Skip to content

Latest commit

 

History

History
590 lines (435 loc) · 26.9 KB

README.md

File metadata and controls

590 lines (435 loc) · 26.9 KB

Welcome to Ecstasy!

This is the public repository for the Ecstasy language (xtclang.org) and the Ecstasy virtual machine (XVM) project.

What is Ecstasy?

Ecstasy

Ecstasy is a new, general-purpose, programming language, designed for modern cloud architectures, and explicitly for the secure, serverless cloud. Actually, to be completely honest, it's the most amazing programming language ever. No, really, it's that awesome.

The Ecstasy project includes a development kit (XDK) that is produced out of this repository, a programming language specification, a core set of runtime modules (libraries), a portable, type-safe, and verifiable Intermediate Representation (IR), a proof-of-concept runtime (with an adaptive LLVM-based optimizing compiler in development), and a tool-chain with both Java and Ecstasy implementations being actively developed.

The Ecstasy language supports first class modules, including versioning and conditionality; first class functions, including currying and partial application; type-safe object orientation, including support for auto-narrowing types, type-safe covariance, mixins, and duck-typed interfaces; complete type inference; first class immutable types; first class asynchronous services, including both automatic async/await-style and promises-based (@Future) programming models; and first class software containers, including resource injection and transitively-closed, immutable type systems. And much, much more.

Read more at https://xtclang.blogspot.com/

Follow us on Twitter @xtclang

Find out more about how you can contribute to Ecstasy.

And please respect our code of conduct and each other.

Binary Installation

For macOS and Linux:

  1. If you do not already have the brew command available, install Homebrew

  2. Add a "tap" to access the XDK CI builds, and install the latest XDK CI build:

brew tap xtclang/xvm && brew install xdk-latest
  1. To upgrade to the latest XDK CI build at any time:
brew update && brew upgrade xdk-latest

For Windows:

Manual local build for any computer (for advanced users):

  • Install Java (version 17 or later) and Gradle

  • Use git to obtain the XDK:

  git clone https://github.com/xtclang/xvm.git
  • cd into the git repo (the directory will contain these files) and execute the Gradle build:
  ./gradlew build

Development

Recommended Git workflow

A note about this section: this workflow is supported by pretty much every common GUI in any common IDE, in one way or another. But in the interest of not having to document several instances with slightly different naming convention, or deliver a confusing tutorial, this section only describes the exact bare bones command line git commands that can be used to implement our workflow, which is also a common developer preference. All known IDEs just wrap these commands in one way or another.

Make sure "pull.rebase" is set to "true" in your git configuration

In order to maintain linear git history, and at any cost avoid merges being created and persisted in the code base, please make sure that your git configuration will run "pull" with "rebase" as its default option. Preferably globally, but at least for the XVM repository.

git config --get pull.rebase

Output should be "true".

If it's not, execute

git config --global pull.rebase true

or from a directory inside the repository:

git config --local pull.rebase true

The latter will only change the pull semantics for the repository itself, and the config may or may not be rewritten by future updates.

Always work in a branch. Do not work directly in master

XTC will very soon switch to only allowing putting code onto the master branch through a pull request in a sub branch.

In order to minimize git merges, and to keep master clean, with a minimum of complexity, the recommended workflow for submitting a pull request is as follows:

1) Create a new branch for your change, and connect it to the upstream:
git checkout -B decriptive-branch-name
git push --set-upstream origin descriptive-branch-name
2) Perform your changes, and commit them. We currently do not have any syntax requirements

on commit descriptions, but it's a good idea to describe the purpose of the commit.

git commit -m "Descriptive commit message, including a github issue reference, if one exists"
3) Push your changes to the upstream and create a pull request, when you are ready for review
git push
Resolving conflicts, and keeping your branch up to date with master

Whenever you need to, and this is encouraged, you should rebase your local branch, so that your changes get ripped out and re-transplanted on top of everything that has been pushed to master, during the time you have been working on the branch.

Before you submit a pull request, you need to rebase it against master. We will gradually add build pipeline logic for helping out with this, and other things, but it's still strongly recommended that you understand the process.

To do a rebase, which has the effect that your branch will contain all of master, with your commits moved to the end of history, execute the following commands:

git fetch 
git rebase origin/master

The fetch command ensures that the global state of the world, whose local copy is stored in the ".git" directory of the repository, gets updated. Remember that git allows you to work completely offline, should you chose to do so, after you have cloned a repository. This means that, in order to get the latest changes from the rest of the world, and make sure you are working in an up-to-date environment, you need to fetch that state from the upstream.

If there are any conflicts, the rebase command above will halt and report conflict. Should this be the case, change your code to resolve the conflicts, and verify that it builds clean again. After it does, add the resolved commit and tell git to continue with the rebase:

git add .
git rebase --continue

If you get entangled, you can always restart the rebase by reverting to the state where you started:

git rebase --abort

After rebasing, it's a good idea to execute "git status", to see if there are heads from both master and your local branch. Should this be the case, you need to resolve the rebase commit order by force pushing the rebased version of you local branch before creating the pull request for review:

git status
git push -f # if needed
Do not be afraid to mess around in your local branch

You should feel free to commit and push as much as you want in your local branch, if your workflow so requires. However, before submitting the finished branch as a pull request, please do an interactive rebase and collapse any broken commits that don't build, or any small commits that just fix typos and things of a similar nature.

  • It is considered bad form to submit a pull request where there are unnecessary or intermediate commits, with vague descriptions.

  • It is considered bad form to submit a pull request where there are commits, which do not build and test cleanly. This is important, because it enables things like automating git bisection to narrow down commits that may have introduced bugs, and it has various other benefits. The ideal state for master, should be that you can check it out at any change in its commit history, and that it will build and test clean on that head.

Most pull requests are small in scope, and should contain only one commit, when they are put up for review. If there are distinct unrelated commits, that both contribute to solving the issue you are working on, it's naturally fine to not squash those together, as it's easier to read and shows clear separation of concerns.

If you need to get rid of temporary, broken, or non-buildable commits in your branch, do an interactive rebase before you submit it for review. You can execute:

git rebase -i HEAD~n

to do this, where n is the number of commits you are interested in modifying.

  • According to the git philosophy, branches should be thought of as private, plentiful and ephemeral. They should be created at the drop of a hat, and the branch should be automatically or manually deleted after its changes have been merged to master. A branch should never be reused.

The described approach is a good one to follow, since it moves any complicated source control issues completely to the author of a branch, without affecting master, and potentially breaking things for other developers. Having to modify the master branch, due to unintended merge state or changes having made their way into it, is a massively more complex problem than handling all conflicts and similar issues in the private local branches.

Status

Version 0.4. That's way before version 1.0. In other words, Ecstasy is about as mature as Windows 3.1 was.

Warning: The Ecstasy project is not yet certified for production use. This is a large and extremely ambitious project, and it may yet be several years before this project is certified for production use.

Our goal is to always honestly communicate the status of this project, and to respect those who contribute and use the project by facilitating a healthy, active community, and a useful, high-quality project. Whether you are looking to learn about language design and development, compiler technology, or the applicability of language design to the serverless cloud, we have a place for you here. Feel free to lurk. Feel free to fork the project. Feel free to contribute.

We only "get one chance to make a good first impression", and we are determined not to waste it. We will not ask developers to waste their time attempting to use an incomplete project, so if you are here for a work reason, it's probably still a bit too early for you to be using this for your day job. On the other hand, if you are here to learn and/or contribute, then you are right on time! Our doors are open.

License

The license for source code is Apache 2.0, unless explicitly noted. We chose Apache 2.0 for its compatibility with almost every reasonable use, and its compatibility with almost every license, reasonable or otherwise.

The license for documentation (including any the embedded markdown API documentation and/or derivative forms thereof) is Creative Commons CC-BY-4.0, unless explicitly noted.

To help ensure clean IP (which will help us keep this project free and open source), pull requests for source code changes require a signed contributor agreement to be submitted in advance. We use the Apache contributor model agreements (modified to identify this specific project), which can be found in the license directory. Contributors are required to sign and submit an Ecstasy Project Individual Contributor License Agreement (ICLA), or be a named employee on an Ecstasy Project Corporate Contributor License Agreement (CCLA), both derived directly from the Apache agreements of the same name. (Sorry for the paper-work! We hate it, too!)

The Ecstasy name is a trademark owned and administered by The Ecstasy Project. Unlicensed use of the Ecstasy trademark is prohibited and will constitute infringement.

All content of the project not covered by the above terms is probably an accident that we need to be made aware of, and remains (c) The Ecstasy Project, all rights reserved.

Layout

The project is organized as a number of subprojects, with the important ones to know about being:

  • The Ecstasy core library is in the xvm/lib_ecstasy directory, and is conceptually like stdlib for C, or rt.jar for Java. When the XDK is built, the resulting module is located at xdk/lib/ecstasy.xtc. This module contains portions of the Ecstasy tool chain, including the lexer and parser. (Ecstasy source files use an .x extension, and are compiled into a single module file with an .xtc extension.)

  • The Java tool chain (including an Ecstasy compiler and interpreter) is located in the xvm/javatools directory. When the XDK is built, the resulting .jar file is located at xdk/javatools/javatools.jar.

  • There is an Ecstasy library in xvm/javatools_bridge that is used by the Java interpreter to boot-strap the runtime. When the XDK is built, the resulting module is located at xdk/javatools/javatools_bridge.xtc.

  • The wiki documentation is online. There is an introduction to Ecstasy that is being written for new users. The wiki source code will (eventually) be found in the xvm/wiki project directory, and (as a distributable) in the xdk/doc directory of the built XDK.

  • Various other directories will have a README.md file that explains their purpose.

To download the entire project from the terminal, you will need git installed. From the terminal, go to the directory where you want to create a local copy of the Ecstasy project, and:

git clone https://github.com/xtclang/xvm.git

(There is excellent online documentation for git at git-scm.com.)

To build the entire project, you need to have gradle, or you use the included Gradle Wrapper from within the xvm directory, which is the recommended method:

./gradlew build

Or on Windows:

C:\> gradlew.bat build

Note that Windows may require the JAVA_TOOLS_OPTIONS environment variable to be set to -Dfile.encoding=UTF-8 in the Environment Variables window that can be accessed from Control Panel. This allows the Java compiler to automatically handle UTF-8 encoded files, and several of the Java source files used in the Ecstasy toolchain contain UTF-8 characters. Also, to change the default encoding used in Windows, go to the "Administrative" tab of the "Region" settings Window (also accessed from Control Panel), click the "Change system locale..." button and check the box labeled "Beta: Use UTF-8 for worldwide language support".

Instructions for getting started can be found in our Contributing to Ecstasy document.

Cleaning the build

You can clean everything in a build by running

./gradlew clean  

However, note that if you restart the build, a lot of intermediary outputs will be cached. The XDK and XTC plugin use the Gradle build system intrinsics to only compile what has provably mutated its outputs and inputs. This should be stable, and is by design. You should really never have to clean your build to test any incremental change. This is even true if you modify the plugin implementation in the XDK repo, as we are using included builds everywhere we should. The end goal is that any change will only require rebuilding, and that rebuild should only build exactly what is necessary.

Should you, for any reason, need to clear the caches, and really start fresh, you can run the script

./bin/purge-all-build-state.sh

Or do the equivalent actions manually:

  1. Close any open XTC projects in your IDEs, to avoid restarting them with a large state change under the hood. Optionally, also close your IDE processes.
  2. Kill all Gradle daemons.
  3. Delete the $GRADLE_USER_HOME/cache and $GRADLE_USER_HOME/daemons directories. NOTE: this invalidates caches for all Gradle builds on your current system, and rebuilds a new Gradle version.
  4. Run git clean -xfd in your build root. Note that this may also delete any IDE configuration that resides in your build. You may want to preserve e.g. the .idea directory, and then you can do git clean -xfd -e .idea or perform a dry run git clean -xfdn, to see what will be deleted. Note that if you are at this level of purging stuff, it's likely a bad idea to hang on to your IDE state anyway.

Debugging the build

The build should be debuggable through any IDE, for example IntelliJ, using its Gradle tooling API hook. You can run any task in the project in debug mode from within the IDE, with breakpoints in the build scripts and/or the underlying non-XTC code, for example in Javatools, to debug the compiler, runner or disassembler.

Augmenting the build output

XTC follow Gradle best practise, and you can run the build, or any task therein, with the standard verbosity flags. For example, to run the build with more verbose output, use:

./gradlew build --info --stacktrace

The build also supports Gradle build scans, which can be generated with:

./gradlew build --scan --stacktrace

Note that build scans are published to the Gradle online build scan repository (as configured through the gradle-enterprise settings plugin.), so make sure that you aren't logging any secrets, and avoid publishing build scans in "--debug" mode, as it may be a potential security hazard.

You can also combine the above flags, and use all other standard Gradle flags, like --stacktrace, and so on.

Tasks

To see the list of available tasks for the XDK build, use:

./gradlew tasks

Versioning and Publishing XDK artifacts

  • Use publishLocalto publish an XDK build to the local Maven repository and a build specific repository directory.
  • Use publishRemoteto publish and XDK build to the xtclang organization package repo on GitHub (a GitHub token with permissions is required).
  • Use publish to run both of the above tasks.

Note: At the moment some publish tasks may have some raciness in execution, due to Gradle issues. Should you get some kind of error during the publishing task, it may be a good idea to clean, and then rerun that task with the Gradle flag --no-parallel.

The group and version of the current XDK build and the XTC Plugin are currently defined in the properties file "version.properties". Here, we define the version of the current XDK and XTC Plugin, as well as their group. The default behavior is to only define the XDK, since at this point, the Plugin, while decoupled, tracks and maps to the XDK version pretty much 1-1. This can be taken apart with different semantic versioning, should we need to. Nothing is assuming the plugin has the same version or group as the XDK. It's just convenient for time being.

The file gradle/libs.versions.toml contains all internal and external by-artifact version dependencies to the XDK project. If you need to add a new plugin, library, or bundle, always define its details in this version catalog, and nowhere else. The XDK build logic, will dynamically plugin in values for the XDK and XTC Plugin artifacts that will be used only as references outside this file.

TODO: In the future we will also support tagging and publishing releases on GitHub, using JReleaser or a similar framework.

Typically, the project version of anything that is unreleased should be "x.y.z-SNAPSHOT", and the first action after tagging and uploading a release of the XDK, is usually changing the release version in "VERSION" in the xvm repository root, and (if the plugin is versioned separately, optionally in "plugin/VERSION") both by incrementing the micro version, and by adding a SNAPSHOT suffix. You will likely find yourself working in branches that use SNAPSHOT versions until they have made it into a release train. The CI/CD pipeline can very likely handle this automatically.

Bleeding Edge for Developers

If you would like to contribute to the Ecstasy Project, it might be an idea to use the very latest version by invoking:

./gradlew installLocalDist

This copies the build from the xvm directory into the brew cellar, or other local installation, that is deduced from the location of the xec launcher on the system PATH.

Note: this would be done after installing the XDK via brew, or through any other installation utility, depending on your platform. This will overwrite several libraries and files in any local installation.

For more information about the XTC DSL, please see the README.md file in the "plugin" project.

Releasing and Publishing

This is mostly relevant to the XDK development team with release management privileges. A version of the workflow for adding XTC releases is described here.

We plan to move to an automatic release model in the very near future, utilizing JRelease (and JPackage to generate our binary launchers). As an XTC/XDK developer, you do not have to understand all the details of the release model. The somewhat incomplete and rather manual release mode is current described here for completeness. It will soon be replaced with something familiar.

XDK Platform Releases

  1. Take the current version of master and create a release branch.
  2. Set the VERSION in the release branch project root to reflect the version of the release. Typically an ongoing development branch will be a "-SNAPSHOT" suffixed release, but not an official XTC release, which just has a group:name:version number
  3. Build, tag and add the release using the GitHub release plugin.

XDK Platform Publishing

We have verified credentials for artifacts with the group "org.xtclang" at the best known community portals, and will start publishing there, as soon as we have an industrial strength release model completed.

The current semi-manual process looks like this:

  1. ./gradlew publish to build the artifacts and verify they work. This will publish the artifacts to a local repositories and the XTC GitHub org repository.
  2. To publish the plugin to Gradle Plugin Portal: ./gradlew :plugin:publishPlugins (publish the plugin to gradlePortal)
  3. To publish the XDK distro to Maven Central: (... TODO ... )

You can already refer to the XDK and the XTC Plugin as external artifacts for your favourite XTC project, either by mnaually setting up a link to the XTC Org GitHub Maven Repository like this:

repositories {
   maven {
     url = https://maven.pkg.github.com/xtclang/xvm
     credentials {
        username = <your github user name>
        token = <a personal access token with read:package privileges on GitHub Maven Packages>
   }
}

or by simply publishing the XDK and XDK Plugin to your mavenLocal repository, and adding that to the configuration of your XTC project, if it's not there already:

repositories {
   mavenLocal()
}

Questions?

To submit a contributor agreement, sign up for very hard work, fork over a giant pile of cash, or in case of emergency: "info at xtclang dot org", but please understand if we cannot respond to every e-mail. Thank you.

Appendix: Gradle fundamentals

We have tried very hard to create an easy-to-use build system based on industry standards and expected behavior. These days, most software is based on the Maven/Gradle model, which provides repositories of semantically versioned artifacts, cached incremental builds and mature support for containerization.

The principle of least astonishment permeates the philosophy behind the entire build system. This means that a modern developer, should be immediately familiar with how to build and run the XDK project, i.e. clone it from GitHub and execute "./gradlew build". It should also import complaint free, and with dependency chains understood by any IDE that has support for Gradle projects. "It should just work", out of the box, and should look familiar to any developer with basic experience as a Gradle user. Nothing should require more than a single command like to build or execute the system or anything built on top of it.

Implementing language support for an alien language on top of Gradle, however, is a fairly complex undertaking, and requires deeper knowledge of the Gradle architecture. It is our firm belief, though, that the user should not have to drill down to these levels, unless he/she specifically wants to. As it is, any open source developer today still needs to grasp some basic fundamentals about artifacts and the Gradle build system. This is not just our assumption; it is actually industry-wide.

We believe the following concepts are necessary to understand, in order to work with XDK projects or the XDK. None of them are at all specific to XTC:

  • The concept of "gradlew" and "mvnw" (or "gradlew.bat" and "mvnw.bat" on Windows) wrappers, and why it should ALWAYS be used instead of a "gradle" binary on the local system, for any repository that ships it with its build.
  • The concept of a versioned Maven artifact, and that its descriptor "group:artifactId:version" is its "global address", no matter how it is resolved on the lower abstraction layer.
  • The concept of release vs snapshot artifact versions in the Maven model.
  • The concept of local (mostly mavenLocal()) and remote artifact repositories, and how they are used by a maven build.
  • The concept of the Maven/Gradle build lifecycle, its fundamental tasks, and how they depend on each other ("clean", "assemble", "build" and "check").
  • The concept of the Gradle/Maven cache, build daemons, and why "clean" is not what you think
    of as "clean" in a C++ Makefile and why is it often better not to use it, in a cached, incrementally built Gradle project.
  • The concept of Maven/Gradle source sets, like "main", "resources" and "test".
  • The concept of a Gradle build scan, and understanding how to inspect it and how to use it to spot build issues.
  • The standard flags that can be used to control Gradle debug log levels, --info, -q, --stacktrace and so on.
  • The concept of goal of self-contained software, which specifies its complete dependencies as part of its source controlled configuration.
    1. On the Maven model level, this means semantically versioned Maven artifacts.
    2. On the software build and execution level, this also means specific versions of external pieces of software, for example Java, NodeJS or Yarn. This also means that we CAN and SHOULD always be able to containerize for development purposes.

Today, it is pretty safe to assume that most open source developers who has worked on any Gradle or Maven based project has at least the most important parts of the above knowledge. We have spent significant architectural effort to ensure that an adopter who wants to become an XTC or XDK user or developer does not need to acquire any knowledge that is more domain specific than concepts listed above. None of these concepts are specific to the XTC platform, but should be familiar to most software developers who have worked on projects with Maven style build systems.

We will also work on IDE Language support as soon as we have enough cycles to do so, which should make getting up to speed with XTC and even less complicated process.