Let s build. like they build. Markus Völter Bernd Kolb

Size: px

Start display at page:

Download "Let s build. like they build. Markus Völter Bernd Kolb"

Anabel Briggs
5 years ago
Views:

1 Let s build like they build Markus Völter voelter@acm.org Bernd Kolb kolb@itemis.de

2 B 0 Motivation

4 Examples 1

5 M Healthcare

6 Context & Motivation Mobile Apps that help patients w/ treatments Monitor side-effects and recommend actions Manage dosage of medications

7 Context & Motivation Mobile Apps that help patients w/ treatments Monitor side-effects and recommend actions Manage dosage of medications Algorithms for recommendations and dosage at the core of these apps. Safety-critical, since they could hurt patients. Customer develops many different apps/algos like this, efficiency of algo development is key.

8 Context & Motivation Health care professionals directly code algos, using a suitable language. typical DSL-based devapproach. Avoids indirections through requirements docs. Speed up dev significantly.}pretty

9 Some Language Impressions I

10 Some Language Impressions II

11 Some Language Impressions III

12 B Insurance

13 Context & Motivation Specify/Program Insurance Programs Write formal code in a DSL mixed with tables and text No tool support whatsoever No testing (except inspection) No reuse No modularity No varibility

14 Context & Motivation Specify/Program Insurance Programs Write formal code in a DSL mixed with tables and text No tool support whatsoever No testing (except inspection) No reuse No modularity No varibility

15 A real Example: Legacy Specify/Program Debug Debugging directly in C Search-for-use by text search Don t trust the documents may be outdated! Insurance Programs Pixelcrap Implement C Code Write formal code in a DSL mixed with tables and text No tool support whatsoever No testing (except inspection) No reuse No modularity No varibility Printed, PDF Developer reads spec Very idiomatic implementation Dev acts as a human compiler and implements it in C

16 Solution Approach Specify/Program/Test/Debug Insurance Programs Write formal code in a DSL mixed with tables and text Now with IDE support and executable tests The same notation!

17 A real Example: Future Specify/Program/Test/Debug Insurance Programs Generate C Code Still exactly the same C code, or improved as needed. Incremental Refinement/Refactoring of languages: Partially automated migration of models Add model natural notations (insurance-specific, math) Add Support for modularity, reuse, variants

18 M Salary

19 Context & Motivation Calculate Salaries + Taxes for Employees Various deployment platforms Complex Business Logic Variability over 16 States Based on evolving Law Dates & Currencies Temporal Data + Arithmetics Reactive Rules

20 Currencies and Dates

21 Temporal Data

22 Result Data and Rules

23 Result Data and Rules

24 IDE Support

25 B 2 Approach in a Nutshell

26 Integration of Fachler

27 [Business Knowledge] It s what makes a business tick. Distinguishes the business. Business Rules (Financial) Calculations Data Structures Mappings or Queries Validations Scientific Processes Contracts Processes UIs}

28 [Business Knowledge] It s what makes a business tick. Distinguishes the business. Contributed not by developers... but typically implemented in software

29 [Business Knowledge] SO HOW DOES IT GET INTO THE SOFTWARE? Contributed not by developers... but typically implemented in software

30 Reality

31 Goal!?

32 Let Business/Domain people contribute directly! Give them expressive, productive tools to do so!

33 Expressivity for Core Domain Knowledge User-Friendly Notation Great Tool/IDE Testing Meaningful Analyses Synthesis of Software

34 B Separation from Technology

35 Outdated Technology Obscure Business Logic

36 Outdated Technology Fachlichkeit burried Non-Understandable Logic in implementation code. Expensive to Evolve

37 Technology & Business Logic now have connected lifecycles.

38 Goal: Separate the Lifecycles Fachlichkeit Technology

39 M 3 Technical

41 Business Logic Outdated Technology Non-Understandable Logic Expensive to Evolve

42 Metamodel for Business Logic Clearly defined data structure to express all business-relevant structures, behaviors and non-functional concerns.

43 Metamodel for Business Logic {Data Structures Behavioral Rules Expressions Validations Special Types (eg. temporal)...

44 Metamodel for Business Logic {Data Structures Behavioral Rules Expressions Validations Special Types (eg. temporal)... Domains often have a rich language/vocabulary anyway; it just needs to be formalized. (DDD: Ubiquituous Languge)

45 Metamodel for Business Logic Semantics Clearly defined data structure to express all business-relevant structures, behaviors and non-functional concerns. Well-defined meaning of this data structure

Metamodel for Business Logic Semantics Clearly defined data structure to express all business-relevant structures, behaviors and non-functional concerns.

46 Metamodel for Business Logic Semantics Clearly defined data structure to express all business-relevant structures, behaviors and non-functional concerns. Well-defined meaning of this data structure IDE Support is possible Evolution is possible Portability is possible Type Checking Solver-Integration Model Checking Contracts

47 Metamodel for Business Logic Semantics Clearly defined data structure to express all business-relevant structures, behaviors and non-functional concerns. Well-defined meaning of this data structure Execution Engine Tech Infrastructure Technical Platform for correct, efficient and scalable execution

48 Metamodel for Business Logic Semantics Clearly defined data structure to express all business-relevant structures, behaviors and non-functional concerns. Well-defined meaning of this data structure generate code, deploy generate code, deploy transfer data, interpret transfer data, interpret Tech Infrastructure Technical Platform for correct, efficient and scalable execution

49 Metamodel for Business Logic Semantics generate code, deploy transfer data, interpret Tech Infrastructure

50 Metamodel for Business Logic Syntax Semantics L a n g u a g e generate code, deploy transfer data, interpret Tech Infrastructure

51 Metamodel for Business Logic Syntax Semantics generate code, deploy transfer data, interpret L a n g u a g e Syntax is critically important for Productivity Communication and Review Domain Expert Integration Only Buttons and Forms don t work! Tech Infrastructure

52 Metamodel for Business Logic Syntax Semantics L a n g u a g e generate code, deploy transfer data, interpret Tech Infrastructure

53 Metamodel for Business Logic Syntax Semantics generate code, deploy transfer data, interpret L a n g u a g e L a n g u a g e W o r b e n c h Tech Infrastructure

54 A Language Workbench a tool for defining, composing and using ecosystems of languages.

55 Other Language Workbenches TU Delft itemis/typefox Rascal CWI Amsterdam The Whole Platform Solmi/Persiani

57 Lessons Learned 4

58 B A Language is not Enough

59 Language Design::More than Lang GOOD Language Abstractions Notations Great IDE Syntax Coloring Code Completion Goto Definition Analyses Relevant Good Errors GREAT Refactorings Aligned with Processes Testing Write Tests Run them Report Back Debuggers Animate Execution Simulators

60 B Notation Notation Notation

61 Math Insurance Math

62 Feature Models Component Architectures

63 Dataflow Diagrams

64 Form Style

65 Form Style Fixed DS

66 M Influences on the Language

67 Language Design::Influences Domain Structure Non Functionals Permissions, IP, Sharing User Skills Sep. of Concerns Different Views Get a better tool :-) Educate, Put results in context Refactor towards Structure Model Purpose Analyze, Generate Tool Capabilities Notations, Editing, Scale Software Engineering Practices

of Concerns Different Views Get a better tool :-) Educate, Put results in

68 Language Design::Influences Domain Structure Non Functionals Permissions, IP, Sharing User Skills Sep. of Concerns Different Views Get a better tool :-) Educate, Put results in context Refactor towards Structure Model Purpose Analyze, Generate Tool Capabilities Notations, Editing, Scale Software Engineering Practices

69 M How to make People precise?

70 {Formulas, Rules Data Structures Tables Values Precision!= Programming Performance Scalability Robustness Deployment }

71 B Does this scale?

72 Does the approach scale? If structure, formalization, and tool support don t scale, then what will?? What are the alternatives? Excel? Wikis? Prose Documents?

73 Do the tools scale? In terms of overall system size? Yes, the system has to be broken down into models of manageable size, as usual. This requires some thought. In terms of team size? Yes, since we rely on established version control systems (git) to deal with groupware aspects; and yes, diff/merge works as expected. In terms of language complexity? Yes, in particular, since you can modularize the language definitions.

74 Can I find the people to do this? Yes, but it is a significant change, so: - it may be a significant education/training effort. - a few people might not get it - a few people may not want to do it.

75 B This is a threat!

76 Precision and Formality Different Processes Higher Efficiency -> New Skills -> Role Change -> Job Loss Automation Focus on Engineering Empower Business Ppl -> Job Loss -> Role Change -> Less Importance

77 Some people are afraid of this. Take them seriously.

78 A change of that must be managed!

79 M Is this the next legacy system?

80 Today s software is tomorrow s legacy system. Or is it?

81 Today s software is tomorrow s legacy system. Business change is hard Technology change is hard Separation of Concerns Keep BL free of technology Make it portable

82 Language Tech 1 V1 Generator 1 Runtime T 1

83 Language Tech 1 V1 V2 V3... Generator 1 Evolution Runtime T 1 Existing models become incompatble with new language Þ Language Versions Migration Scripts

84 Language Tech 1 V1 V2 V3... Evolution Generator 1 Runtime T 1 Generator 2 Runtime T 2 Retargetting Runtime Tech outdated, uncool or slow Þ Keep Lang Technology Keep Models Build new Generator

85 Migration Language Tech 1 Language Tech 2 V1 V2 V3... V4 V5 V Generator 1 Runtime T 1 Evolution Generator 2 Generator 3 Runtime T 2 Evolution Retargetting Language Tech outdated, uncool Þ Build new Tool Migrate Data Simple, because it well-defined domain semantics and free from technology stuff

86 Today s software is tomorrow s legacy system. No, it is not.

87 M In conflict with Agile?

88 MD* and Agile is in Conflict. 1 Project Language Development Depend on, use System Development Later: 2 Project 1 Language Development Dep d on, use Project 2 System Development Project N

89 MD* and Agile is in Conflict. 1 Project Language Development Depend on, use System Development Framework Library Platform Manage like any other intra-project dependency. Evolution of client code is easier than for F/L/P because of migration support!

90 MD* and Agile is in Conflict. Later: 2 Project 1 Framewor k Language Development Library Platform Dep d on, use Project 2 System Development Manage like any other 3rd party depencency: Development Roadmap Issue Tracker Release Notes...

91 MD* and Agile is in Conflict. 3 Project 2 System Development Models and DSLs are an Enabler for Agility: Integration of Domain Experts Living Requirements Decoupled Fachlichkeit & Technik

92 MD* and Agile is in Conflict. 4 Project 1 Language Development Leading LWBs are so productive, you can literally sit with the domain experts and interactively prototype languages (and then clean up later) I ve looked at the implementation of the language in MPS, but I didn t find much. Is this all there is? Where s the magic? [Customer]

93 MD* and Agile is in Conflict. 4 Project 1 Language Development Leading LWBs are so productive, you can literally sit with the domain experts and interactively prototype languages (and then clean up later) I ve looked at the implementation of the language in MPS, but I didn t find much. Is this all there is? Where s the magic? [Customer] Analyz e Build 1 hour Try out Cleanu p Stabiliz e Validat e 1 to 3 days

94 B Skills?

95 Organizations do not have the necessary skills. True. But... Big AI Data R E S T So built it. Evolve. Hire. Buy.

96 M Rockets????

98 Further domain-specific extensions to C. Developed by end-user lang engineer.

99 Infrastructure Specifics in C

100 Infrastructure Specifics in C

101 Generated Documentation

102 Separation of concerns is key to avoid the legacy trap DSLs can isolate business logic completely from technical concerns DSLs can help integrate domain experts with communication/review or even coding Language Workbenches enable DSLs by reducing effort to build, compose and maintain them Migrating to a new LWB is feasible b/c semantics of all models are known, by definition.

Markus Völter

of Markus Völter voelter@acm.org www.voelter.de @markusvoelter Examples 1 Healthcare Context & Motivation Mobile Apps that help patients w/ treatments Monitor side-effects and recommend actions Manage