ThreeFive-O 665cd26ab0 Fix several problems in WritingPesterTests guideline (#9078 )

2019-03-29 15:29:41 -07:00

14 KiB

Executable file

Raw Blame History

Writing Pester Tests

Note that this document does not replace the documents found in the Pester project. This is just some quick tips and suggestions for creating Pester tests for this project. The Pester community is vibrant and active, if you have questions about Pester or creating tests, the Pester Wiki has a lot of great information. As of January 2018, PowerShell Core is using Pester version 4 which has some changes from earlier versions. See Migrating from Pester 3 to Pester 4 for more information.

When creating tests, keep the following in mind:

Tests should not be overly complicated and test too many things.
- Boil down your tests to their essence, test only what you need.
Tests should be as simple as they can.
Tests should generally not rely on any other test.

Examples

Here's the simplest of tests:

Describe "A variable can be assigned and retrieved" {
    It "Creates a variable and makes sure its value is correct" {
       $a = 1
       $a | Should -Be 1
   }
}

If you need to do type checking, that can be done as well:

Describe "One is really one" {
    It "Compare 1 to 1" {
       $a = 1
       $a | Should -Be 1
    }
    It "1 is really an int" {
       $i = 1
       $i | Should -BeOfType System.Int32
    }
}

If you are checking for proper errors, use the Should -Throw -ErrorId Pester syntax. It checks against FullyQualifiedErrorId property, which is recommended because it does not change based on culture as an error message might.

...
It "Get-Item on a nonexisting file should have error PathNotFound" {
    { Get-Item "ThisFileCannotPossiblyExist" -ErrorAction Stop } | Should -Throw -ErrorId "PathNotFound,Microsoft.PowerShell.Commands.GetItemCommand"
}

Note that if Get-Item were to succeed, the test will fail.

However, if you need to check the InnerException or other members of the ErrorRecord, you should use -PassThru parameter:

It "InnerException sample" {
   $e = { Invoke-WebRequest https://expired.badssl.com/ } | Should -Throw -ErrorId "WebCmdletWebResponseException,Microsoft.PowerShell.Commands.InvokeWebRequestCommand" -PassThru
   $e.Exception.InnerException.NativeErrorCode | Should -Be 12175
}

Describe/Context/It

For creation of PowerShell tests, the Describe block is the level of granularity suggested and one of three tags should be used: CI, Feature, or Scenario.

If the tag is not provided, the build process will fail.

Describe

Creates a logical group of tests. All Mocks and TestDrive contents defined within a Describe block are scoped to that Describe; they will no longer be present when the Describe block exits. A Describe block may contain any number of Context and It blocks.

Context

Provides logical grouping of It blocks within a single Describe block. Any Mocks defined inside a Context are removed at the end of the Context scope, as are any files or folders added to the TestDrive during the Context block's execution.

Any BeforeEach or AfterEach blocks defined inside a Context also only apply to tests within that Context.

It

The It block is intended to be used inside of a Describe or Context block. If you are familiar with the AAA pattern (Arrange-Act-Assert), the body of the It block is the appropriate location for an assert.

The convention is to assert a single expectation for each It block. The code inside of the It block should throw a terminating error if the expectation of the test is not met and thus cause the test to fail.

The name of the It block should expressively state the expectation of the test.

Admin privileges in tests

Tests that require admin privileges on Windows must be additionally marked with RequireAdminOnWindows Pester tag.

In the Azure DevOps Windows CI, we run two different passes:

The pass with exclusion of RequireAdminOnWindows tagged tests.
The pass where only RequireAdminOnWindows tagged tests are being executed.

In each case, tests are executed with appropriate privileges.

Tests that need to be run with sudo on Unix systems must be additionally marked with RequireSudoOnUnix Pester tag.

RequireSudoOnUnix tag takes precedence over all other tags like CI, Feature, etc. (which are ignored when RequireSudoOnUnix is present). Tests tagged with RequireSudoOnUnix will run as a separate pass for any Unix test.

Selected Features

Test Drive

A PSDrive is available for file activity during a test and this drive is limited to the scope of a single Describe block. The contents of the drive are cleared when a Context block is exited.

A test may need to work with file operations and validate certain types of file activities. It is usually desirable not to perform file activity tests that will produce side effects outside of an individual test.

Pester creates a PSDrive inside the user's temporary drive that is accessible via TestDrive: or $TestDrive. Pester will remove this drive after the test completes. You may use this drive to isolate the file operations of your test to a temporary store.

The following example illustrates the feature:

function Add-Footer($path, $footer) {
   Add-Content $path -Value $footer
}

Describe "Add-Footer" {
   $testPath="TestDrive:\test.txt"
   Set-Content $testPath -value "my test text."
   Add-Footer $testPath "-Footer"
   $result = Get-Content $testPath

   It "adds a footer" {
       (-join $result) | Should -BeExactly "my test text.-Footer"
   }
}

When this test completes, the contents of the TestDrive: will be removed.

Parameter Generation

$testCases = @(
    @{ a = 0; b = 1; ExpectedResult = 1 }
    @{ a = 1; b = 0; ExpectedResult = 1 }
    @{ a = 1; b = 1; ExpectedResult = 0 }
    @{ a = 0; b = 0; ExpectedResult = 0 }
    )

Describe "A test" {
    It "<a> -xor <b> should be <expectedresult>" -TestCases $testCases {
        param ($a, $b, $ExpectedResult)
        $a -xor $b | Should -Be $ExpectedResult
    }
}

Mocking

Mocks the behavior of an existing command with an alternate implementation. This creates new behavior for any existing command within the scope of a Describe or Context block. The function allows you to specify a script block that will become the command's new behavior.

The following example illustrates simple use:

Context "Get-Random is not random" {
    Mock Get-Random { return 3 }

    It "Get-Random returns 3" {
        Get-Random | Should -Be 3
    }
}

More information may be found here.

Free Code in a Describe block

Code execution in Pester can be very subtle and can cause issues when executing test code. The execution of code which lays outside of the usual code blocks may not happen as you expect. Consider the following:

Describe it {
    Write-Host -For DarkRed "Before Context"
    Context "subsection" {
        Write-Host -for DarkRed "Before BeforeAll"
        BeforeAll { write-host -for Blue "In Context BeforeAll" }
        Write-Host -for DarkRed "After BeforeAll"

        Write-Host -for DarkRed "Before AfterAll"
        AfterAll { Write-Host -for Blue "In Context AfterAll" }
        Write-Host -for DarkRed "After AfterAll"

        BeforeEach { Write-Host -for Blue "In BeforeEach" }
        AfterEach { Write-Host -for Blue "In AfterEach" }

        Write-Host -for DarkRed "Before It"
        It "should not be a surprise" {
            1 | should -Be 1
        }
        Write-Host -for DarkRed "After It"
    }
    Write-Host -for DarkRed "After Context"
    Write-Host -for DarkGreen "Before Describe BeforeAll"
    BeforeAll { Write-Host -for DarkGreen "In Describe BeforeAll" }
    AfterAll { Write-Host -for DarkGreen "In Describe AfterAll" }
}

Now, when run, you can see the execution schedule:

PS# invoke-pester c:\temp\pester.demo.tests.ps1
Describing it
In Describe BeforeAll
Before Context
   Context subsection
In Context BeforeAll
Before BeforeAll
After BeforeAll
Before AfterAll
After AfterAll
Before It
In BeforeEach
    [+] should not be a surprise 79ms
In AfterEach
After It
In Context AfterAll
After Context
Before Describe BeforeAll
In Describe AfterAll
Tests completed in 79ms
Passed: 1 Failed: 0 Skipped: 0 Pending: 0

The Describe - BeforeAll block is executed before any other code even though it was at the bottom of the Describe block. So if some state is set elsewhere in the Describe block, that state will not yet be visible (as the code will not yet been run).

Notice, too, that the BeforeAll block in Context is executed before any other code in that block. Generally, you should have code reside in one of the code block elements of BeforeAll, BeforeEach, AfterEach and/or AfterAll, especially if those blocks rely on some state set by free code elsewhere in the block.

Skipping Tests in Bulk

Sometimes it is beneficial to skip all the tests in a particular Describe block. For example, tests which are not applicable to a platform could be skipped, and they would be reported as skipped.

The following is an example of how this may be done:

Describe "Should not run these tests on non-Windows platforms" {
    BeforeAll {
        $originalDefaultParameterValues = $PSDefaultParameterValues.Clone()
        if ( ! $IsWindows ) {
            $PSDefaultParameterValues["it:skip"] = $true
        }
    }
    AfterAll {
        $global:PSDefaultParameterValues = $originalDefaultParameterValues
    }
    Context "Block 1" {
        It "This block 1 test 1" {
            1 | should -Be 1
        }
        It "This is block 1 test 2" {
            1 | should -Be 1
        }
    }
    Context "Block 2" {
        It "This block 2 test 1" {
            2 | should -Be 1
        }
        It "This is block 2 test 2" {
            2 | should -Be 1
        }
    }
}

Here is the output when run on a Linux distribution:

Describing Should not run these tests on non-Windows platforms
   Context Block 1
    [!] This block 1 test 1 691ms
    [!] This is block 1 test 2 114ms
   Context Block 2
    [!] This block 2 test 1 73ms
    [!] This is block 2 test 2 6ms

and here is the output when run on a Windows distribution:

Describing Should not run these tests on non-Windows platforms
   Context Block 1
    [+] This block 1 test 1 86ms
    [+] This is block 1 test 2 33ms
   Context Block 2
    [-] This block 2 test 1 52ms
      Expected: {1}
      But was:  {2}
      22:             2 | should -Be 1
      at <ScriptBlock>, <No file>: line 22
    [-] This is block 2 test 2 77ms
      Expected: {1}
      But was:  {2}
      25:             2 | should -Be 1
      at <ScriptBlock>, <No file>: line 25

This technique uses the $PSDefaultParameterValues feature of PowerShell to temporarily set the It block parameter -skip to true (or in the case of Windows, it is not set at all)

Multi-line strings

You may want to have a test like:

It 'tests multi-line string' {
    Get-MultiLineString | Should -Be @'
first line
second line
'@
}

There are problems with using multi-line strings with verifying the output results. The reason for it are line-ends.

They cause problems for two reasons:

They are different on different platforms (\r\n on Windows and \n on Unix).
Even on the same system, they depend on the way how the repo was cloned (local git configuration).

Particularly, in the default Azure DevOps CI Windows image, you will get \n line ends in all your files. That causes problems, because at runtime Get-MultiLineString would likely produce \r\n line ends on Windows.

Some workaround could be added, but they are sub-optimal and make reading test code harder.

function normalizeEnds([string]$text)
{
    $text -replace "`r`n?|`n", "`r`n"
}

It 'tests multi-line string' {
    normalizeEnds (Get-MultiLineString) | Should -Be (normalizeEnds @'
first line
second line
'@)
}

When appropriate, you can avoid creating multi-line strings at the first place. These commands create an array of strings:

Get-Content
Out-String -Stream

Pester Do and Don't

Do

Name your file <descriptive_test_name>.tests.ps1.
Keep tests simple:
- Test only what you need.
- Reduce dependencies.
Be sure to tag your Describe blocks based on their purpose:
- Tag CI indicates that it will be run as part of the continuous integration process. These should be unit test like, and generally take less than a second.
- Tag Feature indicates a higher level feature test (we will run these on a regular basis), for example, tests which go to remote resources, or test broader functionality.
- Tag Scenario indicates tests of integration with other features (these will be run on a less regular basis and test even broader functionality than feature tests.
Make sure that Describe/Context/It descriptions are useful.
- The error message should not be the place where you describe the test.
Use Context to group tests.
- Multiple Context blocks can help you group your test suite into logical sections.
Use BeforeAll/BeforeEach/AfterEach/AfterAll instead of custom initiators.
Use Should -Throw -ErrorId to check for expected errors.
Use -TestCases when iterating over multiple It blocks.
Use code coverage functionality where appropriate.
Use Mock functionality when you don't have your entire environment.
Avoid free code in a Describe block.
- Use BeforeAll/BeforeEach/AfterEach/AfterAll.
- See Free Code in a Describe block
Avoid creating or using test files outside of TESTDRIVE:.
- TESTDRIVE: has automatic clean-up.
Keep in mind that we are creating cross platform tests.
- Avoid using the registry.
- Avoid using COM.
Avoid being too specific about the count of a resource as these can change platform to platform.
- Example: Avoid checking for the count of loaded format files, but rather check for format data for a specific type.

Don't

Don't have too many evaluations in a single It block.
- The first Should failure will stop that block.
Don't use Should outside of an It Block.
Don't use the word "Error" or "Fail" to test a positive case.
- Example: Rephrase the negative sentence "Get-ChildItem TESTDRIVE: shouldn't fail" to the following positive case "Get-ChildItem should be able to retrieve file listing from TESTDRIVE".

14 KiB Executable file Raw Blame History