Towards implementation of a rich type system for microservices-based systems

Larisa Safina Innopolis University, 2015

1

Agenda 1. 2. 3. 4. 5.

Project area: Microservices Project area: Jolie programming language Problem statement Adding the choice data type Conclusions and future work

2

Project area: Microservices • • • • • •

Componentization via Services Organized around Business Capabilities Decentralized Data Management Culture of automation Design for failure Evolutionary Design

http://martinfowler.com/articles/microservices.html#InfrastructureAutomation

3

Project area: Jolie programming language http://www.jolie-lang.org/ Native microservice programming Reusability Distributable by design Protocol agnostic Structured workflows Dynamic error handling for parallel code Formal specifications http://www.jolie-lang.org/

4

Jolie programming language

5

Programming example

6

Communication

7

Program structure Deployment part Behavioral part

inputPort CarRentService { Location: "socket://localhost:2000" Protocol: sodep Interfaces: CarRentInterface }

main{ get_car( request )( response ){ get_price; response = price*1,2 } }

8

Type system Native types and subtypes Linked types Undefined types

type car_request: void { .name: string .id: {?} } type customer: void { .name: string .age?: int .cars*: car_request }

9

Interpreter architecture

Fabrizio Montesi. Jolie: a Service-oriented Programming Language. Master’s thesis

10

Jolie Interpreting Algorithm 1. 2. 3. 4.

Command line arguments are read, and the components needed for parsing are started; The source code in input is parsed and an Abstract Syntax Tree (AST) is produced; Well-formedness, semantic checks and optimizations are performed on the AST; The AST is used in order to generate the OOIT and initialize the Communication Core and Runtime Environment components; 5. The Runtime Environment calls the run method of the root node in the OOIT (which corresponds to the main procedure).

11

Problem statement Lack of useful data types in Jolie • Choice types • Regular expressions • etc...

12

Adding new data type. Type choice • Choice type in XML Choice type in Jolie type animal: cat | dog

13

Adding new data type. New programming examples • Working with existent types type numeric: int | double | void Linked types: type linked_type: string type linked_choice: linked_type | raw Subtypes: person_info: void { .id: string | int }

14

Adding new data type. Changes in interpreter Adding choice type to AST

15

Adding new data type. Changes in interpreter • Extending parser: – Scanner – OLParser – OLParserTreeOptimizer – Semantic verifier Extending runtime facilities Additions to building object-oriented interpretation tree process Addition to message types

16

Adding new data type. New programming examples Making functions generic type choice: string | int fun_choice(choice)(choice) fun_choice(request)(response){ response = request }

17

Adding new data type. New programming examples Program behaviour differs based on the type of arguments passed type person: personSSN | personCCN type personSSN:void {.ssn:int} type personCCN:void {.ccn:string} pay(person)(response) { if ( is_defined( person.ssn ) ) { // ask the person registry.. } else { // contact the bank.. } }

18

Adding new data type. Yet Another Example type Old-Software-Corp: void {. name: string address: string } type New-Software-Corp: void {. name: void { .firstname: string .lastname: string address: string phone: int } type corporation: Old-Software-Corp | New-Software-Corp

19

Conclusions and future work Conclusions • Adding choice operator provide more flexibility Future work Fixing Jolie tools regarding new types system Adding regular expressions

20

References 1. Martin Fowler and James Lewis, “Microservices” http://martinfowler.com/articles/microservices.html 1. Jolie programming language documentation http://www.jolie-lang.org/documentation 1. Fabrizio Montesi. Jolie: a Service-oriented Programming Language. 2. XML Schema choice element http://www.w3schools.com/schema/el_choice.asp

21

Thank you! Questions?

22

Backlog

23

Jolie BNF Program ::= Deployment Behaviour Deployment ::= DeploymentInstruction∗ Behaviour ::= BehaviouralBlock∗ main { Process } BehaviouralBlock∗ BehaviouralBlock ::= define id { Process } | init { Process } Process ::= . . . | InputStatement | OutputStatement InputStatement ::= op(x) (One-Way) | op(x)(y) { Process } (Request-Response) OutputStatement ::= op@OPort(x) (Notification) | op@OPort(x)(y) (Solicit-Response)

24

Jolie Type System DeploymentInstruction ::= . . . | type id : TypeDefinition TypeDefinition ::= . . . | NativeType | NativeType { SubTypeList } | NativeType { ? } (Untyped subnodes ) | id (Type link) | undefined (Shortcut for any:{?}) NativeType ::= int | double | string | raw | void | any

SubTypeList ::= SubType | SubType SubTypeList SubType ::= .id Cardinality : TypeDefinition Cardinality ::= [ int , int ] (Range) | [ int , * ] (Lower-bound) |* (Shortcut for [0,*]) |? (Shortcut for25 [0,1])

Jolie Parser

26

Jolie Type Definition TypeDefinition (jolie.lang.parse.ast.types): • TypeInlineDefinition for expressing native types; • TypeDefinitionLink for linked types; • TypeDefinitionUndefined for undefined types.

27

Jolie Parse Process 1. Scanner reads input and create token objects based on the ones defined in enum. 2. OLParser takes created tokens, checks them by the grammar rules and generate corresponding syntax node (jolie.lang.parse.ast.OLSyntaxNode) in abstract syntax tree (AST). 3. OLParseTreeOptimizer (jolie.lang.parse.OLParseTreeOptimizer) takes ready AST and optimize it by reducing the number of nodes or transforming the code to more efficient versions. 4. SemanticVerifier (jolie.lang.parse.SemanticVerifier) checks if the code well-formed and semantically correct.

28

Jolie Runtime Environment

29

Jolie Communication Core

30