1 / 24

The Umple Model-Oriented Programming Technology: Easy-to-Use Open-Source Code Generation

The Umple Model-Oriented Programming Technology: Easy-to-Use Open-Source Code Generation. Code Generation 2013, Cambridge UK Timothy C. Lethbridge, University of Ottawa tcl@eecs.uottawa.ca http://www.eecs.uottawa.ca/~tcl http://www.umple.org. My background. Once worked at Northern Telecom

hugh
Download Presentation

The Umple Model-Oriented Programming Technology: Easy-to-Use Open-Source Code Generation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. The Umple Model-Oriented Programming Technology: Easy-to-Use Open-Source Code Generation Code Generation 2013, Cambridge UK Timothy C. Lethbridge, University of Ottawa tcl@eecs.uottawa.ca http://www.eecs.uottawa.ca/~tcl http://www.umple.org

  2. My background • Once worked at Northern Telecom • Collaborated with originators of ObjecTime / IBM Rational Tools since early 1990’s • Professor of software engineering at University of Ottawa • Have taught software engineering since 1991 • Research focuses • Usability of software engineering tools • Empirical studies • Recent research support: IBM, Ericsson, General Motors, Defense Deptartment Code Generation 2013 - Umple

  3. Motivation • From our studies of practitioners • Code generation is weak in widely-deployed tools • Tools are awkward, do not interoperate • Modellers mostly • do it for documentation • make important errors • From our teaching of modeling • Students learned syntax, but failed to learn • Semantics • Pragmatics • To create implementable models due to lack of feedback Code Generation 2013 - Umple

  4. Umple: Simple, Ample,UML Programming Language • 1. Open source code generator for UML • The only one that has • Infinitely nested state machines, with concurrency • Proper referential integrity and multiplicity constraints on associations • 2. Textual modelling tool set • Command line compiler • Web-based tool (UmpleOnline) for demos and education • Eclipse plugin • 3. Pre-processor to add UML, patterns and other features on top of Java, PhPand other languages Code Generation 2013 - Umple

  5. Exercise: Compiling and Changing a Model • Go to http://helloworld.umple.org • Look at the example at the bottom • Observe: attribute, association, class hierarchy, mixin • Click on Load the above code into UmpleOnline • Observe and modify the diagram • Add an attribute • Make a multiplicity error, then undo • Generate code and take a look • Download, compile and run if you want Code Generation 2013 - Umple

  6. Demo of the command line Umple compiler • Follows conventions of gcc • Used by ant scripts for building Code Generation 2013 - Umple

  7. Umple Philosophy 1-4 • P1. Modeling is programming and vice versa • P2. An Umple programmer should never need to edit generated code to accomplish any task. • P3. The Umple compiler can accept and generate code that uses nothing but UML abstractions. • The above is the inverse of the following • P4. A program without Umple features can be compiled by an Umple compiler. • e.g. input Java results in the same as output Code Generation 2013 - Umple

  8. Umple Philosophy 5-8 • P5. A programmer can incrementally add Umple features to an existing program • Umplification • P6. Umple extends the base language in a minimally invasive and safe way. • P7. Umple features can be created and viewed diagrammatically or textually • P8. Umple goes beyond UML Code Generation 2013 - Umple

  9. A look at selected Umple features • Attributes • http://attributes.umple.org • Associations • http://associations.umple.org • Notice the inline and independent state machines • State Machines • http://statemachines.umple.org • Switch to ‘GraphViz state diagram’ in the options menu Code Generation 2013 - Umple

  10. Umple Architecture 1 • Grammar DSL: Online • http://grammar.umple.org • (we migrated from Antlr to better handle embedded code blocks and gain flexibility) • Umple is written in itself • Key to testing and reliability • http://code.google.com/p/umple/source/browse/trunk/cruise.umple/src/Umple.ump Code Generation 2013 - Umple

  11. Umple Architecture 2 • Umple’s own class diagram generated by itself from itself: • http://metamodel.umple.org • Colours represent key subsystems • Click on classes to see Javadoc, and then Umple Code Code Generation 2013 - Umple

  12. Testing:TDD with100% pass always required • Multiple levels: • Parsing tests: basic constructs • Metamodel tests: ensure it is populated properly • E.g. http://code.google.com/p/umple/source/browse/trunk/cruise.umple/test/cruise/umple/compiler/AssociationTest.java • Implementation template tests: to ensure constructs generate code that looks as expected • Testbed semantic tests: Generate code and make sure it behaves the way it should Code Generation 2013 - Umple

  13. Continuous integration • Small increments compiled frequently http://cc.umple.org • Single trunk, no branches for now • Online QA report:http://qa.umple.org Code Generation 2013 - Umple

  14. Code Generation Approach 1 • Generation relies on ‘walking’ the metamodel • Where possible, new features are added by layering on top of the existing ones • e.g. • There was an existing ‘code injection’ feature • Constraints were implemented almost entirely at the metamodel level by adding ‘before’ and ‘after’ code injections Code Generation 2013 - Umple

  15. Code Generation Approach 2 • Each generator is a separate plugin and can have a different architecture • Use Jet templates • Language independent common templates • http://code.google.com/p/umple/source/browse/#svn%2Ftrunk%2FUmpleToTemplate%2Ftemplates • Language-specific templates • http://code.google.com/p/umple/source/browse/#svn%2Ftrunk%2FUmpleToJava%2Ftemplates • Language-specific translation functions • http://code.google.com/p/umple/source/browse/trunk/cruise.umple/src/Generator_CodeJava.ump Code Generation 2013 - Umple

  16. Code generation lessons from Umple 1 • Process of developing Umple • Agile continuous integration • Language written in itself • Multi-level test-driven development • Process for all users of Umple • Modeling and programming at the same level • Many semantic model checks produce warnings & errors • Round trip engineering is banished • Generation of diagrams from text and text from diagrams • IDE Flexibility: Eclipse / Web / Command line • Incremental umplification Code Generation 2013 - Umple

  17. Code generation lessons from Umple 2 • Language features • Deep attention to detail in state machines and UML associations • Mixins for separation of concerns • Multi-lingual generation and embedding • Generated code looks nice • No ‘Umple runtime’ • Can be inspected to gain confidence • No risk adoption Code Generation 2013 - Umple

  18. Evidence of benefit • Umple in use an a couple of small companies • Umple development in itself • 20+ developers • Empirical studies show • Umple code is easier to understand than Java • Helps students learn to model: CSEE&T Code Generation 2013 - Umple

  19. Results of survey Prof. Lethbridge should use it in future courses Umple in the lab helped me understand concepts Other tools would have been just as effective I would use Umple in the future, assuming bugs fixed I will get bettergrades because of Umple Umple in the classroom helped me understand concepts As of Dec 2010, Umple is too incomplete or buggy CSEE&T 2011

  20. Ability to generate code to represent the model Availability of a textual form for models and diagrams Ability to edit diagram or text & have the other reflect changes CSEE&T 2011

  21. Comparison of Grades • Comparison of 4 offerings of the same course before introducing Umple, and 2 offerings afterwards • Midterm examination grades on UML modeling question rose from 75.8% to 82.3% • 8.6% improvement • Final exam grades on UML modeling questions rose from 76.4% to 83.6% • 9.4% improvement • Highly statistically significant p<0.00002 on T Test • n=332 prior to Umple; n=137 afterwards • Similar results in 4th year course SEG4110 • No difference in non-UML questions CSEE&T 2011

  22. The Near Future • Real time UML/Autosar extensions • Robust C/C++ generation • Application for Eclipse Foundation hosting • Seek industrial uptake Code Generation 2013 - Umple

  23. A Few Direct Web Links • UmpleOnline • http://try.umple.org • User Manual • http://manual.umple.org • Google Code Hosting Site • http://code.umple.org • This presentation: • http://www.site.uottawa.ca/~tcl/presentations/UmpleTutorial-CodeGeneration.ppt Code Generation 2013 - Umple

  24. Questions? Code Generation 2013 - Umple

More Related