Introduction

This document describes the specific testing practices used by the ODS Platform team in maintaining the Ed-Fi ODS / API and its associated utilities (e.g., code generation, migration utility, database deploy).

For more context on Ed-Fi testing practices, see Testing Standards.

Unit Testing

Unit Test Toolkit

• Test framework: NUnit3.
• Mocking framework: FakeItEasy.
• Assertion library: Shouldly.

Unit Test Base Classes

The ODS solution, as well as related NuGet packages and utility applications, share a common project reference to EdFi.TestFixture, which contains valuable helpers described below.

ScenarioFor

This is a base class that aids in construction of tests with automatic dependency injection. It is present primarily for legacy purposes and is not generally used for new tests.

TestDbSet

This class inherits from DbSet  and serves as a test double. TestDbSet is used with testing repository classes that would otherwise use Entity Framework directly.

TestFixtureBase

Most unit test fixtures in the Ed-Fi-ODS  and Ed-Fi-ODS-Implementation repositories inherit from this class. The class provides lightweight scaffolding that aids in structuring low-maintenance test fixtures. The following code sample highlights some of the features. All comments in the example have been added in Confluence, and do not exist in the original source.

// Create an outer "container" for all tests that belong to a particular system.
[TestFixture]
[SuppressMessage("ReSharper", "InconsistentNaming")]
public class ApiConfigurationProviderTests
{
	// Create an inner class class whose name describes the action being tested; 
	// it should inherit from TestFixtureBase.
    public class When_getting_a_sandbox_api_mode : TestFixtureBase
    {
		// Variables and properties can go here, with appropriate scope. Example:
        private IConfigConnectionStringsProvider _configConnectionStringsProvider;
        private IConfigValueProvider _configValueProvider;
        private IDatabaseEngineProvider _databaseEngineProvider;

		// Commonly there are variables here to store the system under test
		// and the test result:
        private ApiConfigurationProvider _systemUnderTest;
        private ApiMode _result;

		// Implement the Arrange() method, which the base class will call before 
		// execution of each and every test.
        protected override void Arrange()
        {
			// This section will include initial values, creation of fakes, and 
			// setup of those fakes to inject expected responses (mocks)
            _configValueProvider = A.Fake<IConfigValueProvider>();
            _configConnectionStringsProvider = A.Fake<IConfigConnectionStringsProvider>();
            _databaseEngineProvider = A.Fake<IDatabaseEngineProvider>();

            A.CallTo(() => _configValueProvider.GetValue(ApiConfigurationConstants.ApiStartupType))
                .Returns(ApiConfigurationConstants.Sandbox);

            A.CallTo(() => _configConnectionStringsProvider.ConnectionStringProviderByName)
                .Returns(new Dictionary<string, string> {{"db", ApiConfigurationConstants.SqlServerProviderName}});

			// Commonly, `_systemUnderTest` is initialized using constructor injection 
			// to insert the fake dependencies created above.
            _systemUnderTest = new ApiConfigurationProvider(_configValueProvider, _databaseEngineProvider);
        }

		// Implement the Act() method, which will perform the action whose outcome is 
		// being tested.
        protected override void Act()
        {
            // If the method under test has a return value, that value is commonly stored 
			// in an instance variable defined above, e.g. `_result`.
            _result = _systemUnderTest.Mode;
        }

		// The following two [Test] cases illustrate the rule of "one assertion per test".
        [Test]
        public void Should_not_be_year_specific()
        {
            _systemUnderTest.IsYearSpecific()
                .ShouldBe(false);
        }

        [Test]
        public void Should_parse_the_api_mode_with_success()
        {
            _result.ShouldBe(ApiMode.Sandbox);
        }
    }

	// Here we have a second inner class to test functionality under different conditions.
    public class When_getting_a_year_specific_api_mode : TestFixtureBase
    {
		// Member properties and instance variables...

        protected override void Arrange()
        {
            // Once-per-test setup
        }

        protected override void Act()
        {
            // Call the system under test
        }

		// These two test are similar to the class above this, but with different expected 
		// results due to differing input values in the setup process.
        [Test]
        public void Should_be_year_specific()
        {
            _systemUnderTest.IsYearSpecific()
                .ShouldBe(true);
        }

        [Test]
        public void Should_parse_the_api_mode_with_success()
        {
            _result.ShouldBe(ApiMode.YearSpecific);
        }
    }

	// Add more test classes as needed to cover the functionality in the system under test.
}

Integration Testing with Postman

Integration Testing Toolkit

ODS / API integration tests use Postman. The following notes have been helpful for the development team during the adoption of Postman in 2019.

Best Practices for Establishing Data for Tests

• Do not use hard-coded values as this makes the tests rigid.
• Interactively inspect the sample ODS database to ensure the data present are adequate for testing targeted functionality.
  • If necessary data exists, then add GET requests to initialize any needed "known" values (e.g., resource Ids, UniqueIds). Save the known values as environment variables (using pm.environment.set ) for subsequent use. Follow the guidelines on Postman variable usage (see below).
  • If necessary data exists but is difficult to locate through the API, add POST requests as necessary.
  • If necessary data is not present, add POST requests as necessary to initialize data in the ODS.
• Make sure all data initialized by POST requests have tests that check the API's response to make sure data is configured as expected (e.g., ensure that requests to create new data return 201 - Created ).
• For verification, write test scripts to be as flexible as possible in verifying targeted functionality.
  • For example, for Composite resource tests, a GET request is used to retrieve a page of data which is then processed in its entirety to ensure that the expected structure of the composite definition is found somewhere (as opposed to everywhere) in the response.
• When creating new data in PUT and POST requests, omit the following elements from the body:
• The link  object on references. This is not used for writing.
• The id  property
• The _etag  property
• All null  values. This just adds noise to the body for maintenance. Keep it clean.

• Do not include any optional values that don't add meaning to the test. For example, some requests in the test have empty strings assigned to properties that are optional and could just be omitted for the test. Including those values can become an unnecessary distraction for future maintainers of these tests because they are not relevant to the functionality being tested. Here's an example, from a Student POST request:

  Code Block
  ... "firstName": "{{supplied:firstName}}", "generationCodeSuffix": "", "identificationDocuments": [ ], "lastSurname": "{{supplied:lastSurname}}", "maidenName": "", "middleName": "", ...

  
  

Environment Variable Naming / Usage

• The names of environment variables should use the following format: (known|supplied)[:scenarioId]:name 
  • Use a known:  prefix for variables that hold values that are "known" from the ODS database. Examples:
    • known:localEducationAgencyId. The known Local Education Agency ID.
    • known:{scenarioId}:studentUniqueId. A known student Unique ID that is the focus of a particular test scenario.
  • Use a supplied: prefix for variables that hold values defined by the test. Examples:
    • supplied:{scenarioId}:birthDate. An arbitrary value created by the test (e.g., for use in creating a new Student or Staff resource).
    • supplied:{scenarioId}:studentUniqueId. An arbitrary Unique ID value created by the test (also, e.g., for use in creating a new Student or Staff resource).
  • The "scenarioId" portion is optional, but enables the variable to be scoped for a specific scenario (which is something that Postman doesn't directly support).

    • Our current approach has been to copy the following boilerplate to the "Pre-request Script" of an initialization request:

      Code Block

      const uuid = require('uuid'); function newGuid() { return uuid.v4().toString().replace(/[^a-zA-Z0-9 ]/g,""); } function createScenarioId() { return newGuid().substring(0,5); }   // Set the scenarioId for the current scenario pm.environment.set('scenarioId', createScenarioId());   // Use the scenarioId to store values for the current scenario const scenarioId = pm.environment.get('scenarioId'); pm.environment.set('supplied:'+scenarioId+':studentUniqueId',  newGuid()); // ... or use string interpolation pm.environment.set(`supplied:${scenarioId}:studentUniqueId`,  newGuid());

      
      

  • The "name" portion should reflect the nature of the data the variable represents.

  • We have removed all usages of pm.variables.set  because this sets transient variables that only exist for the duration of the current execution run. When using the Collection Runner, the variable will be available throughout the current run, but when executing tests interactively, the variable will only exist for the duration of the current request.

    The issue is that pm.variables.get will resolve variables by name all the way up a chain of priority: 

    • Global
    • Environment
    • Collection
    • Local

    We encountered a situation where there was an identically named variable stored in the environment. During interactive request executions, the value being used was from the environment variable, but during Collection Runner test runs the value set by pm.variables.set was used. This caused the tests to fail during interactive execution which was deemed to be an undesirable characteristic of the approach. Thus, the approach of naming variables such that they are scoped using a middle "scenarioId" segment was taken.

• Add a request at the bottom of the collection that cleans up all environment variables. This can be used when running tests interactively to prevent bleed over from one scenario to another. For example, you might have a GET request on {{ApiBaseUrl}} with the following Pre-request script:

  Code Block

  const __ = require('lodash');   const keys = __.keys(pm.environment.toObject()); console.log('Initial keys: ' + JSON.stringify(keys));   const keysToRemove = __.filter(keys, x => __.startsWith(x, 'known:') || __.startsWith(x, 'supplied:'));   __.each(keysToRemove, k => pm.environment.unset(k));   const remainingKeys = __.keys(pm.environment.toObject()); console.log('Remaining keys:' + JSON.stringify(remainingKeys));