pep-006-business-framework-4.md

PEP: 006
Title: Business framework 4.0
Author: Adrien LAUER <adrien.lauer@mpsa.com>  
Status: draft  

Abstract

The goal here is two-fold:

• Refactor business framework legacy parts into something more modern,
• And take advantages of Java 8 features everywhere it makes sense.

Motivation

We want to improve the developer experience of using the business framework. For this we must fix long-standing oddities and use newer features of the language where it makes sense. Functional upgrades of existing patterns can also be considered.

Specification

GenericRepository

The GenericRepository interface is a left-over from older versions and is replaced in purpose by the Repository interface. It should be removed.

GenericFactory / Factory

To be consistent with other patterns, GenericFactory should be renamed Factory and Factory should be removed. The create(...) method should have a default implementation in the interface so it won't need to be implemented by custom factories.

Repository

The Repository interface should be reworked to be more inline with the original spirit of the repository pattern:

A repository represents all objects of a certain type as a conceptual set (usually emulated). It acts like a collection, except with more elaborate querying capability. […] For each type of object that needs global access, create an object that can provide the illusion of an in-memory collection of all objects of that type.

The proposed interface is the following:

@DomainRepository
public interface Repository<A extends AggregateRoot<K>, K> {

    void add(A aggregate);

    Stream<A> get(Specification<A> specification, Options... options);

    Optional<A> get(ID id);

    boolean contains(Specification<A> specification);

    boolean contains(ID id);

    boolean contains(A aggregate);

    long count(Specification<A> specification);

    long size();

    boolean isEmpty();

    long remove(Specification<A> specification);

    boolean remove(ID id);

    boolean remove(A aggregate);

    void update(A aggregate);

    void clear();

}

FluentAssembler

Make FluentAssembler accept streams

Fluent assembler should be able to use streams instead of lists for assembling multiple aggregates/dtos. Combined with the streaming ability of repositories, this would enable advanced use-cases:

public class SomeClass {
    @Inject
    private Repository<Customer, CustomerId> customerRepository;
    @Inject
    private FluentAssembler fluentAssembler;
    
    public void someMethod() {
        fluentAssembler
            .assemble(customerRepository
                .get(CustomerSpecifications.RecentClients, Sorting.Natural)
                .parallelStream()
                .filter(customer -> customer.getGender() === CustomerGender.FEMALE))
            .to(CustomerDTO.class)
            .collect(toList());
    }
}

Assembler interface

Sometimes only one-way assembling is needed. The current Assembler interfaces forces to implemented both ways and should be splitted into two separate interfaces: DTOAssembler and AggregateAssembler.

Assembler parameters

Parameters for assembler methods should be reversed, with the source as first parameter and the target as second parameter.

Pagination

Pagination should be reworked based on new capabilities of repositories, notably the ability to return streams from specifications. A pagination DSL could help developpers to extract a part of the results automatically.

Page pagination

public class SomeClass {
    @Inject
    Paginator paginator;
    @Inject
    FluentAssembler fluentAssembler.
    @Inject @Jpa
    Repository<MyAggregate, String> myRepository;
        
    public void someMethod() {        
        Page<MyDTO> dtoPage = fluentAssembler.assemble(
            paginator
                .repository(myRepository)
                .options(...)
                .page(7)
                .size(10)
                .paginate(someSpec)) // or .paginate()
            .with(ModelMapper.class)
            .to(MyDTO.class);
    }
}

Chunk pagination

public class SomeClass {
    @Inject
    Paginator paginator;
    @Inject
    FluentAssembler fluentAssembler.
    @Inject @Jpa
    Repository<MyAggregate, String> myRepository;
        
    public void someMethod() {        
        Chunk<MyDTO> dtoChunk = fluentAssembler.assemble(
            paginator
                .repository(myRepository)
                .options(...)
                .offset(70)
                .limit(10)
                .paginate(someSpec)) // or .paginate()
            .with(ModelMapper.class)
            .to(MyDTO.class);
    }
}

Key-based chunk pagination

public class SomeClass {
    @Inject
    Paginator paginator;
    @Inject
    FluentAssembler fluentAssembler.
    @Inject @Jpa
    Repository<MyAggregate, String> myRepository;
        
    public void someMethod() {        
        Chunk<MyDTO> dtoChunk = fluentAssembler.assemble(
            paginator
                .repository(myRepository)
                .options(...)
                .after("insertDate", someDate).limit(10)
                .paginate(someSpec)) // .paginate()
            .with(ModelMapper.class)
            .to(MyDTO.class);
    }
}