Hey,

I've found a slight issue which I'm not entirely sure can be addressed in .NET, but essentially, structs can be initialized as default(...), regardless of what constructors exist.

Thus, I can easily create a SSN and (all other value objects) by using

SSN ssn = default;

While this may work fine if you're relying on everyone initializing these value objects using their constructor, would it not be better to make SSN, Name, etc. classes?

You could, in theory, also use SSN ssn = new SSN(), but that is a stylecop error and not a .NET error.

HI @petrspelos,

I am developing the https://github.com/ivanpaulovich/FluentMediator to support the mediator pattern in a decoupled approach.

Would be using it to publish Commands/Queries from the WebAPI into the UseCases a good idea?
Would it be interesting to see on the vNext of Manga project?

Hello,

Could you kindly publish a web interface like the example below?

https://github.com/EduardoPires/EquinoxProject/tree/master/src/Equinox.UI.Site

Or could you show me of an example someone has done.

It would be nice to add /metrics endpoint with Prometheus metrics, following Reporting Metrics To Prometheus In ASP.NET Core is straightforward.

I would like to see support for multiple currencies.

The challenge is that the current implementation uses a Value Object to describe Money in which there is no information regarding the Currency so it is not possible to deposit Euros and Dollars to the same account history.

In the real world the customers deal seamlessly with multiple currencies and the bank is responsible for the exchange service.

How can we do the same in a DDD approach?

My first thought is to change the Money Value Object to include info regarding the Original Currency and the Value in Euro then design a Factory object that is initialized with an Exchange Service.

After that we need to change the controllers to use the Factory instead of instantiating the Money directly.

Thoughts?

Hey all,

I just released

• 📺 DDD Tactical Design and Clean Architecture - Youttube.
• 📖 DDD Tactical Design and Clean Architecture - Slideshare.

Once more I'd like to thank you for bringing ideas, coding and energy to this project 🍺.

Output messages could be complex and dependent on multiple steps to be created. What if we change the BuildOutput() method to a proper builder pattern?

We added the .ruleset files on each project and I had to disable lots of rules so we can move forward.
I think it is time to enable those rules and update the code base accordingly.

hello ivan, congrats for the nice job you ve done.
btw I ve read the code, and I miss an example of account transfering among two diffrents accounts.
how would it be done within that architectural approach?
May it be something related with Unit of Work?

Even though Manga is a monolith application it would be nice to enable OpenTracing to show how to add it following the Clean Architecture principles.

The Jaeger library is a interesting starting point:

• https://github.com/jaegertracing/jaeger-client-csharp
• https://github.com/GokGokalp/OpenTracing-Jaeger-Sample/blob/master/User.API/Startup.cs

Hi Ivan

I ve found something really interesting in your project Mango

I would like to talk detailed about it to future

Hi Ivan,

Great repo, really usefull stuff!

I have one question though, while researching the ports and adapters pattern / hex arch I stumble upon the following issue:

Should each driver interface (port/usecase in your case) only do one thing? Or is it allow to have, lets say, a customer interface which allows for more than one action a customer.

My concern is that a project will become very big if I were to add an interface for each usecase.

Kind regards,

Bob

Having token based authorisation over endpoints. How the endpoints are secured? Does this plan in on the cards?

Hello,

I have been discovering this great project for a few days now and I already wanted to thank you. :)

Would it be interesting to refactor the implementation of the IEquatable interface of value objects?

Here is a more concise implementation possibility used in ReSharper and Rider.

What do you think of that?

namespace Domain.ValueObjects
{
    using System;
    
    public struct Name : IEquatable<Name>
    {
        private readonly string _text;

        public Name(string text)
        {
            if (string.IsNullOrWhiteSpace(text))
                throw new NameShouldNotBeEmptyException("The 'Name' field is required");
            
            _text = text;
        }

        public override string ToString()
            => _text;

        public bool Equals(Name other)
            => _text == other._text;

        public override bool Equals(object obj)
            => obj is Name other && Equals(other);

        public override int GetHashCode()
            => _text != null ? _text.GetHashCode() : 0;
    }
}

Hi Ivan, first of all, great job.

Based on this architectural proposal, how would you suggest adding localization capability to this project? I'm asking it because APS.NET core provides great support for localization, but when I thinking about Domain Layer, Application Layer it is not so clear to me. My questions are:
1 - Where the resource files should be?
2 - How we can resolve the domain model and value objects?
3 - How would you solve it thinking in tests, business rules?

Thank you.

Hello,

First of all great repository! Good basis to get started with a clean arch service. I am wondering why you recently added interfaces on the entities? I've been told it is an an anti pattern. What exactly is the advantage of adding them (or removing) them?

Thanks!

Hello and thank you for your great project.
i have a question why we have Infrastructure reference in Web Ui ?
base on diagrams we should not depend on Infrastructure ?
so if we have to add the reference because of registering DI classes is there any best practice for loos that dependency?

It would be great if we implement a frontend SPA instead of using Swagger for everything.

Running Manga.IntegrationTests.CustomerRegistration.Register_Deposit_Withdraw_Closefails with the following stack trace:

System.IO.FileNotFoundException : Could not load file or assembly 'Microsoft.AspNetCore.Mvc, Version=2.2.0.0, Culture=neutral, PublicKeyToken=adb9793829ddae60'. The system cannot find the file specified.
Stack Trace:
at Manga.WebApi.Startup.ConfigureServices(IServiceCollection services)
--- End of stack trace from previous location where exception was thrown ---
at Microsoft.AspNetCore.Hosting.ConventionBasedStartup.ConfigureServices(IServiceCollection services)
at Microsoft.AspNetCore.Hosting.Internal.WebHost.EnsureApplicationServices()
at Microsoft.AspNetCore.Hosting.Internal.WebHost.Initialize()
at Microsoft.AspNetCore.Hosting.WebHostBuilder.Build()
at Microsoft.AspNetCore.TestHost.TestServer..ctor(IWebHostBuilder builder, IFeatureCollection featureCollection)
at Manga.IntegrationTests.CustomerRegistration..ctor() in C:\dev\code\clean-architecture-manga\manga\test\Manga.WebApi.Tests\CustomerRegistration.cs:line 33

Hi,

First of all thank you very much for this awesome work @ivanpaulovich and contributors.

I would love to see a GraphQL API co-existing with the current API.

Actually there are 2 big libraries in the .NET Core GraphQL community:

• graphql-dotnet https://github.com/graphql-dotnet/graphql-dotnet
• HotChocolate https://github.com/ChilliCream/hotchocolate

I have tested both of them and personally I do prefer HotChocolate, and even having a smaller comunnity I think they are releasing new features faster and documentation is (at least for me) much clearer.

I would love to collaborate on this trying to integrate GraphQL following the clean architecture approach.

Just let me know if you think it would be ok :)

Thanks in advance!

This is great work. Can you implement authorization and authentication best practices with JWT Authentication?

Hello Ivan,

Another question, I see that you've recently added the mediator pattern to execute use cases from controllers. By using the mediator, the controller does not directly call the execute method of the use case, so it does not "know" that the use case exists. On the other hand, what is the downside of the controller "knowing" the use case interface? I see the benefits of the mediator pattern, I just wonder what the benefits are in this case.

What is the benefit of using the mediator pattern for executing use cases?

Thanks!

We could align the Acceptance Tests with the MF naming https://martinfowler.com/articles/microservice-testing/#testing-component-introduction

What about ComponentTests?

Hi,

May be i'm missing something, but interfaces (in the C# sense of word) in this project don't have any property definitions. Could you explain the rational behind that ?

Thanks.

As I noticed, entities shouldn't be created anytime, only by the factory. The only part of an entity that should be created is the entities of the infrastructure layers.

Based on this, why entities aren't abstract?

Hello Ivan,

First of all, thank you very much this repository. Your project perfectly resembles the components that are described in the presentation of Robert C. Martin about clean architecture (https://youtu.be/Nsjsiz2A9mg).

We've set up our first application using this architecture (the version before the mediator was introduced). One of the things we're having discussions about is that it feels a bit magic that the controller executes the use case and then gets the response from the presenter. There is no coupling between the use case interface and the output port interface, which is probably by design, but makes it a bit vague to use. You have to assume that the result of the use case is written to the presenter.

Another approach that we've discussed here is to pass the output port as an argument to the execute method of the use case (considering that we don't use the mediator pattern for executing use cases). This makes explicit that the use case uses the output port to write the results, but of course, couples the use case interface to the output port interface.

What are your thoughts on this?

First off really cool repo! I was poking around in it and noticed your UseCases are quite similar to the pattern we use which revolves around MediatR. Have you considered using MediatR instead of your custom use case abstractions? Just curious if this library would play into the Clean Architecture you are trying to achieve or if it would cause any issues. Thanks!

@ivanpaulovich what do you think about using .net core 3.1 as a target for library projects as well? Some features of C # 8.0 apparently only work on .NET core 3.0/3.1. For example, NullableReferenceType doesn't work as expected. I tried myself. This link has a little more info about this: https://stu.dev/csharp8-doing-unsupported-things/.

In order to decouple Customers and Account, both aggregate should only see each other through their IDs. Please remove IAccount from Customers:

https://github.com/ivanpaulovich/clean-architecture-manga/blob/master/src/Domain/Customers/Customer.cs#L21

I'd like to simplify Dependency Injection of Modules that fallback on each other, For instance, the Entity Framework is replaced by the In Memory Fakes.

It would be nice to make Persistence module smarter to switch from Real Mode to Test Mode.

I noticed that the ValueObjects and DomainExceptions are placed in a cental folder ValueObjects. In another issue (#115) a statement was made that ValueObjects should be tied to one specific entity. Is there a case to be made to place ValueObejcts in the same namespace as their respective Entity?
When a project grows I could imagine this would make them easier to find.

I'm trying to figure out how the following problem should be solved:

I have a Bank entity.
Users of the system could be customer, manager, director or just a bank operator.
All of them will be a customer too, because to work on a bank, you should have an account.

So, should all of them implements ICustomer?
Should I have IManager which implements ICustomer? (always expanding my class, never changing it as OCP said).

Thanks guys

In order to fully harness the benefits of Domain Driven Design a CQRS / Event Sourcing implementation would be awesome. How would you approach this?

Hey Mr. Ivan!

On README.md, the links at Index of Clean Architecture Manga section are broken starting at .NET Core item.

We could injection the IMediator and Presenter through the [FromServices]. Like it is done at: https://docs.microsoft.com/en-us/aspnet/core/mvc/controllers/dependency-injection?view=aspnetcore-3.1#action-injection-with-fromservices

The documentation should be moved to the Wiki and we could add comments linking the source code to the wiki topics.

For instance the design patterns found in the project (Aggregates, Entity, Repository, UseCase) could be linked to existing wiki pages.

Hey all,

I recognise that the aggregate names are obscure, we have Users, Accounts and Customers and if we talk about them separately they seems to be the same thing.

User is the ExternalUser connected to the Authentication Credentials (Security).
Customer is the person details (Support).
Account is the the SavingsAccount or CheckingsAccount from the transactions perspective (Core).

I'd suggest we break the Domain in three Bounded Contexts:

• Core (for Accounts)
• Customers (for Customers)
• Security (for Users)

My intent is to show a real case of a modular system.

The second suggestion is to bring the Boundaries interfaces into the Bounded Contexts.

Is that a good idea?

I think I just learned something...
So, do we really need to implement IEquatable for value objects?

https://stackoverflow.com/a/32797530/2409554

A user should not be able to withdraw a negative amount and deposit a negative amount example:

Current balance: 1200
‘Withdraw -1100 makes current balance 2300’
Same with deposit.

Hello Ivan,

How should we store a password in a value object? Ideally I think the field should contain a hashed password value?

However, how to hash a password without depending on the .NET framework in order not to break the domain's responsibilities?

Also, what are the rules to be implemented as part of this implementation? For a secure password, here are some rules that could be applied.

• 12 characters minimum
• 30 characters maximum
• one number minimum
• one capital letter minimum
• A punctuation mark OR a special character (dollar, hash,...) minimum

Hi!
I've really love how this project is organized, and i've been working with this structure for a while now. But there is something that i personally don't like muc: repeat a controller for multiples usecases.
In your sample, in most of use cases we have AccountsController again. or IOutputPort. It's more organized, but i still think it's redundant.

Do you have any particular reason for this, or can help me clarifying this?

Thanks in advance!

When I see the RegisterUseCase and it depends on 3 repositories from 3 different Bounded Contexts it is unnecessary hard to understand the Register Use Case.

I would like to see the RegisterUseCase as a composition of:

• Verify User Already Registered / Register.
• Opens Savings Account and Deposit Initial Amount.
• Creates User Credentials.

If we add a Bank Entity where Bank should have a Name property, should it be the same Name of Customer?

But if the rules of Customer or Bank should have an additional implementation like: Bank Name should have only 10 characters, should we build UserName VO and BankName VO?

Although it is described on the project's main page that we're using .NET Core 3.1, the global.json file makes reference to 3.0 version. Should the version be upgraded to 3.1?

I'm bringing this issue up because support to 3.0 will end on 2020-03-03, according to Microsoft. I also believe it doesn't impact on the codebase, resulting in only updating the global.json file.

Hi guys,

we can see here how to implement must of the time a one-to-one needed.
If you need a customer, you got it. If you need a account, you got it.
But if I want to do a report with all my customers, showing the total balance for each one, how it should work?

I'm afraid about how the relation between each customer will be with his many accounts.

ps: Sorry if I'm making this wrong, it's my first time participating on github.

Why should we have an Entity like Debit on Domain project (Domain Layer) and another one inside Infrastructure with the same name (Debit) which inherits from the Debit on Domain Project?

I know it's hard to see advantages on this project yet but what are the real advantages using this approach?

As I can see, the UnitTests needs a presenter to test, and because of this, a Presenter was created on the Infrastructure layer.

Ins't this presenter a complet UnitTest responsability like the presenter of the api is?

Hi folks,

I'd like to inform that I will mention your contributions on my Clean Architecture Essentials talk.

Thank you so much for the ideas, pull requests and issues.

I will update you after the meetup 🍺

It would be nice to instantiate ValueObjects the same way we instantiate Entities. I suggest we include methods into EntityFactory to instantiate ValueObjects.

Hi,

I observed that the class CustomerRepository is with wrong name on implementation ( "CustomerReadOnlyRepository")