DEV Community: Harish Babu

Using Domain Types with NHibernate

Harish Babu — Sun, 21 Nov 2021 22:06:21 +0000

Primitive obsession is a code smell. The proposal is to replace strings, and other primitive data types with domain types. For example, user's id in system can be represented by an integer or Guid (UUID), or string (username). However, there are benefits that come along with creating a type called UserId using that in place of primitive types.

One advantage is that you cannot accidentally call the following method with another integer or Guid. You must do the right thing otherwise code will not compile.

public void AcceptedBy(UserId id)
{ ... }

One of the hurdles you will have when replacing this code smell is that now, there might be places in the website where you need to display the integer value or include the Guid in the link to a user's profile. On the other end, you might have to store the user's Id in to some database or make an API call with it.

The first instinct might be to translate the UserId object to the type that's required. Instead, stop the urge to map / translate the type and use the same type through out.

Most of the Entity Relational Object mappers allows you to specify how to translate from an domain type to a database type. For NHibernate, there are a couple of things that needs to be setup.

First, the mapping needs to specify the database type and where to store the value.

public class UserMapping : ClassMap<User>
{
    public UserMapping()
    {
        Id(x => x.Id)
          .CustomType<string>.Access.CamelCaseField(Prefix.None);
    }
}

The Access configuration tells NHibernate to look for a private field on the class User that has the camel cased name of the property and is not prefixed. Therefore, NHibernate will then look for a field with name id in class User.

public class User
{
    string id;
    public virtual UserId Id
    {
        get
        {
            UserId.TryParse(id, out var result);
            return result;
        }
        set => id = value?.Value;
    }
}

The Id property on User class will return the parse value of the string id and set the field appropriately from UserId object. In the example, I expose a get only property for the string. With the mapping and the private field usage, NHibernate will be able to store domain type into and retrieve from database.

I would have preferred if I didn't have the need to expose Value property. However, the creation is the UserId still needs to happen through the constructor. So, its OK for now.

For using the Id property on LINQ queries however with NHibernate, you will need to do one more change.

var userById = session.Query<User>(u => u.Id == new UserId("chekkan"));

The above query will fail with an exception when trying to convert UserId type to string. What I gathered is happening behind the scene is that NHibernate was trying to covert the right hand side of the expression to string because Id column in User table is a nvarchar type. But, UserId cannot be converted to string.

If we try to compare Id to UserId.Value, that's not going to compile.

var userById = session.Query<User>(u => u.Id == new UserId("chekkan").Value);

If we were to rewrite the query in hql however, it will work.

var userToFind = new UserId("chekkan");
var query = session.CreateQuery("from User u where u.Id = :Id");
query.SetParameter("Id", userToFind.Value);
var userById = query.List<User>();

But, its not type safe. We want the LINQqueries to work. For that, we will create an implicit coverter from UserId to string in our UserId type.

public static implicit operator string(UserId id) => id.Value;

With this change, the above LINQ query that was not compiling before will start to compile.

// with implicit string converter
var userById = session.Query<User>(u => u.Id == new UserId("chekkan").Value);

// with explicit string converter
var userById = session.Query<User>(u => 
    (string) u.Id == new UserId("chekkan").Value
);

MiniProfiler for dotnet with NHibernate MSSQL Driver

Harish Babu — Fri, 05 Nov 2021 22:50:08 +0000

I first came across MiniProfiler when I started a new ruby on rails project. It came built in and I found it very useful to see the time requests spend and a breakdown of it. I think it was only visible on development environment.

I found myself asking the question of whether it was available for ASP.NET web application. Turns out, MiniProfiler was written by the folks at stack overflow. And their primary technology stack is .NET. The library supports both full .NET Framework and the new .NET core. Their existing documentation for ASP.NET integration is easy to follow and they even have a samplesrepository.

One thing that was missing was NuGet package for NHibernate integration. I wanted to log SQL queries that were getting executed by NHibernate ORM library. I saw that there was couple of NuGet packages published by community members. But, they were not actively worked on by the maintainers.

The latest version for MiniProfiler.Mvc5 NuGet package at the time of writing is 4.2.22, which I've installed to my Website project. Following the instructions for ASP.NET MVC 5, my Global.asax.cs file looks like below.

public class MvcApplication : System.Web.HttpApplication
{
  protected void Application_Start()
  {
    ...
    InitProfilerSettings();
  }

  protected void Application_BeginRequest()
  {
    if (Request.IsLocal) MiniProfiler.StartNew();
  }

  protected void Application_EndRequest() => MiniProfiler.Current?.Stop();

  private static void InitProfilerSettings()
  {
    MiniProfiler.Configure(new MiniProfilerOptions
      {
        RouteBasePath = "~/profiler",
      }
      .AddViewProfiling() // Add MVC view profiling
    );
  }
}

I've went ahead and did the changes suggested for both the _Layout.cshtml file and Web.config file.

I've installed FluentNHibernate(3.1.0) and NHibernate(5.3.10) packages to the project. I have an MSSQL Server and have a table called employees and will query for all employees and render them in my view. I would like to view the SQL that NHibernate produced and the time it took in my mini profiler.

public class HomeController : Controller
{
  public async Task<ActionResult> Index()
  {
    using (var session = NHibernateHelper.OpenSession())
    {
      var employees = await session.Query<Employee>().ToListAsync();
      return View(employees);
    }
  }
}

And rendering the list in my index view

<div class="row">
    <ul>
        @foreach (var employee in @Model)
        {
            <li>@employee.Name</li>
        }
    </ul>
</div>

Shows me MiniProfiler without any SQL information.

To report SQL statements to MiniProfiler, we can use one of the SQL wrapper classes.

public class MiniProfiledSqlClientDriver : SqlClientDriver
{
  public override DbCommand CreateCommand()
  {
    var dbCommand = base.CreateCommand();
    if (MiniProfiler.Current != null)
    {
      dbCommand = new ProfiledDbCommand(
                        dbCommand, null, MiniProfiler.Current
                      );
    }
    return dbCommand;
  }
}

I have opted to use the ProfiledDbCommand wrapper as is the approach by MRCollective/MiniProfiler.NHibernate repo. Notice that I am inheriting from SqlClientDriver.

I have then registered the MiniProfiledSqlClientDriver with FluentNHibernate.

public static class NHibernateHelper
{
  private static ISessionFactory _sessionFactory;

  private static ISessionFactory SessionFactory
  {
    get
    {
      if (_sessionFactory != null) return _sessionFactory;
        _sessionFactory = Fluently.Configure()
          .Database(MsSqlConfiguration
            .MsSql2012.ConnectionString(c =>
              c.FromConnectionStringWithKey("Default"))
            .Driver<MiniProfiledSqlClientDriver>())
            .Mappings(m => m.AutoMappings
              .Add(AutoMap.AssemblyOf<Employee>(new StoreConfiguration())))
            .BuildSessionFactory();
          return _sessionFactory;
    }
  }

  public static ISession OpenSession() => SessionFactory.OpenSession();
}

With these changes, I've now SQL statements available in MiniProfiler.

It also works for inserts

Summary

I wasn't expecting the insert to work with the code I've added. When working with an older version of NHibernate library, I had to also implement ProfiledSqlClientBatchingBatcherFactory, inherit MiniProfiledSqlClientDriver from IEmbeddedBatcherFactoryProvider, and implement IEmbeddedBatcherFactoryProvider.BatcherFactoryClass as done in MRCollective/MiniProfiler.NHibernate's implementation.

Repo accompanying this blog post can be found at my github repository.

Ruby on Rails GitHub Action Workflow

Harish Babu — Fri, 23 Apr 2021 22:36:43 +0000

The goal is to start running tests that is part of my code base with each push to main and also any pull requests to main branch. This blog post will not cover any linting or deployment of the ruby on rails web application. The final yaml is tested with ruby version 3.0.0 and ruby on rails version 6.1.3.1 at the time of writing. Skip to end of the post to view the finished yaml.

We will start with recommended ruby starter workflow yaml file from GitHub actions page. The starter file provides you with a matrix with various ruby versions. Because, the workflow I am setting up is for a web application that will be deployed to an environment with a specific ruby version install, I will delete the strategy section completely and also specify my ruby version to 3.0.

runs-on: ubuntu-latest
steps:
    - uses: actions/checkout@v2
    - name: Set up Ruby
      uses: ruby/setup-ruby@473e4d8fe5dd94ee328fdfca9f8c9c7afc9dae5e
      with:
        ruby-version: 3.0
        bundler-cache: true # runs 'bundle install' and caches installed gems automatically

I am planning on deploying the application in a docker container running on linux. Therefore, I've set my build environment to be an ubuntu base image. If you are developing on something other than linux (which is my case), you will also need to add support for linux with your bundle.

bundle lock --add-platform x86_64-linux

After the ruby setup step, the next step thats already provided is the test step with bundle rake command. I have my project setup to run with mysql database. Therefore, we need to define a service for mysql for the tests to pass. Read more about service containers at G itHub documentation site.

  test:
    ...
    services:
      # Label used to access the service container
      mysql:
        # Docker Hub image and tag
        image: mysql:8.0
        env:
          MYSQL_ROOT_PASSWORD: my-secret-password
          MYSQL_DATABASE: <db_name> # replace with actual name
          MYSQL_USER: test_user
          MYSQL_PASSWORD: password
        # Set health checks to wait until mysql has started
        options: >-
          --health-cmd "mysqladmin ping -h localhost -u root -p$$MYSQL_ROOT_PASSWORD"
          --health-interval 10s
          --health-timeout 5s
          --health-retries 5

    steps:
      ...

Now a container will be spun up my GitHub action and made available on the port you specified. The options section specifies that the service is only ready when mysqladmin ping command succeeds. The next step is to setup the database required for your tests. To do that, we will make use of the DATABASE_URL environment variable and the bin/rails db:setup command. Make sure to use the prefix bin. You can read more about configuration and order of preference at ruby on rails documentation site.

  - name: Setup test database
      run: bin/rails db:setup
      env:
        DATABASE_URL: mysql2://test_user:password@127.0.0.1:3306/<db_name>

Make sure that DATABASE_URL environment variable starts with mysql2. For some reason, I can't seem to use localhost for the hostname part. If someone could point me to the reason, that will be much appreciated 🙏.

After the database is setup, you might also have to make sure that webpacker is installed. Refer to the documentation for up to date information.

  - name: Install Webpacker
    run: bin/rails webpacker:install

The final yaml file...

name: Ruby

on:
  push:
    branches: [main]
  pull_request:
    branches: [main]

jobs:
  test:

    runs-on: ubuntu-latest

    services:
      # Label used to access the service container
      mysql:
        # Docker Hub image and tag
        image: mysql:8.0
        ports:
          - 3306:3306
        env:
          MYSQL_ROOT_PASSWORD: my-secret-password
          MYSQL_DATABASE: <db_name> #replace with name
          MYSQL_USER: test_user
          MYSQL_PASSWORD: password
        # Set health checks to wait until mysql has started
        options: >-
          --health-cmd "mysqladmin ping -h localhost -u root -p$$MYSQL_ROOT_PASSWORD"
          --health-interval 10s
          --health-timeout 5s
          --health-retries 5

    steps:
    - uses: actions/checkout@v2
    - name: Set up Ruby
      uses: ruby/setup-ruby@473e4d8fe5dd94ee328fdfca9f8c9c7afc9dae5e
      with:
        ruby-version: 3.0
        bundler-cache: true # runs 'bundle install' and caches installed gems automatically

    - name: Setup test database
      run: bin/rails db:setup
      env:
        DATABASE_URL: mysql2://test_user:password@127.0.0.1:3306/<db_name>

    - name: Install Webpacker
      run: bin/rails webpacker:install

    - name: Run tests
      run: bundle exec rake

Update blog to ghost version 4 running in Kubernetes

Harish Babu — Sat, 10 Apr 2021 21:11:25 +0000

My personal blog site is running on the ghost blogging platform at the time of this writing. Its current running version 3.42.x and there was a new major version released recently and its currently on version 4.2.0. There are guides available on ghost documentation site to help make the upgrade when you have installed ghost on a server using the ghost-CLI tooling. However, none exists for ghost running on Kubernetes, or docker containers. ghost-cli update path seemed desirable compared to the clean install option especially because some database migration might be involved I assumed. Also, I didnt want to re-configure my site with google analytics etc.

Follow the initial steps of backing up all the important content as mentioned in the documentation site. Then come back here...

You will have ssh into the docker container in order to copy the content folder across.

kubectl cp <namespace>/<pod_name>:/var/lib/ghost/content ghost-migration-to-4.x/content

I didn't have much luck copying the theme folder in content. But, I haven't made any changes to the theme using the default Casper theme at present. Make sure run the upgrade check for your theme before progressing any further. Download your existing theme from the Admin site and upload the zipped file into the GScan website.

The way I have the blog setup in Kubernetes is by using the base image chekkan/ghost-cloudinary which builds off the ghost-alpine image. I have already gone ahead and published version 4.2.0 of the docker image.

You will need to ssh into the docker container pod and install ghost-cli.

kubectl exec --stdin --tty -n "<namespace>" "<pod_name>" -- /bin/bash

ssh into the pod thats running ghost container

npm install -g ghost-cli@latest

install ghost-cli npm package

Make sure your current working directory is where you've installed ghost. For me, its at /var/lib/ghost.

I used environment variables in deployment spec file together with Kubernetes secrets to configure my database credentials. For some reason, these were not picked up when I ran ghost config get database.connection.host command. So, I decided to configure them again manually.

ghost config --db mysql --dbhost <dbhost> --dbuser <dbuser> \
  --dbpass <dbpass> --dbport <dbport> --dbname <dbname>

Update the ghost config values for database

Running the above command will update the config.production.json file. Review the file to make sure its got the expected values.

su node

switch user to node

ghost-cli stops you from updating as a root user. If you wanted to get back to being a root user again, exit.

Before you can update to version 4, ghost wants you to be in the latest version of the currently installed ghost version.

ghost update v3

updates ghost to the latest version of major version 3

ghost update

updates ghost to the latest version; version 4.2.0 at the time of writing

Once ghost is updated, go ahead and kubectl apply with your deployment spec file thats updated to the same version of ghost.

Faking .NET Framework ConfigurationSection for Unit Tests

Harish Babu — Sat, 30 Jan 2021 21:44:19 +0000

Many are now familiar with using typed configuration that's available in .net core and .net 5 with the help from Options pattern. However, if you are working on a projects targeting .NET Framework, you will know ConfigurationSection from System.Configuration assembly.

ConfigurationSection allows you to group related configurations together in xml. Lets take an example of providing some simple configuring settings to a retry client.

<?xml version="1.0" encoding="utf-8"?>
<configuration>
  <configSections>
    <section name="retrySettings" type="ConfigSectionTest.RetrySettings, ConfigSectionTest" />
  </configSections>
  <retrySettings count="3" delay="00:00:30" failureThreshold="0.5" />
</configuration>

Notice the configSections.section node where we are naming the config section and also pointing to the type where its implemented. retrySettings xml node has various attributes and values assigned for each.

The RetrySettings class is implemented as...

public class RetrySettings : ConfigurationSection
{
  [ConfigurationProperty("count")]
  public int RetryCount => (int)this["count"];

  [ConfigurationProperty("delay", IsRequired = true)]
  public TimeSpan RetryDelay => (TimeSpan)this["delay"];

  [ConfigurationProperty("failureThreshold")]
  public decimal FailureThreshold => (decimal)this["failureThreshold"];
}

With expression bodied property syntax, the class is fairly small. I've specified RetryDelay to be required using the ConfigurationProperty attribute.

In order to use the configuration section RetrySettings, you'd do so by calling ConfigurationManager.

ConfigurationManager.GetSection("retrySettings") as RetrySettings;

Even though you can create an instance of RetrySettings, it is not possible set the values of the properties without modifying the class (providing a setter). You will be able to mock the class by introducing an interface or using an actual appsettings.config file.

However, it maybe easier to create a fake retry settings config section instead of mocking. In order to do this, you can inherit the ConfigurationSection class which gives you access to the DeserializeElement method. This method takes in an xml string which is your section node. The second argument indicates whether to serialize only the collection key property.

class FakeRetrySettings : RetrySettings
{
  public void PopulateConfig(string xmlString)
  {
    using (XmlReader reader = new XmlTextReader(new StringReader(xmlString)))
    {
      reader.MoveToContent();
      DeserializeElement(reader, serializeCollectionKey: false);
    }
  }
}

You can now configure the retry settings as needed to orchestrate your system under test.

var retrySettings = new FakeRetrySettings();
const string xmlString = "<retrySettings count=\"3\" delay=\"00:05:00\" />";
retrySettings.PopulateConfig(xmlString);

You can make it much more easy to work with the RetrySettings class in test project by having a builder class.

class RetrySettingsBuilder
{
  private readonly Dictionary<string, string> properties = new Dictionary<string, string>
  {
    { "count", "3" },
    { "delay", "00:30:00" },
    { "failureThreshold", "0.5" }
  };

  public RetrySettingsBuilder WithProperty(string key, string url)
  {
    if (!properties.ContainsKey(key))
      throw new ArgumentException(nameof(key));

    properties[key] = url;

    return this;
  }

  public RetrySettings Build()
  {
    var xmlString = "<retrySettings ";
    var pairs = properties.Select(pair => $"{pair.Key}=\"{pair.Value}\"");
    xmlString += string.Join(" ", pairs) + " />";

    var settings = new FakeRetrySettings();
    settings.PopulateConfig(xmlString);
    return settings;
  }
}

Builder has some default values already configured. Therefore, if you wanted the test to work and you don't worry about the values, you can create a new instance of the configuration section by calling new RetrySettingsBuilder().Build();. You can override the default value by calling WithProperty method e.g. new RetrySettingsBuilder().WithProperty("count", "2").Build();.