I have done fair bit research and tried all sorts of different ways of getting the Test to pass but now i need some help.
I am trying to test the following repository method:
public class PortalsRepository : BaseRepository<PortalDomainRole>, IPortalsRepository
{
public PortalsRepository(IAuthDbContext context) : base(context)
{
}
public IEnumerable<PortalRole> GetRoles(string domainName)
{
return Set.Include(x => x.PortalDomain)
.Include(x => x.PortalRole)
.Where(x => x.PortalDomain.Name.ToLower() == domainName)
.Select(x => x.PortalRole)
.ToList();
}
}
The Context looks like:
public interface IAuthDbContext : IDbContextBase
{
}
public interface IDbContextBase
{
IDbSet<T> Set<T>() where T : class;
IEnumerable<DbValidationError> GetEntityValidationErrors();
int SaveChanges();
Task<int> SaveChangesAsync();
Task<int> SaveChangesAsync(CancellationToken cancellationToken);
}
My Unit Test Set-up Looks like:
protected override void GivenThat()
{
var mockRolesSet = GetMockDbSet(PortalRoles().AsQueryable());
mockRolesSet.Setup(x => x.Include("PortalRole")).Returns(mockRolesSet.Object);
var mockDomainsSet = GetMockDbSet(PortalDomains().AsQueryable());
mockDomainsSet.Setup(x => x.Include("PortalDomain")).Returns(mockDomainsSet.Object);
var mockPortalDomanRolesSet = GetMockDbSet(PortalDomainRoles().AsQueryable());
mockPortalDomanRolesSet.Setup(x => x.Include("PortalRole")).Returns(mockPortalDomanRolesSet.Object);
mockPortalDomanRolesSet.Setup(x => x.Include("PortalDomain")).Returns(mockPortalDomanRolesSet.Object);
var customDbContextMock = new Mock<IAuthDbContext>();
customDbContextMock.Setup(x => x.Set<PortalRole>()).Returns(mockRolesSet.Object);
customDbContextMock.Setup(x => x.Set<PortalDomain>()).Returns(mockDomainsSet.Object);
customDbContextMock.Setup(x => x.Set<PortalDomainRole>()).Returns(mockPortalDomanRolesSet.Object);
ClassUnderTest = new PortalsRepository(customDbContextMock.Object);
}
My Unit Test Supporting Methods:
public List<PortalDomainRole> PortalDomainRoles()
{
var data = new List<PortalDomainRole>
{
new PortalDomainRole { PortalRoleId = 2, PortalDomainId = 1},
new PortalDomainRole { PortalRoleId = 1, PortalDomainId = 2},
new PortalDomainRole { PortalRoleId = 2, PortalDomainId = 2}
};
return data;
}
public List<PortalDomain> PortalDomains()
{
var data = new List<PortalDomain>
{
new PortalDomain { Name = "google.co.uk", PortalDomainId = 1 },
new PortalDomain { Name = "bbc.com", PortalDomainId = 2 }
};
return data;
}
public List<PortalRole> PortalRoles()
{
var data = new List<PortalRole>
{
new PortalRole {Name = "New Products", PortalRoleId = 1},
new PortalRole {Name = "Classic Products", PortalRoleId = 2}
};
return data;
}
When the unit test executes the method in question, i get:
System.NullReferenceException : Object reference not set to an instance of an object.
Most likely it does not know how to handle the nested include statements - i have followed many online questions and tutorials and now i am stuck.
My answer is probably a bit controversial, but in my experience, the best way to test your repository layer (or whatever you call the code that does the actual data access), is by having it actually call the database during testing.
When you are writing your unit test, you are assuming that Entity Framework works in a specific way. But sometimes it works in different ways than you expect, and thus a test may pass, even though the code doesn't work correctly.
Take this example, that illustrates the problem (the last EF version I worked with was version 4, but I assume that my statement is still true for EF6)
public class Foo {
public DateTime StartDate { get; set; }
public DateTime EndDate { get; set; }
public bool Active {
get { return StartDate < DateTime.Now && EndDate > DateTime.Now }
}
}
public class FooRepository {
public IEnumerable<Foo> ActiveFoos { get { return DataContext.Foos.Where(x => x.Active) } }
}
Testing this FooRepository against a mocked data access will pass, but executing against a real database will throw an exception. That is because EF will try to create an SQL clause for the Where(x => x.Active), but because Active is not a field in the database, EF will have no idea how to translate the query to SQL.
So your unit test provides a false positive. Executing the tests against the database will make it fail, as it should.
Related
I have read the document https://gist.github.com/netoisc/5d456850d79f246685fee23be2469155
which well know how to mock elasticsearch query
But I have a case which return result have innerhits
Documents = searchResult.Hits.Select(
h => {
if (h.InnerHits.TryGetValue("books", out var data)) {
ht.Source.Books = data!.Documents<book>();
}
ht.Source.Books = ht.Source.Books.Where(k=>k.country=="US");
return h.Source;
})
From my mock test
I have do this
var innerHitResult = new Mock<InnerHitsResult>();
innerHitResult.SetupGet(s => s.Documents<book>()).Returns(new List<book>());
var innerHitDictionary = new Dictionary<string, InnerHitsResult> {
{
"books", innerHitResult.Object
}
};
I've got the error on innerHitResult.SetupGet , which error say :
System.ArgumentException: Expression is not a property access: s => s.Documents<book>()
Even I use innerHitResult.Setup is not working
Can I know how to do mock for inner hit ?
In the other simple example
if I have a class :
public class BlogSearchResult
{
public InnerMetaData Hits { get; set; }
public IEnumerable<T> Document<T>() where T : class => Hits.Documents<T>();
}
I want to mock BlogSearchResult which want to do
var blogs = fixture.CreateMany<Blog>();
var blogSearch = new Mock<BlogSearchResult>();
blogSearch.SetupGet(s => s.Document<Blog>()).Returns(blogs);
or
blogSearch.Setup(s => s.Document<Blog>()).Returns(blogs);
They are all return error, is that possible ?
The requirement here is to write test cases for the Opc Ua NodeManager and it uses NodeId from Opc.Ua class.
Methods/properties in NodeId class cannot be moq because they are Non-Overridable methods and have ony get in them.
So I created a wrapper on top of NodeId class and tried to Moq that class. It works fine but now I have 2 methods/Properties
public class NodeIdTestClass : NodeId
{
public NodeIdTestClass()
{
}
public new virtual object Identifier
{
get => base.Identifier;
}
public new virtual ushort NamespaceIndex
{
get => base.NamespaceIndex;
}
public new virtual bool IsNullNodeId
{
get => base.IsNullNodeId;
}
}
//Arrange
var nodeIdMock = new Mock<NodeIdTestClass>() { CallBase = true };
nodeIdMock.Setup(x => x.Identifier).Returns(nodeIdMock.Object.Identifier);
nodeIdMock.Setup(x => x.NamespaceIndex).Returns(1);
nodeIdMock.Setup(x => x.IsNullNodeId).Returns(false);
_nodemanager.SetNamespaces(new string[] { "0", "1", "2" });
//Act
var result = _nodemanager.GetManagerHandle(nodeIdMock.Object);
The problem :
Is there something wrong with code?
Do you need a mock?
Wouldn't this be enough?
var nodeId = new NodeId(value: 17, namespaceIndex: 1);
var result = _nodemanager.GetManagerHandle(nodeId);
Could some one please help me to resolved this? i'm trying to change CustomAsync to MustAsync, but i couldn't make things to work. Below is my custom method
RuleFor(o => o).MustAsync(o => {
return CheckIdNumberAlreadyExist(o)
});
private static async Task<ValidationFailure> CheckIdNumberAlreadyExist(SaveProxyCommand command)
{
if (command.Id > 0)
return null;
using (IDbConnection connection = new SqlConnection(ConnectionSettings.LicensingConnectionString))
{
var param = new DynamicParameters();
param.Add("#idnumber", command.IdNumber);
var vehicle = await connection.QueryFirstOrDefaultAsync<dynamic>("new_checkDuplicateProxyIdNumber", param, commandType: CommandType.StoredProcedure);
return vehicle != null
? new ValidationFailure("IdNumber", "Id Number Already Exist")
: null;
}
}
To make it work with the latest version of the FluentValidation, I had to use the codes like below.
RuleFor(ws => ws).MustAsync((x, cancellation) => UserHasAccess(x)).WithMessage("User doesn't have access to perform this action");
Please notice the lambda expression here MustAsync((x, cancellation) => UserHasAccess(x)), without this I was always getting an error as cannot convert from 'method group' to 'Func<Worksheet, CancellationToken, Task<bool>>
Below is my custom UserHasAccess function.
private async Task <bool> UserHasAccess(Worksheet worksheet) {
var permissionObject = await _dataProviderService.GetItemAsync(worksheet.FileItemId);
if (permissionObject is null) return false;
if (EditAccess(permissionObject.Permission)) return true;
return false;
}
I'm assuming you're using a version of FluentValidation prior to version 6, as you're not passing in a Continuation Token, so I've based my answer on version 5.6.2.
Your example code does not compile, for starters, as you're missing a semi-colon in your actual rule. You are also evaluating two different properties on the SaveProxyCommand parameter.
I've built a very small POC based on some assumptions:
Given 2 classes:
public class SaveProxyCommand {
public int Id { get; set; }
}
public class ValidationFailure {
public string PropertyName { get; }
public string Message { get; }
public ValidationFailure(string propertyName, string message){
Message = message;
PropertyName = propertyName;
}
}
And a validator:
public class SaveProxyCommandValidator : AbstractValidator<SaveProxyCommand>{
public SaveProxyCommandValidator()
{
RuleFor(o => o).MustAsync(CheckIdNumberAlreadyExists)
.WithName("Id")
.WithState(o => new ValidationFailure(nameof(o.IdNumber), "Id Number Already Exist"));
}
private static async Task<bool> CheckIdNumberAlreadyExists(SaveProxyCommand command) {
if (command.Id > 0)
return true;
var existingIdNumbers = new[] {
1, 2, 3, 4
};
// This is a fudge, but you'd make your db call here
var isNewNumber = !(await Task.FromResult(existingIdNumbers.Contains(command.IdNumber)));
return isNewNumber;
}
}
I didn't include the call to the database, as that's not part of your problem. There are a couple of things of note here:
You're not setting the .WithName annotation method, but when you're setting up a validation rule for an object you have to do this, as FluentValidation expects you to specify specific properties to be validated by default, if you pass in an entire object it just doesn't know how to report errors back.
Must/MustAsync need to return a bool/Task<bool> instead of a custom object. To get around this, you can specify a custom state to be returned when failing validation.
You can then get access to this like this:
var sut = new SaveProxyCommand { Id = 0, IdNumber = 3 };
var validator = new SaveProxyCommandValidator();
var result = validator.ValidateAsync(sut).GetAwaiter().GetResult();
var ValidationFailures = result.Errors?.Select(s => s.CustomState).Cast<ValidationFailure>();
The above does not take into account empty collections, it's just an example of how to dig into the object graph to retrieve custom state.
As a suggestion, fluentvalidation works best if you set up individual rules per property, instead of validating the entire object. My take on this would be something like this:
public class SaveProxyCommandValidator : AbstractValidator<SaveProxyCommand>{
public SaveProxyCommandValidator()
{
RuleFor(o => o.IdNumber).MustAsync(CheckIdNumberAlreadyExists)
.Unless(o => o.Id > 0)
.WithState(o => new ValidationFailure(nameof(o.IdNumber), "Id Number Already Exist"));
}
private static async Task<bool> CheckIdNumberAlreadyExists(int numberToEvaluate) {
var existingIdNumbers = new[] {
1, 2, 3, 4
};
// This is a fudge, but you'd make your db call here
var isNewNumber = !(await Task.FromResult(existingIdNumbers.Contains(numberToEvaluate)));
return isNewNumber;
}
}
This read more like a narrative, it uses the .Unless construct to only run the rule if Id is not more than 0, and does not require the evaluation of the entire object.
I have the following three methods in the CompanyApplication class (along with the supporting factories and services listed):
public ResultSet<CompanyDto> AddCompany(CompanyDto companyDto)
{
var result = new CompanyDto();
var company = new Company();
Mapper.Map(companyDto, company);
using (ITransaction t = _transactionFactory.Create())
{
company = _companyService.Add(company);
t.Commit();
}
Mapper.Map(company, result);
return new ResultSet<CompanyDto>(1, new[] { result });
}
public ResultSet<CompanyContactDto> AddCompanyContact(CompanyContactDto companyContactDto)
{
var result = new CompanyContactDto();
var company = new Company();
var contact = new CompanyContact();
Mapper.Map(companyContactDto, contact);
using (ITransaction t = _transactionFactory.Create())
{
var contactCompanies = FindByIdJoin<Company, CompanyDto>(companyContactDto.CompanySK);
Mapper.Map(contactCompanies.Data.First(), company);
company.CompanyContacts.Add(contact);
company = _companyService.Update(company);
t.Commit();
}
Mapper.Map(contact, result);
return new ResultSet<CompanyContactDto>(1, new[] { result });
}
public ResultSet<T_DtoType> FindByIdJoin<T_DbType, T_DtoType>(long id)
{
IAbstractRepository<T_DbType> repository = EnsureRepository<T_DbType>();
T_DbType entity = repository.FindByIdJoin(id);
return (entity == null ? null : MapResultSetToDto<T_DbType, T_DtoType>(entity));
}
There are other objects in play here, which is why the FindByIdJoin has been made a separate method in the CompanyApplication class.
I have set up the testing class with some mocks and an instance of the CompanyApplication class:
private Mock<ICompanyRepository> _mockCompanyRepository;
private Mock<ICompanyDomain> _mockCompanyService;
private Mock<ITransactionFactory> _mockTransactionFactory;
private Mock<ITransaction> _mockTransaction;
private CompanyApplication _companyApplication;
[Setup]
public void SetUp()
{
_mockCompanyRepository = new Mock<ICompanyRepository>(MockBehavior.Strict);
_mockCompanyService = new Mock<ICompanyDomain>(MockBehavior.Strict);
_mockTransactionFactory = new Mock<ITransactionFactory>(MockBehavior.Strict);
_mockTransaction = new Mock<ITransaction>(MockBehavior.Strict);
_companyApplication = new CompanyApplication(
_mockCompanyRepository.Object,
_mockCompanyService.Object,
_mockTransactionFactory.Object);
}
I am successfully able to test the FindByIdJoin and AddCompany methods directly in Moq like this:
[Test]
public void CanFindCompanyByIdJoin()
{
var data = new Company {ObjectId = 1, Name = "Company1"};
_mockCompanyRepository.Setup(x => x.FindByIdJoin(It.Is<long>(arg => arg == data.ObjectId)))
.Returns(data);
var result = _companyApplication.FindByIdJoin<Company, CompanyDto>(data.ObjectId);
Assert.AreEqual(data.ObjectId, result.Data.First().ObjectId);
}
[Test]
public void CanAddCompany()
{
var data = new Company {ObjectId = 1, Name = "Company1"};
_mockCompanyService.Setup(x => x.Add(It.Is<Company>(arg => arg.ObjectId == data.ObjectId)))
.Returns(data);
_mockTransactionFactory.Setup(x => x.Create()).Returns(_mockTransaction.Object);
_mockTransaction.Setup(x => x.Commit());
_mockTransaction.Setup(x => x.Dispose());
var dto = new CompanyDto {ObjectId = 1, Name = "Company1"};
var result = _companyApplication.AddCompany(dto);
_mockCompanyService.Verify(t => t.Add(It.IsAny<Company>()));
}
Those two tests pass just fine. However, I'm having trouble coming up with a test for AddCompanyContact, because it calls FindByIdJoin as part of its flow, and that seems to be getting in the way.
Specifically, is there a way to mock var contactCompanies = FindByIdJoin<Company, CompanyDto>(companyContactDto.CompanySK) in a test for the AddCompanyContact method?
Thanks!
There is two alternatives that i see depending on the amount of work that you want to do.
Wrap that call into a object and instantiate it using a IOC container. This is the one that i feel would take the most effort if you are not using one already.
Turn that call into a Func and make a method without that parameter that does the call. This approach has the disadvantage that the top call will be untestable but will allow access to the rest of the method.
Example Below:
public ResultSet<CompanyContactDto> AddCompanyContact(CompanyContactDto companyContactDto)
{
AddCompanyContact(CompanyContactDto, ()=>
{
return FindByIdJoin<Company, CompanyDto> companyContactDto.CompanySK);
}
}
public ResultSet<CompanyContactDto> AddCompanyContact(CompanyContactDto companyContactDto, Func<WhateverTheMethodReturns> findIdReplacement)
{
var result = new CompanyContactDto();
var company = new Company();
var contact = new CompanyContact();
Mapper.Map(companyContactDto, contact);
using (ITransaction t = _transactionFactory.Create())
{
var contactCompanies = findIdReplacement();
Mapper.Map(contactCompanies.Data.First(), company);
company.CompanyContacts.Add(contact);
company = _companyService.Update(company);
t.Commit();
}
Mapper.Map(contact, result);
return new ResultSet<CompanyContactDto>(1, new[] { result });
}
I was over complicating the problem... since AddCompanyContact calls FindByIdJoin, all I needed to do was mock the same interface that is used in FindByIdJoin.
Lesson learned: mock interfaces, not classes.
I have stumbled upon a riddle which I can't explain, maybe someone here will be able to.
Here is a (rather lengthy but complete) code snippet:
public class Foo
{
public int ID { get; set; }
public string Name { get; set; }
public virtual ICollection<Foo> InnerFoo { get; set; }
}
public class AppContext : DbContext
{
public IDbSet<Foo> Foos { get; set; }
}
public class Initializer : DropCreateDatabaseAlways<AppContext>
{
protected override void Seed(AppContext context)
{
var list = new List<Foo>
{
new Foo {Name = "one", InnerFoo = new List<Foo>{new Foo {Name = "Four"}}},
new Foo {Name = "two"},
new Foo {Name = "three"},
};
list.ForEach(f => context.Foos.Add(f));
}
}
public class Filter
{
public static Expression<Func<Foo, bool>> GetPredicate()
{
return p => p.Name != null && p.Name.Length > 3;
}
}
class Program
{
static void Main(string[] args)
{
Database.SetInitializer(new Initializer());
using (var ctx = new AppContext())
{
var predicate = Filter.GetPredicate();
var list = ctx.Foos.Where(f => f.InnerFoo.AsQueryable().Where(predicate).Count() > 0).ToList(); // this works
// var list = ctx.Foos.Where(f => f.InnerFoo.AsQueryable().Where(Filter.GetPredicate()).Count() > 0).ToList(); // this doesn't
foreach (var s in list)
{
Console.WriteLine(s.Name);
}
}
}
}
The line that's commented out doesn't work - throws an exception at runtime - "Internal .NET Framework Data Provider error 1025.". I'm using EntityFramework.4.1.10715.0
Can anyone tell me why?
Side question: I'm looking for a way to keep filtering expressions that are used in several different classes.
The problem is that your inner Where is already within the context of an "outer Where" on ctx.Foos - so the call to Filter.GetPredicate() ends up as part of the expression tree, and the Entity Framework has no idea what it means or how to translate it into SQL.
That's why it's happening... I'm not sure of the best solution right now though, unless you can extract the predicate into a separate variable where you need it.
(As an aside, it's generally more expressive to use Any(...) than ...Count() > 0 - and in LINQ to Objects it can make a huge difference.)
You hinted towards this yourself already, but just to make it clear to future readers.
You can use the function to generate the predicate as you're doing, but you need to store the expression in an intermediate expression within your function.
i.e. change:
var list = ctx.Foos.Where(f => f.InnerFoo.AsQueryable().Where(Filter.GetPredicate()).Count() > 0).ToList();
into:
var pred = Filter.GetPredicate();
var list = ctx.Foos.Where(f => f.InnerFoo.AsQueryable().Where(pred).Count() > 0).ToList();
I'm still puzzled as to why this happens this way.