Welcome! This tutorial was designed specifically for developers, who are completely new to MPS and prefer a guided tour through the MPS landscape. You will walk the beaten path one step at a time, following clear marks that show you where to go next. The information is structured so that you progress from simpler concepts to the more involved ones and at the end of the journey you'll understand MPS and will be able to use it effectively on your projects.
We admit that there are easier tasks in the world than learning MPS. Language design is a complex domain and projectional editing needs some getting used to. It helps a lot if you have prior experience with DSLs or language workbenches, but it is not necessary for understanding this tutorial. Just expect more bumps on the road. There certainly will be a lot of them - partly also because MPS has its own problems and bugs. So you are in for a tough experience that may take quite a big amount of your time.
Bear in mind that many have walked through here before and succeeded. If you persist you'll be rewarded - you'll become a language designer who can prepare custom-tailored DSLs, integrate them into projects and develop high-value software with them.
So, take on good shoes, lace them tight and start going ...
- Ten steps to mastery
- Step 1 - Install MPS
- Step 2 - MPS from the programmer's perspective
- Step 3 - MPS from the language designer's perspective
- Step 4 - Creating your first language
- Step 5 - Why MPS
- Step 6 - Projectional editing tips and tricks
- Step 7 - Shapes tutorial
- Step 8 - Calculator tutorial
- Step 9 - Other sample projects
- Step 10 - Generator tutorial
- Reviewing the progress
- Moving up the ladder
- Documentation sources
Ten steps to mastery
Step 1 - Install MPS
Plain reading about MPS will never teach you as much as trying it for real yourself. You might have done this already, but if not, grab MPS from the download site and install it (You may ignore the IntelliJ IDEA MPS plugin for now).
With MPS installed you will be able to continue on the Fast Track.
Step 2 - MPS from the programmer's perspective
Raw time estimate - 45 minutes
Before you continue, you should probably get to know the most distinct feature of MPS - the projectional editor. Check out a video that explains all the differences between traditional textual editors and the MPS projectional editor and lists the numerous benefits of this approach.
Since we programmers like trying things, how about taking the MPS editor for a simple test ride first? You may get the first impressions instantly, if you start MPS and open the Robot Kaja sample project. This project defines an artificial language to command a virtual robot. The robot can move within a rectangular area, drop and collect marks and check the surrounding cells for walls and marks. You write Scripts using the built-in commands as well as new commands that you create on the fly. When you run the script a simulator written in Java is started to visualize the execution of your program.
To open the project, start your fresh MPS, click on the Open Sample Project button and pick robot_Kaja. The sample projects have been installed into your HOME/MPSSamples folder.
We also shot an instructional screen-cast to help you start. Check it out.
Step 3 - MPS from the language designer's perspective
Raw time estimate - 40 minutes
Now you must be curious about how the robot sample language has been created. Please, open again the Robot Kaja sample project and play our explanatory screen-cast showing you round the language definition fundamentals. As you go through the video, pause frequently and look around the sample project yourself. This way you will understand a typical language structure and see the relationships between its individual parts.
At this point you should understand that language definition consists of several aspects. Some of these are mandatory, some are optional:
- Structure - Defines the kinds of nodes (called Concepts) that may be used in user models. Each node in the program (model) refers to its concept. Concepts specify, which properties, children and references nodes may have. Concepts can extend other Concepts and implement ConceptInterfaces.
- Constraints - Restricts the relationships between nodes as well as the allowed values for properties beyond the rules defined in Structure. Constraints typicaly define:
- the target scope for references (a collection of allowed nodes a reference can point to)
- situations, in which a node can be a child/parent/ancestor of another node
- allowed values for properties
- property accessor methods (getters and setters)
- Behavior - Just like classes in OOP hold methods, Concepts may define methods and static methods that can be invoked on nodes in a polymorphic way. Nodes thus carry behaviour alongside their properties and relationships.
- Editor - Instead of defining a parser that would translate code from an editable form (i.e. text) into the tree-like structure that a computer could manipulate, MPS offers the concept of projectional editor, which let's the user edit the AST directly. The Editor aspect enables language designers to create a UI for editing their concept concepts.
- Actions - Since the MPS editor manipulates the underlying AST directly, some editing actions, such as typing certain characters or strings, translate into non-trivial AST changes. For example, typing the plus sign next to a number should insert a PlusExpression, hook its left child to the original number and set the cursor to the empty cell for the right child.
The Actions aspect holds definitions of such customised AST transformations.
- Intentions - All modern IDEs assist developers with instant code manipulating action available under a handy key-shortcut (Alt + Enter in MPS). Language authors can define such little code transformations for their languages in the Intentions aspect.
- Typesystem - Language that need to type-check their code need to provide type-system rules. The MPS type-system engine will evaluate the rules on-the-fly, calculate types for nodes and report
errors, wherever the calculated type differs from the expectations.
So called checking rules may additionally be defined to verify non-typesystem assertions about the model.
- Dataflow - The ability to understand the flow of values and the flow of control through language constructs helps languages report issues such as unreachable code or potential null-pointer error. Language designer can leverage the Dataflow aspect to define the flow for each concept, which MPS will then use to calculate the dataflow for the whole program.
- Generator - Models written in one or more languages get ultimately translated into runnable code in some target general-purpose language and platform, such as Java. Along the way models get gradually transformed so that repeatedly concepts get replaced with concepts from a lower level of abstraction until the bottom-line level is reached. The rules for translating concepts and their proper ordering is defined in the Generator aspect.
- TextGen - During code generation after the Generator has reached the bottom-line AST representation, the TextGen phase kicks in and translates all nodes in the model into their textual representation and saves the resulting textual source files on disk.
Step 4 - Creating your first language
Raw time estimate - 3 hours
The next screen-cast to watch, MPS basics - creating your first language, will show you the typical workflow when creating a new language - building the structure, defining the editor and implementing the generator. You will see the creation process of the Constants language, which is also included as a sample project with MPS distributions.
Now you're recommended to watch a series of short high-level videos that will give you a brief introduction into each of the aspects of a language definition:
- Introduction to JetBrains MPS, part 1: Projects - This episode covers the MPS project setup and organisation of modules and models including their dependencies.
- Introduction to JetBrains MPS, part 2: Structure - This episode provides a brief theoretical background into models, meta-models and abstract syntax trees and then applies the knowledge to the MPS structure aspect.
- Introduction to JetBrains MPS, part 3: Constraints - This episode details the constraints aspect of language definition in MPS, mainly how to restrict properties, links and how to define scopes.
- Introduction to JetBrains MPS, part 4: Behavior - This episodes adds a few useful tips on adding functionality to concepts and nodes.
- Introduction to JetBrains MPS, part 5: Editor - Make'em see your code - defining editors that project the AST on the screen
- Introduction to JetBrains MPS, part 6: Actions - Polish the editors - smoothing the editing experience by defining transformations and substitutions
- Introduction to JetBrains MPS, part 7: Intentions - Assist the developer with context-sensitive hints and refactorings
- Introduction to JetBrains MPS, part 8: Generator - Converting models - defining model-to-model transformations
- Introduction to JetBrains MPS, part 9: Text-Gen - Here's what I've written - converting models to text
- Introduction to JetBrains MPS, part 10: Dataflow - Go with the flow - defining dataflow definitions so that MPS could automatically detect issues in code structure
- Introduction to JetBrains MPS, part 11: Type-system - They are just my type - defining types and type-system rules to validate expressions early-on
You may also watch these as a single YouTube playlist.
Step 5 - Why MPS
Raw time estimate - 30 minutes
Now that you understand the internal workings of MPS you may like to slow down a bit and read a few documents describing how people use MPS. We've collected a few case studies that may inspire you to your own ideas on building something great around MPS.
- You may find a complete list of user stories in the Use cases section of the MPS documentation.
- You may also check out a list of all academic publications on MPS that the MPS users ahem published over time
Step 6 - Projectional editing tips and tricks
Raw time estimate - 30 minutes
When editing code in MPS you will run into situations when your coding habits interfere with the projectional nature of the MPS editor. I'm actually pretty confident that you have already discovered such surprises during your experiments. Code in MPS is not text, but instead it is a projection of the abstract syntax (AST). This gives several benefits to language designers and, at the same time, makes the editing experience slightly different and more-restrictive than in the commonplace text editors. This may or may not be considered to be an advantage, nevertheless, there is a learning curve to projectional editing. In general, in projectional code you more rely on the assistance of the editor. Code-completion, block selection and intention actions should become your close friends in order to become an efficient projectional coder. There's a good and a bad side to it:
- good - in less than a week of programming in the MPS editor people typically get back to their full speed of coding they experienced before in text-based IDEs
- bad - projectional editing is highly addictive and you may find text-based editors less compelling and less helpful than you thought they'd been before
Step 7 - Shapes tutorial
Raw estimate - 2 hours
Here's an opportunity for you to quickly get your own language. Check out the MPS entry-level Shapes Tutorial and build a language that enables non-programmers to build scenes of graphical shapes.
Step 8 - Calculator tutorial
Raw time estimate - 8 hours
Now you must be curious to find out more details about language design in MPS, right? Having given you all the fundamental information and some level of experience, the introductory tutorials should have left you in a good position to dive deeper and build a sample language all the way from the ground up. The Calculator tutorial will guide you through the process of language creation. You start with an empty project and end up with a language to build Java Swing-based visual applications. Try it out and have fun.
Step 9 - Other sample projects
MPS comes with several sample projects that you may play with to get a better grasp of language development. I recommend you tried them out, perhaps in the order as they are listed below. Find out what they do and how they are implemented. They may teach you important lessons. Don't worry if you sometimes cannot figure out how a particular feature has been implemented. Most likely you will find an answer in one of the following exercises and then you can revisit the sample and the thing will just click instantly.
Here's a list of the samples that we've prepared for you:
- SampleJavaExtensions (updated) - a collection of small handy Java enhancements, such as parallel for loop, unless statement, Money literal, decision tables and Constants definition. The sample will teach you how to extend existing general purpose languages with custom constructs. You may also like to watch a short screen-cast introducing the individual extensions.
- Complex - defines a Java extension to work with complex numbers
- XML Literals - allows XML literals to be used directly in Java code as first-class language elements. The language builds on top of jetbrains.mp.xml and generates into into org.jdom.* API. The user-inserted XML code may refer back to the surrounding Java context to easily parametrize.
- math - provides fancy-looking math language extension to Java, which allows you to enter complex numbers, matrixes, intervals, summation, product and other common mathematical operations in a natural way.
- LanguagePatterns (updated) - a set of sample languages illustrating implementations of common language patterns in MPS. It is useful for beginners to see in isolation concrete implementations of typical language constructs. This sample project is gradually being updated with new patterns.
- MultipleProjections - an example of a simple state machine language that focuses on the mechanism of multiple projections per concept. The programmer, who uses the language, can on-the-fly switch between several pre-defined notations as she progressed with her state-machine definition. All the available notations are defined in the language or its extensions.
- MultiTarget - illustrates how to have a single language generated into multiple different target languages.
- Custom Aspects - shows the ways to define custom aspects for language definition. Check out the Custom language aspect cookbook for details.
- Component Dependencies - a minimalistic example of a language for expressing dependencies between components of a system, which provides three alternative editor notations:
- Customized Debugger - an example of hooking a languages with the Java debugger, even when the language neither extends nor generates into BaseLanguage.
- Mind maps - an intuitive example of defining a diagramming notation, including a set of customized graphical symbols
- State machine (Secret Compartment language/DSL book sample) - an example of a state machine language, including a state machine test language. An overview screen-cast is available to help you get started.
- The Simplest Language - a minimalistic example generating a Hello world application printing out a customized message. Shows the very basics of language generation.
- Migrations - a simple example of migration scripts to migrate two interconnected languages using both the smodel API and TransformStatement.
- Lambda Calculus - a language simulating lambda calculus, with projectional editor and generation into Java
- Agreement - an example of a business-oriented DSL
- Expressions - a simplified expression language to allow you to play with the type system and the Type-system Trace tool. The language gives you a bare-bone expressions language with minimal editor support, but with well structured type-system rules, which you can help you learn the core of how MPS calculates types. Read more in the Typesystem Debugging documentation.
- Fixed Length Reader - illustrates how to create a language for describing binary data formats. Comes with a stub for a hypothetical runtime library that would extract elements from binary data using the supplied data format description (called configuration).
- HighLevel Languages - an example of defining a custom set of language definition languages. Illustrates how to define your own way to codify structure and editor for a language.
- Sample Language - A very intuitive and minimalistic example showing how to implement a common pattern of a root concept containing definitions as well as references to them. In this sample, we simulate Java classes, method definitions and method calls. However, such a pattern is quite frequent in many languages and it is worthwhile to spend some time investigating. You may also check out a brief screen-cast showing you round this sample.
- Formula Language - a sample definition of an expression language, including editor actions and a generator
- NotesOrganizer - a sample DSL plugin for IntelliJ IDEA including a build script, that simulates a task/todo list. Tasks can have different priorities, states and categories, which by themselves can also be customized. The task lists can be filtered using several kinds of Swing UI components and the user can choose among several alternative visualizations of the task lists. Shows as an example of filtering node collections, alternative editors and direct incorporation of Swing components into form-like notations.
- Attributes - a sample illustrating the transparent handling of attributes in text-gen and the generator. The two CommentAttribute node attribute concepts are passed through the generator unchanged so that their text-gen can convert them to text.
- Sample Facet - illustrates a simple make facet that intercepts the make process.
- Lightweight DSL - examples of usages of the jetbrains.mps.baselanguage.lightweightDSL language to build internal "lightweight" DSLs embedded into BaseLanguage classes.
- Heating controller - a simple language providing two notations to specify heating plans for a hypothetical home heating controller. The default notation allows for textual representation, while a language extension provides an alternative tabular notation.
- Progress Indicator - a set of actions that illustrate the proper way to handle progress bars in actions, how to make actions cancellable and enable them to run in the background.
- xmlPersistence - a non-trivial example of implementing custom persistence using the new Open API.The plugin contains three solutions (a build script, a persistence implementation and a plugin descriptor). The custom persistence implementation will load/save XML documents. Please check out our Custom Persistence Cookbook for details.
Samples to tutorials:
- Shapes (updated) - the language created as part of the introductory Shapes Tutorial.
- Robot Kaja (updated) - the language for commanding a virtual robot used in many of the introductory screencasts. Now includes a sample Context Assistant implementation.
- Calculator - the language created as part of the Calculator Tutorial.
- Generator demo - the set of languages developed as part of the Generator Tutorial.
Step 10 - Generator tutorial
Raw time estimate - 12 hours
The Calculator tutorial should have left you with good overall knowledge of language design. The Generator tutorial will now teach you more subtle options that the MPS generator offers. You will learn about the different types of macros, generator scripts, utility models and more.
Reviewing the progress
Congratulations! By completing the above tutorials and exercises you've made a serious dent into the language design world. Your knowledge and command of MPS could now be qualified somewhere above the Advanced beginner stage of the Dreyfus model of skill acquisition. This is definitely a good time to celebrate and perhaps also to let us know about you success.
Moving up the ladder
The next stage in the Dreyfus model of skill acquisition is Competent. Once you reach that stage you'll be able to work on your own without concrete guidelines and supervision. But for this to happen, you'll need to spend more time solving various problems with MPS. If you already have an idea of a language to implement, go ahead and do it. If not, try to come up with a useful little language. You could as well try to re-implement one or more of the MPS samples and then see how closely you got. Or just pick a sample as the starting point and move it forward - add language features, implement a generator, add type-system rules, polish the editing experience - whatever feels like the right challenge for you.
Additionally, you should try these in order to become Competent in MPS and language design:
- Study the Common language patterns
- Study the cookbooks to deepen your knowledge of the various aspects of language design
- Get to know quotations and anti-quotations
- Familiarise yourself with the languages bundled in MPS, such as closures, collections, concept functions, regexp, tuples, smodel, build language and other
- Learn how to build MPS projects from the command line, how to package languages into IntelliJ IDEA as well as MPS plugins, how to use the plugins and how to build standalone DSL IDEs
- Get a feel for the user guide so you can quickly find the information that you need
- Learn some of the keyboard shortcuts by heart
- Become active in the MPS community and try to answer the forum questions of people that walk behind you on the same path to MPS mastery
After becoming Competent you should definitely continue gaining experience. The more you work with MPS and the more divert your projects are, the faster you'll be moving up on the Dreyfus model of skill acquisition scale. If you set yourself a goal to become Proficient and ultimately an Expert in MPS, you might consider trying:
- Studying and practising advanced topics, such as:
- Helping MPS expand into there areas by creating a base language for a different target language/platform - Python, Ruby, Dart, Haskell, Go, etc
- Speaking about MPS at events
- Helping us evolve the documentation
- Contributing code to the MPS project
Although there's not much detailed documentation to guide you at this experience level and you'll have to be self-driven most of the time, we'll always be happy to discuss, advise and cooperate with you.
Here's a list of documentation sources that you may need on your path to MPS mastery
Fabien Campagne wrote The MPS Language Workbench: Volume I and Volume II books, which you can purchase both electronically and as paper-back. Get them here.
The publications page collects all relevant papers about MPS and the related field.
The get slightly simplified practical information as well as some tips and tricks on individual aspects of language design, you may try some of the cookbooks.
- Common language patterns - if you need guidelines on how to implement certain functionality
- Editor cookbook
- Generator cookbook
- Description comments - a cookbookd showing how to leverage attributes to add support for description comments attached to arbitrary nodes
- Cookbook - Type System
- SModel language reference documentation to show you how to query your models
- Regular expressions
- Finding your way out - a brief collection of guidelines that should help you move forward when you get stuck somewhere.
Full-blown user guides will provide exhaustive information on their respective subject.
- Java developer's User Guide (IntelliJ IDEA)
- DSL developer's User Guide (standalone MPS-based IDEs)
- Language designer's User Guide
Over time the MPS team has created a large collection of screen-casts covering many aspects of MPS use and language design. You may check them out all nicely sorted at the MPS screen-cast page or directly at the MPS channel of JetBrains TV.
The MPS team offers hands-on trainings. Check out the details:
The MPS documentation page will give you all the remaining pointers to case studies, tutorials, FAQs and other pieces of MPS documentation.
To set common terminology and avoid confusion, we wrote down the most frequent terms for you:
Abstract Syntax Tree (AST)
a logical representation of code in memory (and disk) in the shape of a tree forest that describes hierarchies of nodes. These nodes have a notion of a parent-child relationship. Additionally, two nodes can be mutually connected with explicit references that go across the hierarchy structure.
a projectional clone of Java 6 (with optional extensions for Java 7 and 8). It follows the Java specification and is 1:1 compatible with Java 6. Additionally, MPS provides several handy extensions to BaseLanguage, such as dates, collections, closures and many others.
the process of transfromation code from one model (AST) into another model. For example, code describing a set of business rules can be transformed into plain Java so that it can be compiled with javac and run as part of an enterprise application.
A package of related languages that have been grouped for user convenience.
Domain Specific Language (DSL)
a language dedicated to a particular problem domain, typically created with the aim of simplicity and greater expressivity compared to a general purpose language.
a packaged library ( a zip file) containing all the required elements in order to use a language either inside either IntelliJ IDEA or MPS.
an editor that allows the user to edit the AST representation of code directly, while mimicing the behavior of a text editor to some extent. The user sees text on teh screen and edits it, however, in reality the text is only an illusion (projection) of an AST.
The top-level organization element of an MPS project that typically groups several models together. It can have three basic types: Solution, Language and DevKit and may depend on other modules and models.
A lower-level organizational element grouping individual concepts. It may depend on other models.
A solution that is required by a language, sometimes also called a library. Runtime solutions may contain normal models as well as stubs for Java sources, classes or jar files external to MPS.
A language aspect defining all types (concepts) of AST nodes that can be used in the language together with their relationships.
A definition that describes the abstract structure of a syntax element. E.g. the IfStatement concepts says that an if holds a boolean Expression and up-to two StatementLists.
A language aspect holding additional restrictions on concepts, their properties and relationships.
Allows the language designer to define behavior of the language concepts.
Holds vizualization definitions of individual language concepts. Since the way concepts are viewed and edited on the screen can be customized, the editors specify how the user will interact with the language.
The set of elements that are visible and applicable to a particular position within a program. Typically only a sub-set of all elements of a particular kind can be used at any given program location.
A set of rules that validate and infer types of concepts in a program.
User-invoked commands that may perform changes to the code. Actions can be attached to keyboard shortcuts or menu items.
Context-sensitive actions offered to the language user through a small pop-up window triggered by the Alt + Enter key shortcut. These actions typically perform a relatively local refactoring to the code under carret or a selected block of code.
Surround With intention actions
Intentions applicable to a selected block of code that wrap the block by another concept. E.g. Surround with Try-Catch.
A potentially substantial automated change in code structure triggered by a user action.