Updated all Pester tests in the project to support Pester v5. The main benefit that we get out of this, besides just moving on to the currently supported version of the test framework, is that we can now execute individual tests directly within VS Code.
In the process of doing this migration:
* Removed repository Code of Conduct tests from GitHubMiscellaneous.tests.ps1 since the command has been deprecated by GitHub
* Updated GitHubReleases.tests.ps1 to account for pre-releases
* Updated run-unitTests.yaml and CONTRIBUTING.md to have the Pester v5.3.3 install command, as well as to update its invocation commands.
* Removed the language-specific error messages in some `-Throws` tests because the wording keeps changing.
* Updated this project's VSCode settings.json to take advantage of the fact that the tests are now Pester 5 compatible. This allows each individual block to have a Run/Debug link associated with it. ([More info](https://pester.dev/docs/usage/vscode)).
* Disabled all of the Projects/ProjectCards/ProjectColumns tests because Classic Projects have been deprecated by GitHub (#380 for more info)
Fixes#194
Co-authored-by: Howard Wolosky <HowardWolosky@users.noreply.github.com>
All state-changing commands (with the exception of `New-*`) are now silent by default. Users can pass-in `-PassThru` to make them return the result that used to be returned by default.
Users can revert back to the previous behavior by leveraging the new configuration value: `DefaultPassThru`.
Resolves#217
# Description
Comprehensively adds pipeline support throughout the entire module
This change required examining every method, and in some cases a few breaking changes (noted below) had to be introduced in order to make support work consistently end-to-end. UT's were added for every function, which resuled in a few bugs that were identified and fixed (also noted below).
## Breaking changes
### Unavoidable Breaks
* `Get-GitHubRepositoryBranch`: `Name` parameter is now `BranchName`.
> A `GitHub.Repository` object has a `name` property which is the name of the repo. Piping in the repo object to this method would confuse it, making it search for a branch with the name of the repo, as opposed to the desired effect of just listing out all branches in the repo.
* `*-GitHub*Label`: `Name` parameter is now `Label`.
> Similar to above.
> `Name` was too generic and was causing unintended conflicts when passing in certain objects over the pipeline (like a `GitHub.Repository`) which also has a `name` property, making it impossible to pipe in a repository to list all labels (since it was trying to query for a label in that repository named the same as the repo.
* `*-GitHubLabel`: `Milestone` parameter is now `MilestoneNumber`.
> A `GitHub.Issue` contains a `milestone` object property, and PowerShell was complaining that there was no way to convert that to an `[int64]` when an Issue was passed in via the pipeline. The only way to avoid this was to use `MilestoneNumber` which is the name of the property we add to `GitHub.Milestone` objects, and it's what we alias to in all other methods that interact with milestones.
* `*-GitHubIssue`: `Milestone` parameter is now `MilestoneNumber`.
> A `GitHub.Issue` contains a `milestone` object property, and PowerShell was complaining that there was no way to convert that to an `[int64]` when an Issue was passed in via the pipeline. The only way to avoid this was to use `MilestoneNumber` which is the name of the property we add to `GitHub.Milestone` objects, and it's what we alias to in all other methods that interact with milestones.
* `Get-GitHubLicense`: `Name` parameter is now `Key`.
> The `key` is what you can query against, but a License object has _both_ a `name` and a `key` property, which caused piped object queries to fail with the existing parameter name.
* `Get-GitHubCodeOfConduct`: `Name` parameter is now `Key`.
> Similar to above.
> The `key` is what you can query against, but a Code of Conduct object has _both_ a `name` and a `key` property, which caused piped object queries to fail with the existing parameter name.
* `New-ProjectCard`: Signature has changed. Instead of specifying `ContentId` and `ContentType`, you now just directly specify either `IssueId` or `PullRequestId`.
> Pipeline input would not have worked without this change. Additionally, this simply provides a cleaner interface for users in general, requiring one less input.
* `Get-GitHubUserContextualInformation`: Signature has changed. Instead of specifying `SubjectId` and `SubjectType`, you now just directly specify either `OrganizationId`, `RepositoryId`, `IssueId` or `PullRequestId`.
> Similar to above.
> Pipeline input would not have worked without this change. Additionally, this simply provides a cleaner interface for users in general, requiring one less input.
### Breaks With Workarounds
* A number of changes to the Comments functions
* `GitHubComments.ps1` was renamed to `GitHubIssueComments.ps1` given that there are 5 different types of comments, these functions focus on `Issue` comments, and there doesn't currently appear to be a path forward for creating a single unified API set for all comment types.
* All of these methods were renamed from `*-GitHubComment` to `*-GitHubIssueComment`, however they were also aliased back to their previous names (for now). You should really migrate to these new names however.
* The main parameter was renamed from `CommentId` to `Comment`, however it was aliased back to `CommentId`.
* `*-GitHubProject`: `ArchivedState` parameter is now `State` (but aliased back to `ArchivedState` to help with migration).
* `*-GitHubProject`: `Name` parameter is now `ColumnName` (but aliased back to `Name`)
* `*-GitHubProjectColumn`: `Name` parameter is now `ColumnName` (but aliased back to `Name`)
* `Move-GitHubProjectCard`: `ColumnId` parameter is now `Column` (but aliased back to `ColumnId` to support pipeline input).
* `Get-GitHubRelease`: `ReleaseId` parameter is now `Release` (but aliased back to `ReleaseId` to support pipeline input).
* `Rename-GitHubRepository` has been effectively deprectaed. It was built on top of the same endpoint that `Set-GitHubRepository` uses, but was only modifying a single property (the name). For now, the function remains, but it now just internally calls into `Set-GitHubRepository` to perform the same action.
* `Set-GitHubRepositoryTopic`: `Name` parameter is now `Topic` (but aliased back to `Name`).
* `Get-GitHubUser`: `User` parameter is now `UserName` (but aliased back to `User` and `Name`).
* `Get-GitHubUserContextualInformation`: `User` parameter is now `UserName` (but aliased back to `User` and `Name`).
## Bug Fixes:
* Events had been using a typo'd version of the `sailor-v` AcceptHeader. This was fixed as part of the migration to using the AcceptHeader constants.
* `Lock-GitHubIssue` was not correctly providing the `lock_reason` value to the API, so that metadata was getting lost. This was caught by new UT's and then fixed.
* `Update-GitHubLabel` was modified to accept colors both with and without a leading # sign, just like `New-GitHubLabel` already supported.
## New Features:
* `Get-GitHubGitIgnore` has a new `RawContent` switch which allows you to directly get back the content of the actual .gitignore file.
* `Set-GitHubRepository` has been updated to allow users to change the name (rename) the repository at the same time as making any of their other changes. If a new name has been specified, users will be required to confirm the change, or to specify `-Confirm:$false` and/or `-Force` to avoid the confirmation.
* `Get-GitHubRepositoryCollaborator` has gained a new parameter `Affiliation` allowing you to filter which collaborators you're interested in querying for.
## Minor updates:
* Fixed the casing of the "microsoft" OwnerName in any examples
* Fixed the casing of `GitHub` in the `Assignee` method names (was `Github` intead of `GitHub`
* Added new configuration property (`DisablePipelineSupport`) to assist with pipeline development
* All `AcceptHeader` usage has migrated to using script-wide constants
* `Split-GitHubUri` can now return both components in a single function call if no switch is specified.
* Added `Join-GitHubUri` which can reverse what `Split-GitHubUri` does, based on the configured `ApiHostName`.
* Adds type information (via `.OUTPUTS` and `[OutputType()]` for every function.
* Documentation updates reflecting the new pipeline support.
## Test changes:
* `Set-StrictMode -Version 1.0` has been specified for all tests to catch access to undeclared variables (which should help catch common typo errors)
* The workarounds for PSScriptAnalyzer's false complaint for rule `PSUseDeclaredVarsMoreThanAssignments` have been removed, and all test files now suppress that rule.
* A test file now exists for every module file, and some tests were moved around to the appropriate file (away from GitHubAnalytics.tests.ps1 which had previously been a dumping ground for tests).
* A _substantial_ number of new tests have been added. Every function should now be tested with limited exceptions (like for Organizations and Teams where we don't yet have the support in the module to get the account in the appropriate state to be able to query with those existing functions).
> Note: These tests fully pass with Pester 4.10.1. Some additional work may be done prior to merging to get them passing with Pester 5.0, but the priority is to get this merged into master soon in order to unblock the pipeline of existing pull requests that have been waiting on this change.
#### Issues Fixed
- Fixes#63
#### References
[The whole API set](https://developer.github.com/v3/)
Tests were failing on Mac and Linux, but not Windows ([recent test run](https://dev.azure.com/ms/PowerShellForGitHub/_build/results?buildId=83887&view=logs&j=0da5d1d9-276d-5173-c4c4-9d4d4ed14fdb)). That's because Windows CI was running against PoSh 5.x while Linux and Mac were running on PoSh 7.x.
There's a slight difference in behavior for how those two treat arrays.
The real root cause for this was the behavior of `Invoke-GHRestMethodMultipleResult`. When creating `$finalResult`, it was always blindly adding the result to the existing array:
587e204262/GitHubCore.ps1 (L648)
`...`
587e204262/GitHubCore.ps1 (L670)
Oddly enough, this created a difference in behavior between PowerShell versions when making the result an array on the caller side. Now I ensure that I don't add anything to `$finalResult` unless there's actually a value. With that change, we can now be sure that when we grab the result as an array, it'll be appropriately empty or populated (and not populated with a single `$null` entry, thus making `Count` 1, erroneously).
I removed the attempt to force the results to be an array, because this is pointless. PowerShell will always unwrap an array of 0 or 1 in a return result. If you want to ensure that a result is always an array, you have to [wrap the result in an object](https://stackoverflow.com/a/60330501) or you have to do wrap the result in an array on the caller side.
587e204262/GitHubCore.ps1 (L684-L685)
I also normalized some naming in all of the tests, so that when we're getting back a singular result (by querying for a specific item) that we use a singular variable name, and a plural variable name otherwise.
With this change, we should now be passing CI on all OS platforms and across PowerShell 4+.
Resolves#198
Somehow a number of PSScriptAnalyzer issues snuck in. This fixes them.
The main PSScriptAnalyzer issues needing to be fixed were:
* `PSReviewUnusedParameter` - This one came up a lot due to our heavy
usage of `Resolve-RepositoryElements` and `Resolve-ParameterWithDefaultConfigurationValue`
which both end up referencing their parameters by grabbing them off
the stack. That means that `NoStatus` and `Uri` are frequently
never directly referenced. So, exceptions were added. There were
two cases (in GitHubAnalytics) where there was a false positive due
to PowerShell/PSScriptAnalyzer#1472
* `PSUseProcessBlockForPipelineCommand` - We had a number of functions
that took pipeline input, but didn't actuall use the `process` block.
This actually caught a bug with `Group-GitHubIssue` and
`Group-GitHubPullRequest`. Added correct `process` block usage for
most of the functions, but removed pipeline support for those where
it didn't actually make sense anymore.
* `PSUseDeclaredVarsMoreThanAssignments` - These are false positives
in the Pester tests due to the usage of `BeforeAll`. There wasn't
an obvious way to use `SuppressMessageAttribute` in the Pester test,
so I used a hacky workaround to "use" the variable in the `BeforeAll`
block. I could have added the suppression to the top of the file,
but I still want to catch real issues in those files later.
* `PSAvoidOverwritingBuiltInCmdlets` - It turns out that there's a bug
with PSDesiredStateConfiguration in PS Core 6.1.0 where it was exporting
internal functions. This was thus a false-postive flag for Write-Log.
See PowerShell/PowerShell#7209 for more info.
Also, it turns out that `Group-GitHubPullRequest` hadn't actually been
exported, so I fixed that too.
Mark
Howard Wolosky
Giuseppe Campanelli
Jess Pomfret