The following are naming conventions to keep in mind when coming up with a name for your cmdlet.
Cmdlet names should follow the Verb-Noun format, where the verb is from the list of approved PowerShell verbs, and the noun is a specific noun describing a resource within your service.
For ARM cmdlets, the noun must be prefixed with Az
.
From the Strongly Encouraged Development Guidelines:
Use Pascal case for cmdlet names. In other words, capitalize the first letter of the verb and all terms used in the noun. For example, "Clear-ItemProperty".
Do capitalize both characters of two-character acronyms. For example, New-AzVM, Remove-AzCosmosDBTable.
Do capitalize only the first character of acronyms with three or more characters, which aligned with Pascal case. for example, Restart-AzVmss, New-AzSqlDatabase.
From the Strongly Encouraged Development Guidelines:
Nouns used in cmdlet naming need to be very specific so that the user can discover your cmdlets. Prefix generic nouns such as "server" with a shortened version of the product name. For example, if a noun refers to a server that is running an instance of Microsoft SQL Server, use a noun such as "SQLServer". The combination of specific nouns and the short list of approved verbs enable the user to quickly discover and anticipate functionality while avoiding duplication among cmdlet names.
To enhance the user experience, the noun that you choose for a cmdlet name should be singular. For example, use the name
Get-Process
instead ofGet-Processes
. It is best to follow this rule for all cmdlet names, even when a cmdlet is likely to act upon more than one item.
If your cmdlet is performing a PATCH operation (i.e., a partial replacement on the server), then the cmdlet should use the verb Update
.
If your cmdlet is performing a PUT operation (i.e., a full replacement on the server), the cmdlet should use the verb Set
.
If you there is a separate nomenclature for your service and/or resource, or if you would like to shorten the name of the cmdlet so it's easier to remember, you can add an alias attribute to your cmdlet to allow for this functionality.
Specified by the OutputType
attribute, this piece of metadata lets the user know what the type of the object returned by the cmdlet is (found in the Outputs section of a cmdlet's help content). The type specified here should always be a single element and not an enumeration of elements (e.g., PSVirtualMachine
instead of List<PSVirtualMachine>
).
If the cmdlet returns an object, the type of the object returned must be defined; the output type for a cmdlet should never be object
, PSObject
, PSCustomObject
or the like. Returning these types of objects makes it difficult for the user to anticipate what properties will be found on the object returned from the cmdlet, as well as makes it impossible for the breaking change analyzer to detect if a breaking change was introduced to the cmdlet as the type is not defined.
In order to preserve proper piping scenarios, the output type for a cmdlet should never be a string
. If a cmdlet is expected to return a string
, the suggestion is to introduce a new type that encapsulates the string
information as a property and return that object. The PowerShell language revolves around objects and passing them around cmdlets; returning string
objects can introduce inconsistencies in the piping experience for users.
In most cases, cmdlets will be returning an object corresponding to a resource that a user is performing an action on. Rather than returning the .NET SDK type for that resource (exposing .NET SDK types in PowerShell cmdlets is strongly discouraged), we suggest creating a new class that wraps this .NET SDK type, allowing for breaking changes in the underlying type while avoiding breaking changes in the PowerShell type.
For example, the Get-AzVM
cmdlet uses the .NET SDK to retrieve objects of the VirtualMachine
type, but a new class, PSVirtualMachine
, was created to wrap the type from the .NET SDK, and is returned by the cmdlet. If, in the future, the VirtualMachine
type in the .NET SDK has a property removed, that property can still be maintained in PowerShell by adding it to the PSVirtualMachine
type and recreating the value, thus avoiding a breaking change in the corresponding cmdlet(s).
In the case where your cmdlet doesn't return any output (e.g., deleting, starting, stopping a resource), the cmdlet should implement the -PassThru
parameter and the OutputType
should be set to bool
. The -PassThru
parameter is a SwitchParameter
set by the user to signal that they would like to receive output from a cmdlet which does not return anything. If the -PassThru
parameter is provided, you should return the value true
so the user is made aware that the operation was successful. If the operation was unsuccessful, then the cmdlet should throw an exception.
From the Strongly Encouraged Development Guidelines:
By default, many cmdlets that modify the system, such as the
Stop-Process
cmdlet, act as "sinks" for objects and do not return a result. These cmdlet should implement the-PassThru
parameter to force the cmdlet to return an object.
The code below shows how this should look in a cmdlet:
[Cmdlet(...), OutputType(typeof(bool))]
public class MySampleCmdlet : MyBaseCmdlet
{
// other parameters omitted
[Parameter(Mandatory = false)]
public SwitchParameter PassThru { get; set; }
public override void ExecuteCmdlet()
{
// other code omitted
if (this.PassThru.IsPresent)
{
WriteObject(true);
}
}
}
When returning a collection of objects, the cmdlet should enumerate the collection. This ensures that the objects are written to the pipeline one at a time, which is the expected behavior for PowerShell cmdlets.
There are two ways to accomplish this: (a) call WriteObject()
for each object in the collection, or (b) use WriteObject()
with the enumerateCollection
parameter set to true
. The enumerateCollection
parameter is a boolean that, when true, will enumerate the collection and write each object to the pipeline.
The code below shows how this should look in a cmdlet:
var resources = Client.ListResources();
// option a: call WriteObject() for each object in the collection
foreach (var resource in resources)
{
WriteObject(resource);
}
// option b: use WriteObject() with the enumerateCollection parameter set to true
WriteObject(resources, true);
If a cmdlet makes any changes to an object on the server (e.g., create, delete, update, start, stop a resource), the cmdlet should implement ShouldProcess
. This property adds the -WhatIf
and -Confirm
parameters to the cmdlet:
-WhatIf
is aSwitchParameter
that, when provided by the user, doesn't execute the part of the cmdlet responsible for making the changes to the object, but rather displays a message alerting the user of the action that is to be performed on the object-Confirm
is aSwitchParameter
that, when provided by the user, prompts the user for confirmation that they want to continue with the execution of the cmdlet.
The code below shows how this should look in a cmdlet:
[Cmdlet(..., SupportsShouldProcess = true), OutputType(typeof(...))]
public class MySampleCmdlet : MyBaseCmdlet
{
// parameters omitted
public override void ExecuteCmdlet()
{
// other code omitted
if (ShouldProcess(targetResource, actionMessage))
{
// make the change
}
}
}
More information about ShouldProcess
can be found in the Should Process and Confirm Impact document.
The -Force
parameter is reserved for special scenarios where additional confirmation from the user is required. From the above document on Should Process and Confirm Impact document:
Some cmdlets require additional confirmation. For example, if a cmdlet would destroy existing resources in some circumstances, the cmdlet might detect that condition and prompt the user to verify before continuing. Overwriting an existing resource during resource creation, overwriting a file when downloading data, deleting a resource that is currently in use, or deleting a container that contains additional resources are all example of this pattern. To implement additional confirmation, and allow scripts to opt out of additional prompts, the above pattern is enhanced with calls to
ShouldContinue()
and the-Force
parameter.
All long running operations must implement the -AsJob
parameter, which will allow the user to create jobs in the background. For more information about PowerShell jobs and the -AsJob
parameter, read this doc.
To implement the -AsJob
parameter, simply add the parameter to the end of the parameter list:
[Parameter(Mandatory = false, HelpMessage = "Run cmdlet in the background")]
public SwitchParameter AsJob { get; set; }
Once you add the parameter, please manually test that the job is created and successfully completes when the parameter is specified. Additionally, please ensure that the help files are updated with this parameter.
To ensure that -AsJob
is not broken in future changes, please add a test for this parameter. To update tests to include this parameter, use the following pattern:
$job = Get-AzSubscription -AsJob
$job | Wait-Job
$subscriptions = $job | Receive-Job
To set a custom job name, please use SetBackgroupJobDescription
. The default job description is: "Long Running Operation for '{cmdlet name}' on resource '{resource name}'"
In most Azure PowerShell cmdlets, there is a bare minimum of three parameter sets that need to be implemented.
This parameter set should be implemented by every cmdlet - in most cases, the user provides the name of the resource that they are acting upon (-Name
) and the resource group in which they are acting in (-ResourceGroupName
).
The interactive parameter set will always be the default parameter set for a cmdlet (specified by the DefaultParameterSetName
property in the Cmdlet
attribute). This means that when PowerShell is unable to determine which parameter set a user is in, it will default to the interactive parameter set and prompt the user to provide values for the missing mandatory parameters.
This parameter set should be implemented by every cmdlet - the user is able to provide a -ResourceId
string or GUID from the Azure Portal, or from one of the generic resources cmdlets (more information about this scenario can be found in the piping-best-practices.md
document), and act upon the given resource associated with the id. The typical -Name
and -ResourceGroupName
parameters are replaced by a single -ResourceId
parameter of type string.
This parameter should be implemented by most cmdlets - the user is able to take the object returned from the Get
, New
, or Set
cmdlets (or other cmdlets that return the common resource) and provide it to the -InputObject
parameter for a cmdlet that acts upon the same resource (more information about this scenario can be found in the piping-best-practices.md
document). The typical -Name
and -ResourceGroupName
parameters are retrieved from the -InputObject
that the user is passing through.