understand the principles and ideas behind it, we recommend you first read the Polylith documentation.

Organising code as a Polylith can be done manually, which was actually how it all began. With that said, there is no magic behind this way of organising the code. It's not a framework nor a library, just a simple yet powerful way to work with code at the system level.

The reason we built this Leiningen plugin was to make life easier for you as a developer by making the work more efficient and fun.

Enjoy the ride!

Table of Contents

Installation

If you haven't done it already, start by installing Leiningen.

The next thing to do is to add the Polylith plugin to ~/.lein/profiles.clj. After editing the file it should look something like this:

{:user {:plugins [[polylith/lein-polylith "LATEST"]]}}

This ensures that the Polylith plugin can be called from anywhere in the file system and not just from the workspace root where the project.clj file with the Polylith declaration resides:

...
:plugins [[polylith/lein-polylith "0.2.1"]]

If called from the workspace root then it will use 0.2.1 in this case, otherwise it will use the latest version of the plugin.

Latest version

Clojars Project

Content

This documentation aims to be a practical guide to this plugin with lots of code examples. We encourage you to follow the code examples and try it out yourself. We will guide you through the steps of creating a workspace with systems composed of bases, components and libraries.

You will get an understanding of how the Polylith workspace is structured and how systems and environments are your views to components and bases by using symbolic links.

You will also learn how to test your building blocks frequently by using the test and build commands and how to keep libraries, systems and interfaces syncronized with the sync command.

We will walk you through how dependencies are detected and how the plugin keeps track of the time for the last successful test or build. We will explain how this information is used to build, compile and test your systems incrementally to shorten the feedback loop, how the dependencies can be listed with the deps command and used from several commands to stop you from introducing circular dependencies in your code.

We will explain the value of components and how they bring context to your development experience, which will help you build decoupled and scalable systems from day one.

Finally we will show what's next in the pipeline and how the extra power and flexibility will take your development experience to the next level.

Help

Go to Commands to read how to use the built-in help.

Realworld Example

If you want to have a look at a full-blown system, go to the RealWorld project where you can compare it with implementations made in other languages.

Workspace

The workspace is the top-level container for all your code and everything you need to create Polylith systems.

Let’s start by creating the example workspace with the top namespace se.example:

$ lein polylith create w example se.example

By default, the create workspace command adds a top-level folder that’s version controlled by git. If you don’t use git, then you can disabled this feature by adding the -git parameter to the command:

$ lein polylith create w example se.example -git

The workspace directory structure will end up like this:

example                         # root dir
  bases                         # empty dir
  components                    # empty dir
  environments
    development                 # the development environment
      ...
  interfaces
    project.clj                 # the workspace interfaces project file
    src                         # source dir
      se                        # top namespace: se.example
        example                 # empty dir (no source code added yet)
  images
    logo.png                    # used by readme.md
  project.clj                   # project file for the workspace
  readme.md                     # documentation
  systems                       # empty dir

When you get used to it, you will love this structure because everything lives where you expect to find it. The bases live in bases, components in components, systems in systems, development environments in environments and workspace interfaces in the interfaces’ src directory beneath the top namespace se.example.

If you, for example, have the top namespace com.a.b.c and the component user then all its namespaces will live under the namespace path com.a.b.c.user. If the user component has a core namespace then it will automatically get the namespace com.a.b.c.user.core.

Right now the plugin doesn’t support changing the name of the top namespace. The advice is therefore to think carefully when deciding the name of the top namespace.

The plugin doesn’t use configuration files, annotations, dependency injection or similar to assemble its parts. Instead it uses symbolic links directly in the file system. So just look in the file system or use the info command to inspect all the parts of the workspace.

If you want to mix different languages without the risk of introducing name conflicts, we recommend you to include the name of the language in the top namespace, e.g.: com.a.b.c.clojure, com.a.b.c.java or com.a.b.c.scala. If you always include it, it allows you to safely work with several languages (one per workspace) in the future.

The development environment contains these files:

      docs                      # empty dir
      interfaces                # link to interfaces/src
      project-files
        bases                   # empty dir
        components              # empty dir
        interfaces-project.clj  # link to interfaces/project.clj
        systems                 # empty dir
        workspace-project.clj   # link to project.clj
      project.clj
      resources                 # empty dir
      src                       # source dir
        se                      # top namespace: se.example
          example               # empty dir (no source code added yet)
      test                      # source test dir
        se                      # top namespace: se.example
          example               # empty dir (no source code added yet)

The development directory will be the root directory when you work with the code in your development environment. Whenever you create a new component or base, the development environment will be updated automatically, letting you concentrate on writing code!

To make sure that git doesn’t remove empty directories, the plugin will put a .keep file in the empty ones. It’s important that they are not removed when you clone a repository (workspace) otherwise the plugin will not work properly.

Let’s have a look at some files in the workspace so we get an idea of what’s in there.

The Leiningen project.clj file defines which version of Polylith plugin to use, the name of the top namespace and the Clojure version to use:

(defproject se.example/example "1.0"
  :description "A Polylith workspace."
  :plugins [[polylith/lein-polylith "0.2.1"]]
  :polylith {:top-namespace "se.example"
             :clojure-version "1.9.0"})

The interfaces/project.clj file describes how to compile all the workspace interfaces and looks like this:

(defproject se.example/interfaces "1.0"
  :description "Component interfaces"
  :dependencies [[org.clojure/clojure "1.9.0"]]
  :aot :all)

The environments/development/project.clj file is the project file for the development environment and looks like this:

(defproject se.example/development "1.0"
  :description "The main development environment."
  :dependencies [[org.clojure/clojure "1.9.0"]])

We will talk more about these files later.

Go to the newly created workspace:

$ cd example

If we now run the info command, it will list the contents of the workspace:

$ lein polylith info

interfaces:
components:
bases:
systems:
environments:
  development

...which is empty at the moment because we haven’t added anything yet.

The prompt

A nice way of working with the plugin is to start a prompt:

$ lein polylith prompt

This will enter a prompt within the example namespace:

example$>

The benefits are that commands will execute instantly and that you don’t need to repeat “lein polylith” before every command. All commands can be executed from any directory within the workspace root during the prompt session.

If you haven't started a prompt, all commands except help and create (workspace) must be executed from the workspace root.

We can execute the info command again by just typing info:

example$> info
interfaces:
components:
bases:
systems:
environments:
  development

Type quit or exit to exit:

example$> exit

Everywhere in this documentation where we say you should type something like lein polylith command we just mean command if you are inside a prompt.

Feel free to start a prompt, it will save you some time and typing!

System

A system consists of a base at the bottom with components and libraries above:
System

This is well explained in the Polylith documentation. The idea is to compose systems with smaller building blocks instead of putting everything into one place as a monolith. To describe a system we need to start with the base.

Base

The base is the foundation of a system:
Base

A system can only have one base. The base exposes a public API to the outer world, illustrated as sockets in the metaphor. In the cmd-line base that we will soon create, the public API will consist of a single main function that prints out “Hello world!”:

(defn -main [& args]
  (println "Hello world!"))

Let’s create the cmd-line system:

$ lein polylith create s cmd-line cmd-line

This will create a system with the name cmd-line that has a base with the name cmd-line. The development environment and the cmd-line system will now both contain the cmd-line base (in blue):

Run the info command again to see the result:

$ lein polylith info

interfaces:
components:
bases:
  cmd-line *
systems:
  cmd-line *
    cmd-line *   -> base
environments:
  development
    cmd-line *   -> base

We can see that the cmd-line base and the cmd-line system were created. The cmd-line base is automatically added to the cmd-line system and to the development environment. The “-> base” means that they both link to a base. The * symbols indicate that something has changed since the last successful test or build (more on that later).

The workspace now looks like this on disk:

example
  bases
    cmd-line                # new 
      project.clj           # new (cmd-line build file)
      readme.md             # new (almost empty documentation)
      resources
        cmd-line            # new (empty dir - to put cmd-line resources)
      src/se/example
        cmd-line            # new
          core.clj          # new (cmd-line core namespace)
      test/se/example
        cmd-line            # new
          core_test.clj     # new (cmd-line core-test namespace)
  environments
    development
      docs
        cmd-line-readme.md  # new (link to bases/cmd-line/readme.md)
      project-files
        bases
          cmd-line-project.clj  # new (link to bases/cmd-line/project.clj)
        systems
          cmd-line-project.clj  # new (link to systems/cmd-line/project.clj)
      resources
        cmd-line     # new (link to bases/cmd-line/resources/cmd-line)
      src/se/example
        cmd-line     # new (link to bases/cmd-line/src/se/example/cmd-line)
      test/se/example
        cmd-line     # new (link to bases/cmd-line/test/se/example/cmd-line)
  systems
    cmd-line
      build.sh
      project.clj
      readme.md
      resources
        cmd-line  # link to bases/cmd-line/resources/cmd-line
      src/se/example
        cmd-line  # link to bases/cmd-line/src/se/example/cmd-line

If we have a look at some of the generated files, we first find bases/cmd-line/src/se/example/cmd_line/core.clj:

(ns se.example.cmd-line.core
  (:gen-class))

;; A stand alone base example. Change to the right type of base.
(defn -main [& args]
  (println "Hello world!"))

...with its corresponding test bases/cmd-line/test/se/example/cmd_line/core_test.clj:

(ns se.example.cmd-line.core-test
  (:require [clojure.test :refer :all]
            [se.example.cmd-line.core :as core]))

;; Add tests here...
(deftest hello-world-example-test
  (let [output (with-out-str (core/-main))]
    (is (= "Hello world!\n"
           output))))

...then we also have systems/cmd-line/project.clj which is the project file for the cmd-line system:

(defproject se.example/cmd-line "0.1"
  :description "A cmd-line system."
  :dependencies [[org.clojure/clojure "1.9.0"]]
  :aot :all
  :main se.example.cmd-line.core)

...and systems/cmd-line/build.sh which is the script that performs the final step to build an artifact for the system, in this case an uberjar:

#!/usr/bin/env bash
set -e

lein uberjar

Let’s run the build command:

$ lein polylith build
updated: bases/cmd-line/project.clj

Changed components:
Changed bases: cmd-line
Changed systems: cmd-line

Compiling workspace interfaces
Created /Users/joakimtengstrand/IdeaProjects/example/interfaces/target/interfaces-1.0.jar
Wrote /Users/joakimtengstrand/IdeaProjects/example/interfaces/pom.xml
Installed jar and pom into local repo.

Compiling bases/cmd-line

Compiling systems/cmd-line

Start execution of tests in 1 namespaces:
lein test se.example.cmd-line.core-test

lein test se.example.cmd-line.core-test

Ran 1 tests containing 1 assertions.
0 failures, 0 errors.

Building systems/cmd-line
Created /Users/joakimtengstrand/IdeaProjects/example/systems/cmd-line/target/cmd-line-0.1.jar
Created /Users/joakimtengstrand/IdeaProjects/example/systems/cmd-line/target/cmd-line-0.1-standalone.jar

set :last-success in .polylith/time.edn

Execution time: 31.2 seconds

A build performs these steps:

  1. Checks for circular dependencies and quits if found.
  2. Calculates the components and bases to build, based on what has changed since the last successful test or build.
  3. Calls sync and makes sure that:
    • all library dependencies in project.clj files are in sync
    • the workspace interfaces are in sync with the component interfaces
    • all systems have the libraries and components they need
  4. AOT-compiles changed components, bases and systems to check that they compile against the workspace interfaces.
  5. Runs tests for all bases and components that have been affected by the changes.
  6. Executes build.sh for all changed systems to make sure they have a working build script and no missing components or libraries.
  7. If the entire build is successful, then execute the success command that updates the time for the last successful test or build.

We can now execute the jar and see if it works:

$ java -jar systems/cmd-line/target/cmd-line-0.1-standalone.jar
Hello world!

Yes it did!

If we run the info command again:

$ lein polylith info

interfaces:
components:
bases:
  cmd-line
systems:
  cmd-line
    cmd-line   -> base
environments:
  development
    cmd-line   -> base

...we can see that all the *s are gone. The reason is that no files have been changed since the successful build that we just did. Let’s have a look at the file where the time of the last successful test or build is stored:

cat .polylith/time.edn
{:last-success 1529166522000}

1529166522000 is the number of milliseconds since 1970. You can see a more readable version of it by executing the settings command:

$ lein polylith settings
...
bookmarks:
  2018-06-16 18:28:42 last-success (1529166522000)

The plugin uses the date format: yyyy-mm-dd hh:mm:ss.

Component

Components are the main building blocks in the Polylith world which are used to compose systems:
Component

A component consists of an implementation, an interface and dependencies to other components and libraries.

The component interface will be explained in the next section.

Let’s create the user component:

$ lein polylith create c user

If you leave out the third argument of create, the component interface will get the same name as the component, which is user in this case.

The user component was created and added to the development environment:

Run the info command again to see the result:

$ lein polylith info

interfaces:
  user *
components:
  user *
bases:
  cmd-line
systems:
  cmd-line
    cmd-line   -> base
environments:
  development
    user *     -> component
    cmd-line   -> base

We can see that the user interface and the user component were created (and are marked as changed). The user component was also added to the development environment where “-> component” indicates that it links to the user component.

The user-related parts in the workspace now look like this on disk:

example
  components
    user                   # new (user component dir)
      readme.md            # new (empty documentation)
      project.clj          # new (user build file)
      resources
        user               # new (empty dir - to put user resources)
      src/se/example
        user               # new
          core.clj         # new (user core namespace)
          interface.clj    # new (user interface namespace)
      test/se/example
        user               # new
          core_test.clj    # new (user core test namespace)
  environments
    development
      docs
        user-readme.md     # new (link to components/user/readme.md)
      project-files
        components
          user-project.clj  # new (link to components/user/project.clj)
      resources
        user                # new (link to components/user/resources/user)
      src/se/example
        user             # new (link to components/user/src/se/example/user)
      test/se/example
        user             # new (link to components/user/test/se/example/user)
  interfaces
    src/se/example
      user                 # new (public user interface dir)
        interface.clj      # new (user interface namespace)

The symbolic links in environments/development form a view to your components and bases that makes it possible to work with the Polylith code from one place.

Some example code was created in components/user/src/se/example/user/core.clj:

(ns se.example.user.core)

;; add your functions here...
(defn add-two [x]
  (+ 2 x))

...and its test in components/user/test/se/example/user/core_test.clj:

(ns se.example.user.core-test
  (:require [clojure.test :refer :all]
            [se.example.user.interface :as user]))

;; add your tests here...
(deftest test-add-two
  (is (= 42 (user/add-two 40))))

You are not forced to put your code in the core namespace. It’s just an example and can be changed to some other name and/or divided up into several namespaces.

If we go to the development directory, we can now execute all the tests (the cmd-line base test and the user component test):

$ cd environments/development
$ lein test

lein test se.example.cmd-line.core-test

lein test se.example.user.core-test

Ran 2 tests containing 2 assertions.
0 failures, 0 errors.

An alternative way of running the tests is:

$ cd ../..
$ lein polylith test
updated: components/user/project.clj

Changed components: user
Changed bases:
Changed systems:

Compiling workspace interfaces
Created /Users/joakimtengstrand/IdeaProjects/example/interfaces/target/interfaces-1.0.jar
Wrote /Users/joakimtengstrand/IdeaProjects/example/interfaces/pom.xml
Installed jar and pom into local repo.

Compiling components/user

Start execution of tests in 1 namespaces:
lein test se.example.user.core-test

lein test se.example.user.core-test

Ran 1 tests containing 1 assertions.
0 failures, 0 errors.

set :last-success in .polylith/time.edn

Execution time: 13.3 seconds

You may notice that the cmd-line base wasn’t compiled and that its test wasn’t executed. The reason is that it hasn’t changed since the last successful test or build, which is something that the plugin detects to speed up testing.

The project file components/user/project.cj was updated when the sync step was executed. When we ran the build command for the first time it updated bases/cmd-line/project.clj. The is because the formatting and/or spacing differs between the project file and the generated file from the sync command. This will only happen the first time you run sync.

If you execute the sync again or any of the other commands that include the sync as a step (build, compile and test) the project file will be left untouched:

$ lein polylith sync

Component interface

If you have read the Polylith documentation you may already have an idea of what a component interface is in the Polylith world, shown in light green here:

When we created the user component, the user’s interface was also created in components/user/src/se/example/user/interface.clj:

(ns se.example.user.interface
  (:require [se.example.user.core :as core]))

;; delegate to the implementations...
(defn add-two [x]
  (core/add-two x))

As you can see, the function in the interface delegates to the actual implementation, which lives in another namespace.

What you often do is to delegate calls to other namespaces in the component to keep the interface clean and tidy. There are no restrictions here, it’s up to you how to arrange the code the way you want and you are free to put all the code in the interface namespace if that is what you think is best. Most often you will have hundreds of lines of code in a component and then it’s best to just delegate.

Interfaces are there for a reason and they solve a number of problems:

  • Guarantee isolation for each component by only exposing the interface.
  • Make components interchangeable within a system so that you can replace them with other components that conform to the same interface.
  • Encourage reuse by sharing the workspace interface for all components within the workspace.
  • Enable navigation between components in the development environment.

A nice side-effect of using components is that you can leave all the functions public in the Clojure code. Normally you want to protect some functions by declaring them as private but with components only the interface is exposed anyway and all other functions are hidden automatically. This can also be handy when you stop at a breakpoint to evaluate a function. If it’s private then you need to use a special syntax to access it but if all functions are public, you don’t have that annoying problem.

Workspace interface

When we created the user component, the workspace interface interfaces/src/se/example/user/interface.clj was also created:

(ns se.example.user.interface)

;; add your function signatures here...
(defn add-two [x])

An important difference between a component interface and a workspace interface is that the latter only contains empty functions.

The lack of dependencies results in a flat roof:

...while the signatures reside at the bottom:

The workspace interfaces are used to guarantee that all the components and bases only depend on functions (or variables and macros) in other component's interface. Each base and component have their own project build file with dependencies to their libraries + the empty workspace interfaces. If you try to access the implementation of a component from another component or base, it will result in compilation errors when executing the compile, test or build commands.

An interface is a specification and a contract, similar to interfaces in the object oriented world, and defines how a base or component can connect to them.

The functions in the workspace interface should be empty but the component interface serves as a bridge between the interface signature and its own implementation:

...
(defn add-two [x]
  (core/add-two x))  ; delegates to the implementation

Note that the signature of the component interface and its corresponding workspace interface must match exactly, otherwise you will get compilation errors when compiling the code.

You are free to arrange your interface namespaces in any way, even to use other names than interface. Instead of one namespace you may want to split them up like: mycomponent/interface, mycomponent/x/interface and mycomponent/y/interface. The recommendation is to have one namespace with the name interface, but if you have good reasons not to, there are no restrictions.

Compose a system

Now we have a cmd-line base and a user component:

$ lein polylith info

interfaces:
  user
components:
  user
bases:
  cmd-line
systems:
  cmd-line
    cmd-line   -> base
environments:
  development
    user       -> component
    cmd-line   -> base

Development

So far we have used the console when interacting with the system and the code but very often you want more than that, like an IDE or an advanced editor.

The development project is just an ordinary Leiningen project and can be edited from any development environment of your choice. Here we will use IntelliJ IDEA with the Cursive plugin as an example.

Open example/environments/development from your preferable text editor or IDE or follow the instructions here if you use Cursive.

How to create a project in Cursive:

  1. Select File > New > Project from Existing Sources
  2. Open the directory example/environments/development
  3. Select Import project from external model + Leiningen
  4. Keep the suggested root directory
  5. Keep suggested projects to import: se.example/development:1.0
  6. Select project SDK
  7. Select project name and location:
    • Project name: example
    • Project file location (e.g.): ~/examples/example/environments/development
  8. Select File > Project Structure… > Modules > Development. Select the interfaces directory and press Test to make it a test directory.

The reason we marked the interfaces directory as a test directory is that we want the IDE to treat its content as source code so that we can get the colouring when we edit it. We don’t want to put it in the src folder because then it can shadow the interfaces of the components (that have the same namespaces and signatures).

Now we have the example project created and we are ready to work with it from the IDE:

You may notice that the cmd-line and the user namespace that live under the se.example top namespace are marked with an arrow. This is because they are symbolic links. This is just a trick that enables us to work with the codebase as if it was a single system similar to a monolith.

With this setup we can start the REPL once and keep on working without restarting it, even though we are working with physically and logically separated components and bases. It also enables refactoring support between components and all the other benefits you get when everything resides in one place, like searching and highlighting unused code.

When working with a Polylith environment in your IDE, you’ll lose the git integration features that you would have with a monolith (the same is true for Microservices). That’s due to the symbolic links confusing the git plugin in your IDE. There are two solutions to this issue: 1) open the individual component and base projects in separate IDE windows, and manage git from there, or 2) use git from the command line or an external GUI tool (such as SourceTree).

An improvement would be to create a Polylith plugin for each of the popular IDEs, which includes VCS support.

Emacs/CIDER Support

If you use Emacs/CIDER you should set vc-follow-symlinks to nil and ensure that find-file-visit-truename is nil. It can be done with per-directory local variables - just create .dir-locals.el file in the workspace root directory with this content:

((nil . ((find-file-visit-truename . nil)
         (vc-follow-symlinks       . nil))))

Also you might want to white-list vc-follow-symlinks as safe file variable by adding (put 'vc-follow-symlinks 'safe-local-variable (lambda (x) (memq x '(t nil ask)))) to your configs.

To see why this is necessary, please check this issue.

Working with the code

Let’s add some code to the user component and the cmd-line base and let cmd-line base use the component.

Change the user interface user/interface.clj in interfaces to:

(ns se.example.user.interface)

;; We "forgot" to add the hello! function.

...and the implementation user/core.clj in src to:

(ns se.example.user.core

(defn hello! [user]
  (println (str "Hello " user "!")))

...and its interface user/interface.clj in src to:

(ns se.example.user.interface
  (:require [se.example.user.core :as core]))

(defn hello! [user]
  (core/hello! user))

...and the user test user/core-test in test to:

(ns se.example.user.core-test
  (:require [clojure.test :refer :all]
            [se.example.user.interface :as user]))

(deftest print-user-test
  (let [output (with-out-str (user/hello! "Bill"))]
    (is (= "Hello Bill!\n"
           output))))

...and the base cmd-line/core.clj in src to:

(ns se.example.cmd-line.core
  (:require [se.example.user.interface :as user])
  (:gen-class))

(defn -main [& args]
  (user/hello! "Victoria"))

...and its test cmd-line/core-test.clj in test to:

(ns se.example.cmd-line.core-test
  (:require [clojure.test :refer :all]
            [se.example.cmd-line.core :as core]))

(deftest hello-world-example-test
  (let [output (with-out-str (core/-main))]
    (is (= "Hello Victoria!\n"
           output))))

Now we can run all the tests from the IDE or the console:

$ cd environments/development
$ lein test

lein test se.example.cmd-line.core-test

lein test se.example.user.core-test

Ran 2 tests containing 2 assertions.
0 failures, 0 errors.

The cmd-line system only contains the base.

We can now either add user to cmd-line with:

$ lein polylith add user cmd-line

...or let the sync command detect the missing component and add it for us:

$ lein polylith sync
Added these definitions to 'interfaces/src/se/example/user/interface.clj':
  (defn hello! [user])
Added component 'user' to system 'cmd-line'.

This will add the user component to the cmd-line system:

It also detected the missing user workspace interface and added it:

(ns se.example.user.interface)

;; We "forgot" to add the hello! function.

(defn hello! [user])

Let’s look again:

$ lein polylith info

interfaces:
  user *
components:
  user *
bases:
  cmd-line
systems:
  cmd-line
    user *     -> component
    cmd-line   -> base
environments:
  development
    user *     -> component
    cmd-line   -> base

Dependencies

In our example, the cmd-line base has a dependency to the user component via the :require statement in cmd-line/core.clj.

We can list all existing dependencies to other components by executing the deps command from the workspace root. By default, the name of the component interfaces are listed:

$ cd ../..
$ lein polylith deps

cmd-line:
  user
user: 

This means that cmd-line depends on user and that user doesn’t depend on other components at all.

We can also list function dependencies with:

$ lein polylith deps +f

cmd-line:
  se.example.user.interface/hello!
user:

For the plugin to work correctly, you need to use the :as syntax in the :require part of the ns statement, for example:

(ns se.example.standalone.core
  (:require [se.example.user.interface :as user])
  …)

The plugin can then detect dependencies to the user in function calls like this:

(user/hello! “Victoria”)

It’s not only possible to expose functions and macros defined by defn and defmacro but also values defined by def. The def statements need to be added to both the workspace interface and all component interfaces in the same way as with defn. It can make the code that accesses the interface look a bit cleaner if you have constants like this:

(def con-string database/connection-string)

...instead of:

(def con-string (database/connection-string))

The use of def statements can be especially useful when declaring Clojure spec definitions. The use of defmacro can be handy if you delegate to other macros like an info macro in a logging library.

Indirect dependencies

The success command is the last step to be executed in the build command if everything went ok. We can execute it to fake a successful build:

$ lein polylith success
set :last-success in .polylith/time.edn
$ lein polylith info

interfaces:
  user
components:
  user
bases:
  cmd-line
systems:
  cmd-line
    user       -> component
    cmd-line   -> base
environments:
  development
    user       -> component
    cmd-line   -> base

Now all * have disappeared.

Now make a small change to user/core.clj for example by executing the touch command:

$ touch components/user/src/se/example/user/core.clj 

Now run the info command again:

$ lein polylith info

interfaces:
  user
components:
  user *
bases:
  cmd-line (*)
systems:
  cmd-line
    user *        -> component
    cmd-line (*)  -> base
environments:
  development
    user *        -> component
    cmd-line (*)  -> base

We can see that user is changed and that cmd-line is marked with an (*). The reason is that it depends on something that has changed but is unchanged itself. This will ensure that its tests are executed when running the test or build command.

Now let’s create the address component:

$ lein polylith create c address

This creates the address component and adds it to the development environment:

Circular dependencies

Change the interface address/interface.clj in interfaces to:

(ns se.example.address.interface)

(defn process! [address])

...and the interface address/interface.clj in src to:

(ns se.example.address.interface
  (:require [se.example.address.core :as core]))

(defn process! [address]
  (core/process! address))

Now let’s see what happens if we depend on user in address/core.clj:

(ns se.example.address.core
  (:require [se.example.user.interface :as user]))

(defn process! [address]
  (user/hello! "Carl"))

...and if we depend back on address in user/core.clj:

(ns se.example.user.core
  (:require [se.example.address.interface :as address]))

(defn hello! [user]
  (address/process! "first street"))

Let’s execute the info command again:

$ lein polylith info
interfaces:
  address *
  user
components:
  address *
  user *   
bases:
  cmd-line (*)
systems:
  cmd-line
    user *        -> component
    cmd-line (*)  -> base
environments:
  development
    address *     -> component  (circular deps: address > user > address)
    user *        -> component  (circular deps: user > address > user)
    cmd-line (*)  -> base       (circular deps: cmd-line > user > address > user)

The circular dependencies are detected so that we can take appropriate actions. If you run the cmd-line/core-test from the IDE, it will complain:

If you run the tests from the console, you will get a similar error:

$ cd environments/development
$ lein test

java.lang.Exception: Cyclic load dependency: [ /se/example/address/interface ]->/se/example/user/core->/se/example/user/interface->/se/example/address/core->[ /se/example/address/interface ]->/se/example/address/core_test

If you don’t have test coverage of these components (although that will of course never happen!) then it can slip through to the production code. To avoid that, the check for circular dependencies is also executed as a step in the build command.

Circular dependencies will not only detect simple cases like a > b > a, but also a > b > c > d > a.

Now restore user/core.clj in src:

(ns se.example.user.core

(defn hello! [user]
  (println (str "Hello " user "!")))

...and change address/core.clj in src to:

(ns se.example.address.core)

(defn process! [address]
  (println "process address:" address))

...and change its test address/core-test.clj in test to:

(ns se.example.address.core-test
  (:require [clojure.test :refer :all]
            [se.example.address.interface :as address]))

(deftest process-address-test
  (let [output (with-out-str (address/process! "first street"))]
    (is (= "process address: first street\n"
           output))))

Now we should be able to run the tests again:

$ lein polylith test

lein test se.example.address.core-test

lein test se.example.cmd-line.core-test

lein test se.example.user.core-test

Ran 3 tests containing 3 assertions.
0 failures, 0 errors.

Context

When you have worked with the Polylith for a while, we believe you will soon find it very pleasant to work with. One reason is that you can start the REPL once without the need to restart it. Another reason is that you have direct access to everything from one place, the development environment.

Another slightly subtle detail is that the components via their interfaces give you context. This might sound like a small thing, but it is very helpful. It not just helps with the design of the system but it also improves the way you work with the code.

Object oriented languages give you context by using objects. Let’s say you work in an object oriented language and that you want to save the object userToBeSaved. If you type userToBeSaved followed by a “.”, the intellisense in the IDE will show you a list of available methods for that object, for example persist:

userToBeSaved.persist(db)

...or use a service:

userService.persist(db, userToBeSaved)

With the Polylith you actually get the same level of support from the IDE. Here you would instead import the user interface and then write:

(user/

Here the IDE will list all available functions in the user interface and one of them would be persist!:

(user/persist! db user-to-be-saved)

Libraries

In the Polylith world, each library uses the same version everywhere within a workspace. So if you have a system called backend and another system called backend2 and both have the user component that uses the library [clj-time “0.14.2”] both backend and backend2 will now use the same version of that library.

The plugin can help you keep all your libraries in sync just by calling the sync command. The sync is also called as a step in the test, compile and build command to make sure you don’t forget it.

Now open project-files/components/user-project.clj from the development environment. This is a symbolic link to example/components/user/project.clj.

Add the [clj-time “0.14.2”] library:

(defproject se.example/user "0.1"
  :description "A user component"
  :dependencies [[org.clojure/clojure "1.9.0"]
                 [se.example/interfaces "1.0"]
                 [clj-time "0.14.2"]]
  :aot :all)

If we look in project-files/systems/cmd-line-project.clj, which is a symbolic link to systems/cmd-line/project.clj, it looks like this:

(defproject se.example/cmd-line "0.1"
  :description "A cmd-line system."
  :dependencies ([org.clojure/clojure "1.9.0"])
  :aot :all
  :main se.example.cmd-line.core)

...and the project.clj file in the root of the development project looks like this:

(defproject se.example/development "1.0"
  :description "The main development environment."
  :dependencies [[org.clojure/clojure "1.9.0"]])

Syncing dependencies

To be able to work with the user component from the development environment, the clj-time library now needs to be added to the development project file. The good thing is that the sync command can help us with that (make sure you are at the workspace root):

$ lein polylith sync
updated: environments/development/project.clj
updated: components/user/project.clj
updated: components/address/project.clj
updated: systems/cmd-line/project.clj

The environments/development/project.clj file now looks like this:

(defproject se.example/development "1.0"
  :description "The main development environment."
  :dependencies [[clj-time "0.14.2"]
                 [org.clojure/clojure "1.9.0"]])

...and systems/cmd-line/project.clj like this:

  (defproject se.example/cmd-line "0.1"
    :description "A cmd-line system."
    :dependencies ([clj-time "0.14.2"]
                   [org.clojure/clojure "1.9.0"])
    :aot :all
    :main se.example.cmd-line.core)

The sync command not only helps you add the missing libraries, it also sorts the dependency list in alphabetical order. You can also have keywords like :extensions in a dependency.

The development environment project file (environments/development/project.clj) is the master for controlling library versions. However, the individual project files in each component or base are where you add new libraries.

The rules of thumb are:

  • add new libraries to component or base project files.
  • update library versions in the development project file.

Let’s update clj-time to 0.14.4 in environments/development/project.clj to:

(defproject se.example/development "1.0"
  :description "The main development environment."
  :dependencies [[clj-time "0.14.4"]
                 [org.clojure/clojure "1.9.0"]])

...and run sync to see what happens:

$ lein polylith sync
  updated: components/user/project.clj
  updated: systems/cmd-line/project.clj

The clj-time version was updated for the user component:

$ cat components/user/project.clj
(defproject se.example/user "0.1"
  :description "A user component"
  :dependencies [[clj-time "0.14.4"]
                 [org.clojure/clojure "1.9.0"]                 
                 [se.example/interfaces "1.0"]]
  :aot :all)

...and added to the cmd-line system:

$ cat systems/cmd-line/project.clj
(defproject se.example/cmd-line "0.1"
  :description "A cmd-line system."
  :dependencies [[clj-time "0.14.4"]
                 [org.clojure/clojure "1.9.0"]]
  :aot :all
  :main se.example.cmd-line.core)

To release often in a controlled way is a good thing and to keep the code and its libraries in sync is also a good thing. The best thing is that the plugin helps you with both!

Testing

This plugin encourages a test-centric approach when working with your code. The introduction of components (that are well isolated and more manageable in size compared to systems) makes testing less complex, faster and more fun.

The test and build commands, that only compiles and runs what has changed since the last successful test or build, makes testing an iterative process that allows you to grow the software in small controlled steps.

Workflow

A natural way of working with the Polylith is this:

  1. Change code and/or add tests.
  2. Run tests from your development environment (e.g. IDE).
  3. Repeat 1 and 2 till you become confident.
  4. Execute lein polylith test
    • if fails => go to 1.
    • if ok => commit to local git repo
  5. Go to 1 or push to global git repo.

This gives a really fast feedback loop from the REPL (1 and 2). The test command (4) will only compile and run tests for affected components since the last time you executed the test command. This will save more and more time the bigger your codebase grows compared to having to compile everything every time. Finally you push your changes (5) and then the same plugin code will be executed on the CI server so if it worked locally you can be quite confident that it will also work on the server (the sync command makes sure the code also passes the last build step).

Test changes

To understand how the plugin figures out what components and bases that have changed we can start by looking at the diff command:

$ lein polylith diff
  interfaces/src/se/example/address/interface.clj
  systems/cmd-line/src/se/example/user/core.clj
  systems/cmd-line/project.clj
  components/address/src/se/example/address/interface.clj
  components/address/src/se/example/address/core.clj
  components/address/readme.md
  components/address/test/se/example/address/core_test.clj
  components/address/project.clj
  components/user/src/se/example/user/core.clj
  components/user/project.clj
  environments/development/src/se/example/address/interface.clj
  environments/development/src/se/example/address/core.clj
  environments/development/src/se/example/user/core.clj
  environments/development/interfaces/se/example/address/interface.clj
  environments/development/docs/address-readme.md
  environments/development/project-files/systems/cmd-line-project.clj
  environments/development/project-files/components/address-project.clj
  environments/development/project-files/components/user-project.clj
  environments/development/test/se/example/address/core_test.clj
  environments/development/project.clj

To calculate this list, the plugin starts with going through all the files in the workspace recursively. Then it takes the value of the :last-success key in .polylith/time.edn and uses that timestamp to filter out all files with a later timestamp.

If the CI variable is set, it’s an indication that the code is executed by the continuous integration server. In that case it takes the SHA1 value from the :last-success key in .polylith/git.edn and then performs a git diff --name-only last-successful-hash current-hash. This approach covers one special case that the local time.edn solution doesn’t cover, namely when a file or directory has been deleted but nothing else was changed.

Commands like build, changes, info and test all start by calling diff internally to calculate changed components and bases to figure out what components and bases to operate on. They all support sending in a bookmark or a timestamp.

The timestamp is the time in milliseconds since 1970, e.g. 1529504135000 instead of 2018-06-20 16:15:35 (the 20th of April 2018, 4:15 pm and 35 seconds). The bookmarks are found in .polylith/time.edn or .polylith/git.edn depending on environment.

So if we figure out when the last successful test or build was:

$ cat .polylith/time.edn
{:last-success 1529504135000}

...then these three commands are equivalent:

$ lein polylith diff
$ lein polylith diff 1529504135000
$ lein polylith diff last-success

To see all changed files since 1970-01-01, simply execute:

$ lein polylith diff 0

To test everything since 1970-01-01, execute:

$ lein polylith test 0

Design

The Polylith makes it easier to compose well-designed systems by providing decoupled building blocks in manageable sizes. It’s just code that directly connects with other code and can therefore easily be used everywhere, like lego bricks.

Let’s take the RealWorld system as an example:

Here we have the rest-api base at the bottom followed by components and libraries on top (not shown here). This design ensures that all dependencies always point in the same direction (up) which is a property it shares with the layered architecture, the hexagonal architecture, and others, but with more flexibility built in.

The design helps you focus on building useful building blocks that are easy to combine like functions without introducing circular dependencies. It also makes it easier to find things and to reason about your systems.

“All problems in computer science can be solved by another level of indirection”
Butler Lampson

The statement means something like “It is easier to move a problem around than it is to solve it” (see indirections). With the Polylith design you get decoupling “for free” via the component’s interfaces which reduces the need for layers to a minimum. It also makes reuse and replacement of code a natural part of your daily work.

Test doubles

If we also add address to the cmd-line system:

$ lein polylith add address cmd-line

Part of the workspace will now look like this:

It’s possible to have more than one component that conforms to an interface. This can be useful if we want to create a separate test system where one of the components is replaced by a fake implementation (see test doubles).

Let’s create a fake implementation of the address interface:

$ lein polylith create c address-fake address
FYI: the component address-fake was created but not added to development because it's interface address was already used by address.

The workspace will now look like this:

The reason why address-fake wasn’t added to the development environment was that it already contained a component that conformed to the address interface, namely the address component. This is not allowed by the plugin because the JVM would then only pick one of them anyway (the one that comes first in the classpath) while the other component would be shadowed and ignored.

If we run the info command again we can see that the address-fake component has an interface (address) with a different name than the component itself (address-fake):

$ lein polylith info
interfaces:
  address *
  user
components:
  address *     
  address-fake *   > address
  user *        
bases:
  cmd-line (*)
systems:
  cmd-line *
    address *     -> component
    user *        -> component
    cmd-line (*)  -> base
environments:
  development
    address *     -> component
    user *        -> component
    cmd-line (*)  -> base

Now create the cmd-line-test system using the cmd-line base and then add user and address-fake to it:

$ lein polylith create s cmd-line-test cmd-line
$ lein polylith add user cmd-line-test
$ lein polylith add address-fake cmd-line-test

Now we have one more system, the cmd-line-test:

In this first release of the Polylith we only support having one development environment and we don’t support adding or removing bases against the development environment either. The reason is that we wanted to release an MVP. A consequence of these decisions is that it’s harder to edit and test components that are not part of the development environment (like address-fake in this case).

What you can do in the meanwhile is to edit address-fake either from its own project or from the cmd-line-test system project.

You can now either add unit tests to address-fake or create system-level tests by manually creating a test/se/example directory structure under example/systems/cmd-line-test, adding your tests there and making sure you execute these from your CI build.

Versioning and branching

The polylith architecture gives you plenty of alternatives to avoid branching. You may need to make extensive changes to a component that is hard to do without being affected or affect other team members or teams. One way of doing that is to add a new component that fulfills the same interface. Now you can work with that component in isolation and keep up with changes in the corresponding workspace interface. When the new component is ready to use you can just switch it in to replace the old one.

When the plugin supports more than one environment, this will be even smother, but in the meantime you can still do it by editing the component from the component project itself, as you would do if it was a Microservice architecture.

Another situation that can arise is when you have different teams working on new functionality that takes a long time to implement and probably will break existing code. In these situations an alternative is to temporarily introduce new versions in the interfaces of the components you want to change. If you for example have the interface com.abc.user.interface containing your function signatures, then you can introduce the namespace com.abc.user.v2.interface containing the changes that otherwise would break your code. When all development has finished, you can either choose to keep that interface or merging it back to com.abc.user.interface and remove com.abc.user.v2.interface.

Mix languages

The Polylith allows us to run multiple languages side by side where each language lives in their own workspace. This will work especially well if we run different languages on top of the same platform, e.g. the JVM (see list of JVM languages).

Let's say we have the languages A, B and C. The first thing to remember is to have different names of the top namespace for each language, so that we don't run into name conflicts. We would end up with top namespaces like: com.mycompany.a, com.mycompany.b and com.mycompany.c. Each language will have their own workspace and they will compile each component to its own library, alternatively compile all components into one big jar like a.jar, b.jar or c.jar.

So if component com.mycompany.a.authentication is used by com.mycompany.b.user, then com.mycompany.b.user will include either a-authentication.jar or a.jar in its library list and delegate to functions in authentication from the user component interface.

With this setup, all components can be used between the languages by first compiling them into libraries. We could also use the Java Native Interface to share code between languages that don't run on top of the JVM, or use something like Neanderthal if we want to integrate with the GPU.

An alternative approach would be to use the GraalVM or similar.

In practice

Here are some tips on how to think when working with the Polylith architecture based on our first two years' experience working with the Polylith in production.

A good principle is to keep your components small. In our experience a size between 100 and 1000 lines of code is a good rule of thumb in Clojure. But as long as you keep the interface cohesive and manage to divide the implementation into well named namespaces, you will be fine even with larger components.

Another advantage of keeping the components small is that it also reduces the build time. The reason is that all the components and bases that depend on a changed component (directly or indirectly) have to execute their tests when building the workspace. If you for example have a huge common component, you run the risk of a majority of your components depending on that component and therefore have to run their tests during build time every time that component changes. If you instead keep the components small, less code will be affected by a change which will in turn reduce the number of executed tests.

To give good names to your systems, bases and components is well spend time. It will affect how you think, reason and communicate about your system. It also makes it easier to find what you are looking for.

It's recommended to wrap API's with components. If you have a REST API to an external system "x" then you should wrap it with the component "x-api" that exposes a nice interface. Then you may have another component "x" where you put all the business logic. To keep them separate is a good thing.

We created a dedicated base, called migration, that was not part of any system but used from the development environment. In the beginning we used it to migrate an old relational database to Datomic by executing code in the migration base from our development environment. It turned out that we could put other type of scripts here too, like cleaning data in the database or to execute "jobs". Basically everything that was executed once and didn't need its own server. We kept all these scripts in the version control systems to keep track of what changes had been done and to steal code from when it was time to create another similar script.

Continuous Integration

As we mentioned before, Polylith uses timestamps to decide if a file has changed since the last success. This works fine on your local machine, however when you push your code to a git repository, all the timestamps will be reset. Once your Continuous Integration (CI) system pulls your code, it will assume everything has changed since the last build.

Don’t worry, the Polylith tool solves this too.

Most popular CI products have an environment variable named CI. The Polylith tool uses that variable to detect if the commands are running on your local machine or on your CI. On your CI, instead of looking at the timestamps, it uses git commit SHA1s to figure out what has changed. Every time the CI runs the success command, it saves the current SHA1 to .polylith/git.edn file. You can cache this file for the next build so the Polylith tool can detect changes since the last build. You can also test this locally if you set the CI environment variable on your computer before running any commands.

You can find a complete CircleCI configuration file that handles Polylith builds in the RealWorld example.

Commands

The goal for this documentation has so far been to give an overall understanding of what problems the plugin tries to solve and how to use it. This section zooms in and explains each command separately. The individual help texts listed here are taken from the built-in help command described here.

Commands

To list all available commands, write:

$ lein polylith help

Or just:

$ lein polylith
  Polylith 0.2.1 (2018-11-19) - https://github.com/tengstrand/lein-polylith

  lein polylith CMD [ARGS]  - where CMD [ARGS] are:

    add C S               Adds a component to a system.
    build N [A] [S]       Builds changed systems and create artifacts.
    changes E P [A]       Lists changed components, bases or systems.
    compile P [A] [S]     Compiles changed components, bases and systems.
    create X N [F]        Creates a component, system or workspace.
    delete c N            Deletes a component.
    deps [A]              Lists dependencies.
    diff P [A] [F]        Lists all changes since a specific point in time.
    help [C]              Show this help or help for specified command.
    info P [A]            Lists interfaces, components, bases, systems and environments.
    prompt                Starts a prompt for current workspace.
    remove C S            Removes a component from a system.
    settings              Shows polylith settings.
    success [B]           Sets last-success or given bookmark.
    sync [F]              Syncs library dependencies, system components and workspace interfaces.
    test P [A] [S]        Executes affected tests in components and bases.

  Examples:
    lein polylith add mycomponent targetsystem
    lein polylith build
    lein polylith build -sync -compile -test -success
    lein polylith build 1523649477000
    lein polylith build mybookmark
    lein polylith changes b
    lein polylith changes c 1523649477000
    lein polylith changes s mybookmark
    lein polylith compile
    lein polylith compile -sync
    lein polylith compile 1523649477000
    lein polylith compile mybookmark
    lein polylith create c mycomponent
    lein polylith create c mycomponent myinterface
    lein polylith create s mysystem mybase
    lein polylith create w myworkspace -
    lein polylith create w myworkspace com.my.company
    lein polylith create w myworkspace com.my.company -git
    lein polylith delete c mycomponent
    lein polylith deps
    lein polylith deps +c
    lein polylith deps +f
    lein polylith deps development
    lein polylith deps mycomponent +f
    lein polylith deps myenvironment +c
    lein polylith diff
    lein polylith diff 1523649477000
    lein polylith diff mybookmark
    lein polylith diff mybookmark +
    lein polylith help
    lein polylith help info
    lein polylith info
    lein polylith info 1523649477000
    lein polylith info mybookmark
    lein polylith prompt
    lein polylith remove mycomponent mysystem
    lein polylith settings
    lein polylith success
    lein polylith success mybookmark
    lein polylith sync
    lein polylith test
    lein polylith test -compile
    lein polylith test 1523649477000
    lein polylith test mybookmark
    lein polylith test -compile -success

As described above, you can read more about a specific command, e.g. prompt, by typing:

$ lein polylith help prompt
  Starts a prompt for current workspace.
  ...

If you plan to execute the code from a script, here comes some useful information. If a command halts because of a problem, it will perform a (System/exit 1) call. The return code can therefore be used to check if something went wrong. Possible problems could for example be that an unsolvable interface declaration or a circular dependency was detected.

add

  Adds a component to a system.

  lein polylith add COMPONENT SYSTEM
     COMPONENT = Component to add.
     SYSTEM    = Add COMPONENT to SYSTEM.

  example:
    lein polylith add mycomponent mysystem

build

  Builds system artifacts.

  The following steps are performed:
    - checks for circular dependencies and quits if found.
    - calculates what components and bases to build based on what has
      changed since the last successful test or build.
    - calls 'sync' and makes sure that all dependencies in project.clj
      files are in sync and that all systems have all components they need.
    - AOT-compiles changed components, bases and systems to check that they compile
      against the workspace interfaces and have all the libraries they need.
    - runs tests for all bases and components that have been affected by the changes.
    - executes build.sh for all changed systems to make sure they have a working
      build script and no missing components or libraries.
    - if the entire build is successful, then execute the success command
      that updates the time for the last successful test or build.

  lein polylith build [ARG] [SKIP]
    ARG = (omitted) -> Since last successful test or build, stored in bookmark
                       :last-success in WS-ROOT/.polylith/time.edn
                       or :last-success in WS-ROOT/.polylith/git.edn
                       if you have CI variable set to something on the machine.
          timestamp -> Since the given timestamp (milliseconds since 1970).
          git-hash  -> Since the given git hash if the CI variable is set.
          bookmark  -> Since the timestamp for the given bookmark in
                       WS-ROOT/.polylith/time.edn or since the git hash
                       for the given bookmark in WS-ROOT/.polylith/git.edn
                       if the CI variable is set.

    SKIP = (omitted)      -> Executes all steps.
           -circular-deps -> Skips checking for circular dependencies step
           -sync          -> Skips dependency sync step
           -compile       -> Skips compilation step
           -test          -> Skips test step
           -success       -> Skips success step

  'lein polylith build 0' can be used to build all files in the workspace
  (or at least changes since 1970-01-01).

  examples:
    lein polylith build
    lein polylith build -compile
    lein polylith build 1523649477000
    lein polylith build mybookmark
    lein polylith build 1523649477000 -compile -test

changes

  Shows what has been changed since a specific point in time.

  lein polylith changes ENTITY [ARG]
    ENTITY = i[nterface] -> Shows changed interfaces
             c[omponent] -> Shows changed components
             b[ase]      -> Shows changed bases
             s[ystem]    -> Shows changed systems

    ARG = (omitted) -> Since last successful test or build, stored in bookmark
                       :last-success in WS-ROOT/.polylith/time.edn. or
                       :last-success in WS-ROOT/.polylith/git.edn if
                       you have the CI variable set to something on the machine.
          timestamp -> Since the given timestamp (milliseconds since 1970).
          git-hash  -> Since the given git hash if the CI variable is set.
          bookmark  -> Since the timestamp for the given bookmark in
                       WS-ROOT/.polylith/time.edn or since the git hash
                       for the given bookmark in WS-ROOT/.polylith/git.edn
                       if the CI variable is set.

  'lein polylith changes 0' can be used to build all files in the workspace
  (or at least changes since 1970-01-01).

  example:
    lein polylith changes i
    lein polylith changes c
    lein polylith changes component
    lein polylith changes b
    lein polylith changes s 1523649477000
    lein polylith changes s mybookmark

compile

  Compiles changes since a specific point in time.

  The following steps are performed:
    - checks for circular dependencies and stops if found.
    - calculates what components and bases to process based on what has
      changed since the last successful test or build.
    - calls 'sync' and makes sure that all dependencies in project.clj
      files are in sync and that all systems have all components they need.
    - AOT compile changed components, bases and systems to check that they compile
      and fulfill workspace interfaces and have all libraries they need.

  lein polylith compile [ARG]
    ARG = (omitted) -> Since last successful test or build, stored in bookmark
                       :last-success in WS-ROOT/.polylith/time.edn
                       or :last-success in WS-ROOT/.polylith/git.edn
                       if you have the CI variable set to something on the machine.
          timestamp -> Since the given timestamp (milliseconds since 1970).
          git-hash  -> Since the given git hash if the CI variable is set.
          bookmark  -> Since the timestamp for the given bookmark in
                       WS-ROOT/.polylith/time.edn or since the git hash
                       for the given bookmark in WS-ROOT/.polylith/git.edn
                       if the CI variable is set.

    SKIP = (omitted)      -> Executes all steps.
           -circular-deps -> Skips checking for circular dependencies step.
           -sync           -> Skips dependency sync step.

  'lein polylith compile 0' can be used to compile all files in the workspace
  (or at least changes since 1970-01-01).

  example:
    lein polylith compile
    lein polylith compile -sync
    lein polylith compile 1523649477000
    lein polylith compile mybookmark
    lein polylith compile mybookmark -sync

create

  Creates a component:

  lein polylith create c[omponent] NAME [INTERFACE]
    NAME = Component name
    INTERFACE = Interface name. Same as component name if omitted.
  --------------------------------------------------------
  Creates a system:

  lein polylith create s[ystem] NAME BASE
    NAME = System name.
    BASE = Base name.
  --------------------------------------------------------
  Creates a workspace:

  lein polylith create w[orkspace] WS NS [FLAG]
    WS = Workspace name.
    NS = Namespace name or '-' to omit it.
         It's recommended and good practice to give a namespace.
    FLAG = (omitted) -> version control the workspace with git.
           -git      -> don't version control the workspace
                        (may occur in any order).

  example:
    lein polylith create c mycomponent
    lein polylith create c mycomponent myinterface
    lein polylith create component mycomponent
    lein polylith create component mycomponent myinterface
    lein polylith create s mysystem mybase
    lein polylith create system mysystem mybase
    lein polylith create w myworkspace -
    lein polylith create w myworkspace com.my.company
    lein polylith create w myworkspace com.my.company -git
    lein polylith create workspace myworkspace com.my.company

delete

  Deletes a component and its interface if no other components use it.

  lein polylith delete c NAME
    NAME = component to delete

  example:
    lein polylith delete c mycomponent

deps

  Lists dependencies used in at least one environment or system.

  lein polylith deps [NAME] [FLAG]
    NAME = (omitted) -> Lists all dependencies.
           else      -> Filters dependencies by given system, environment
                        base or component name.
    FLAG = (omitted)    -> Lists interface dependencies.
           +c[omponent] -> Lists component dependencies.
           +f[unction]  -> Lists function dependencies.

  To work correctly, :require with a corresponding :as alias
  must be used to specify dependencies in each namespace, e.g.:
    (ns my.namespace.core
      (:require [my.namespace.user.interface :as user]
                [my.namespace.email.interface :as email]))

  example:
    lein polylith deps
    lein polylith deps +c
    lein polylith deps +component
    lein polylith deps +f
    lein polylith deps +function
    lein polylith deps development
    lein polylith deps mycomponent +f
    lein polylith deps myenvironment +c

diff

  Lists all files and directories that have changed in the workspace
  since a specific point in time. Deleted files and directories
  are not detected locally (when the CI environment variable is not set).

  lein polylith diff [ARG] [FLAG]
    ARG = (omitted) -> Since last successful test or build, stored in bookmark
                       :last-success in WS-ROOT/.polylith/time.edn
                       or :last-success in WS-ROOT/.polylith/git.edn
                       if you have the CI variable set to something on the machine.
          timestamp -> Since the given timestamp (milliseconds since 1970).
          git-hash  -> Since the given git hash if the CI variable is set.
          bookmark  -> Since the timestamp for the given bookmark in
                       WS-ROOT/.polylith/time.edn or since the git hash
                       for the given bookmark in WS-ROOT/.polylith/git.edn
                       if the CI variable is set.

    FLAG = +        -> Show time information.
                       (the + sign may occur in any order in the argument list).

  'lein polylith diff 0' can be used to list all files in the workspace
  (or at least changes since 1970-01-01).

  example:
    lein polylith diff
    lein polylith diff +
    lein polylith diff + 1523649477000
    lein polylith diff 1523649477000
    lein polylith diff 1523649477000 +
    lein polylith diff mybookmark

info

  Shows the content of a Polylith workspace and its changes since
  the last successful test or build or a given point in time.

  Each row is followed by an * if something has changed.
  Each row is followed by a (*) if nothing has changed but it
  depends on one or more components that have changed.

  lein polylith info [ARG]
    ARG = (omitted) -> Since last successful test or build, stored in bookmark
                       :last-success in WS-ROOT/.polylith/time.edn
                       or :last-success in WS-ROOT/.polylith/git.edn
                       if you have the CI variable set to something on the machine.
          timestamp -> Since the given timestamp (milliseconds since 1970).
          git-hash  -> Since the given git hash if the CI variable is set.
          bookmark  -> Since the timestamp for the given bookmark in
                       WS-ROOT/.polylith/time.edn or since the git hash
                       for the given bookmark in WS-ROOT/.polylith/git.edn
                       if the CI variable set.

  example:
    lein polylith info
    lein polylith info 1523649477000
    lein polylith info mybookmark

prompt

  Starts a prompt for current workspace.

  Allows faster execution of all the commands from any directory.

  Type 'exit' or 'quit' to exit the prompt.

  examples:
    lein polylith prompt

remove

  Removes a component from a system.

  lein polylith remove COMPONENT SYSTEM
    COMPONENT = Name of the component
    SYSTEM    = Name of the system

  example:
    lein polylith remove mycomponent mysystem

settings

  Shows workspace settings with various information.

  examples:
    lein polylith settings

success

  Depending on whether the environment variable CI is set or not:

    If CI is not set (when executed from the local development environment):
      Sets the BOOKMARK in WS-ROOT/.polylith/time.edn to current time
      in milliseconds.

    If CI is set (when executed from the continuous integration server):
      Sets the BOOKMARK in WS-ROOT/.polylith/git.edn to current Git SHA1.

  lein polylith success [BOOKMARK]
    BOOKMARK = (omitted) -> last-success
               name of the bookmark

  examples:
    lein polylith success
    lein polylith success mybookmark

sync

  These steps are performed:

  1. Adds missing libraries to the development environment.
     The way it does that is to first check which components and bases
     are part of the development environment. Then it goes through
     those components and bases and collects a list of all their dependencies
     from each project.clj file. That list is compared with the dependencies
     in environments/development/project.clj and missing libraries are added.

  2. Makes sure that the library versions for all components
     and bases are in sync with the library versions in
     environments/development/project.clj.

  3. Makes sure that each system has a library list that reflects
     the sum of all libraries of its components and bases.

  4. Adds missing components to systems if possible/needed.
     This can be performed only if each interface belongs to exactly
     one component, otherwise an error message is displayed.

  5. Adds missing def/defn/defmacro definitions to workspace interfaces.
     All namespaces for each interface under the interfaces directory are
     parsed and all def/defn/defmacro definitions are collected into a set.
     Then all def/defn/defmacro definitions, that exist in the corresponding
     component interface(s) but not in the workspace interface, are added.
     If a new arity of a function or macro has been added or changed,
     then an error message is shown, informing you to update it manually.

  examples:
    lein polylith sync

test

  Executes component and base tests.

  The following steps are performed:
    - checks for circular dependencies and stops if found.
    - calculates what components and bases to process based on what has
      changed since the last successful test or build.
    - calls 'sync' and makes sure that all dependencies in project.clj
      files are in sync and that all systems have all components they need.
    - AOT compile changed components, bases and systems to check that they compile
      and fulfill workspace interfaces and have all libraries they need.
    - runs tests for all bases and components that have been affected by the changes.
    - if all the tests are successful, then execute the success command
      that updates the time for the last successful test or build.

  lein polylith test [ARG] [SKIP]
    ARG = (omitted) -> Since last successful test or build, stored in bookmark
                       :last-success in WS-ROOT/.polylith/time.edn
                       or :last-success in WS-ROOT/.polylith/git.edn
                       if you have the CI variable set to something on the machine.
          timestamp -> Since the given timestamp (milliseconds since 1970).
          git-hash  -> Since the given git hash if the CI variable is set.
          bookmark  -> Since the timestamp for the given bookmark in
                       WS-ROOT/.polylith/time.edn or since the git hash
                       for the given bookmark in WS-ROOT/.polylith/git.edn
                       if CI variable set.

    SKIP = (omitted)      -> Executes all steps.
           -circular-deps -> Skips checking for circular dependencies step.
           -sync          -> Skips dependency sync step.
           -compile       -> Skips compilation step.
           -success       -> Skips success step

  'lein polylith test 0' can be used to test all files in the workspace
  (or at least changes since 1970-01-01).

  examples:
    lein polylith test
    lein polylith test -compile
    lein polylith test 1523649477000
    lein polylith test mybookmark
    lein polylith test mybookmark -compile
    lein polylith test -compile -success
    lein polylith test -compile test

What’s next

Here are some of the planned features for the plugin:

  • Support for more than one environment.
  • Support for adding and removing components and bases to/from environments.
  • Support for adding and removing bases to/from systems.
  • Support for renaming interfaces, components, bases and systems.
  • Support for creating a base only used by an environment.
  • Support for executing tests in other environments than development.
  • Introduce the 'doc' command that produces interactive HTML documentation where you can visualise your systems, bases, components, interfaces and how everything fits together:

System

Thanks

A big thanks goes to James Trunk for your support during the recent years and for always having time to listen to my sometimes crazy ideas. Thanks also for your excellent work with the Polylith presentation and documentation and the amazing level of quality that you put into everything you do.

Another big thanks goes to Furkan Bayraktar not just because you are a nice person but also because you are the fastest and most talented developer I have worked with during my 20+ years in this industry. Thanks for your contribution to the Polylith plugin and the RealWorld example.

Thanks to Kim Kinnear, the creator of zprint whose library helps us to update the dependencies in project.clj files.

I also want to thank my son Mattias Tengstrand for his excellent work with the Polylith logo.

My last thanks goes to Rachel for the help with the English texts.

Joakim Tengstrand

Contact

Feel free to contact me:

Twitter: @jtengstrand
Email: joakim[dot]tengstrand[at]gmail[dot]com

You can also get in touch with us in the Polylith forum.

License

Distributed under the Eclipse Public License, the same as Clojure.