I've seen this, but neither answer actually answers the question using named argument in the call to the function
This is for Powershell 7. No workflow. And really Named Parameters this time! Also not covered in the standard documentation here .. Starting to think that this isn't possible.
this works, but it is requires brackets with positional arguments, not named arguments
Function MyFunction ([ref][string]$MyRefParam) {
$MyRefParam.Value = "newValue";
}
$myLocal = "oldValue"
Write-Host $myLocal # Outputs: oldValue
MyFunction ([ref]$myLocal);
Write-Host $myLocal # Outputs: newValue
Of course we all know that the best way to call a Powershell function is with named argument MyFunction -Arg1 23 rather than positional arguments MyFunction 23 or MyFunction(23) . These would not get through our Pull Requests!
But this doesn't work
Function MyFunction ([ref][string]$MyRefParam) {
$MyRefParam.Value = "newValue";
}
$myLocal = "oldValue"
Write-Host $myLocal # Outputs: oldValue
MyFunction -MyRefParam ([ref]$myLocal) # Outputs Cannot process argument transformation on parameter 'MyRefParam'. Reference type is expected in argument.
Write-Host $myLocal # Outputs: oldValue
Is there another way to provide the ref type in this syntax? So far I've tried combinations of [ref] and [ref][string] in both the param definition and in the call - I can't get anything to let Powershell see that I am really passing a Ref
Thanks for any help!
Use only a [ref] type constraint in your parameter declaration, i.e. remove the additional [string] type constraint (it doesn't do anything, and arguably shouldn't even be allowed - see bottom section):
Function MyFunction ([ref] $MyRefParam) {
$MyRefParam.Value = "newValue"
}
$myLocal = 'oldVaue'
MyFunction -MyRefParam ([ref] $myLocal)
$myLocal # -> 'newvalue'
You cannot type a [ref] instance: [ref] isn't a keyword that modifies a parameter declaration (as ref would be in C#), it is a type in its own right, System.Management.Automation.PSReference, and its value-holding property, .Value is of type object, i.e. it can hold any type of object.
The upshot:
You cannot enforce a specific data type when a value is passed via [ref].[1]
Conversely, you're free to assign a value of any type to the .Value property of the [ref] instance received.
Taking a step back:
[ref] is primarily intended for supporting calls to .NET APIs with ref / out parameters.
Its use in pure PowerShell code is unusual and best avoided, not least due to the awkward invocation syntax.
That said, the difference in behavior with your double type constraint between positional and named parameter binding is certainly curious - see Mathias R. Jessen's excellent explanation in the comments.
However, the real problem is that you're even allowed to define a multi-type constraint (unusual in itself) that involves [ref], because there appears to be an intentional check to prevent that, yet it doesn't surface in parameter declarations.
You can make it surface as follows, however (run this directly at the command line):
# ERROR, because [ref] cannot be combined with other type constraints.
PS> [ref] [string] $foo = 'bar'
Cannot use [ref] with other types in a type constraint
That the same check isn't enforced in parameter declarations has been reported in GitHub issue #16146.
[1] Behind the scenes, a generic type that is non-public is used when an actual value is passed with a [ref] cast (System.Management.Automation.PSReference<T>), which is instantiated with whatever type the value being cast is. However, this generic type's .Value property is still [object]-typed, allowing for modifications to assign any type.
Related
I'm trying to invoke the List[T](IEnumerable) directly adding an item to the initial List like so, where T is a PowerShell class I've written (the below example uses the class name Thing:
$someObject = Get-Thing # returns a single object
$list = [List[Thing]]::new(#( $someObject ))
However, this yields an error suggesting it can't find the overload for this constructor:
Cannot find an overload for "List`1" and the argument count: "1".
Setting List[T] to the Object class works, however:
$someObject = Get-Thing
$list = [List[Object]]::new(#( $someObject ))
While this works, I'm unsure why I'm unable to use my PowerShell class as the type. My understanding is that only context-bound types and (by default) nested types are unable to be used with generics, but the following shows that my class is not a ContextBoundObject:
class Thing {
$Name
Thing($name) {
$this.Name = $name
}
}
$thing = [Thing]::new('Bender')
$thing -is [System.ContextBoundObject] # ==> False
I'm not certain if a PowerShell class would be a nested type of some sort, and about_Classes does not mention nested types.
I'm unsure why I'm unable to use my PowerShell class as the type
The array subexpression operator #() returns its results as [object[]] - a type which satisfies the argument type [IEnumerable[object]] - which is why it always works when you use [object] as the type parameter for the receiving collection type.
So, what to do about that?
If the array consists only of [Thing]'s, you can explicitly cast to a more specific collection type that implements [IEnumerable[Thing]]:
$list = [List[Thing]]::new([Thing[]]#( $someObject ))
To complement Mathias R. Jessen's helpful answer, which explains the problem well and offers an effective solution:
PowerShell's casts are not only syntactically more convenient, but also more flexible when it comes to on-demand type conversions.
Indeed, using a cast instead of calling a constructor, via the static ::new() method, does work:
using namespace System.Collections.Generic
class Thing { [string] $Name; Thing([string] $name) { $this.Name = $name } }
# Both of the following work:
# Single [Thing] instance.
$list = [List[Thing]] [Thing]::new('one')
# Multiple [Thing] instances, as an array, via the grouping operator, (...)
# #(...), the array subexpression operator, works too, but is unnecessary.
$list = [List[Thing]] ([Thing]::new('one'), [Thing]::new('two'))
PowerShell's automatic type conversions, as also used in casts:
Unfortunately, as of this writing the rules aren't documented, but a comment in the source-code provides a high-level overview, as does the (pretty low-level) ETS type converters documentation, which can be summarized as follows, in descending order of precedence:
First, engine-internal, fixed conversion rules may be applied (see source-code link above).
A notable internal rule concerns to-string conversions: while any .NET type supports it by an explicit call to its .ToString() method (inherited from the root of the object hierarchy, System.Object), PowerShell applies custom rules:
If a type has a culture-sensitive .ToString(<IFormatProvider>) overload, PowerShell passes the invariant culture deliberately, to achieve a culture-invariant representation, whereas a direct .ToString() call would yield a culture-sensitive representation - see this answer for details; e.g., in a culture where , is the decimal mark, [string] 1.2 returns '1.2' (period), whereas (1.2).ToString() returns '1,2' (comma).
Collections, including arrays, are stringified by concatenating their (stringified) elements with a space as the separator (by default, can be overridden with preference variable $OFS); e.g., [string] (1, 2) returns 1 2, whereas (1, 2).ToString() returns merely System.Object[].
Also, PowerShell converts freely:
between different number types (when possible).
between numbers and strings (in a culture-invariant manner, recognizing only . as the decimal mark when converting from a string).
and allows any data type to be converted to (interpreted as) as Boolean - see the bottom section of this answer for the rules.
Next, TypeConverter or (PSTypeConverter) classes that implement custom conversions for specific types are considered.
If the input type is a string ([string]), a static ::Parse() method is considered, if present: first, one with a culture-sensitive signature, ::Parse(<string>, <IFormatProvider>), in which case the invariant culture is passed, and, otherwise one with signature ::Parse(<string>).
Next, a single-argument constructor is considered, if the input type matches the argument's type or is convertible to it.
If an implicit or explicit conversion operator exists for conversion between the input and the target type.
Finally, if the input object implements the System.IConvertible interface and the target type is a supported-by-the-implementation primitive .NET type except [IntPtr] and [UIntPtr] or one of the following types: [datetime], [DBNull], [decimal].
In C#, it's possible to get the default value of any type using the default operator:
var i = default(int); // i == 0
// in C# 7.1+
int j = default; // j == 0
Is there a similar construct in PowerShell, and if so what is it? As far as I've been able to determine in my Googling and testing, default is only recognized by PS when present in switch blocks.
PowerShell has no direct language construct for it because it doesn't need it -- due to its loose typing you are almost never required to produce a value of a specific type and there is no support for creating generic types or functions. Untyped variables start off as $null if you do nothing special. Typed variables start off as whatever value you explicitly give them, and that's generally sufficient due to PowerShell's liberal rules for conversion ([int] "" and [int] $null are both 0).
Only in rare cases does this fail, like attempting to declare a variable of type DateTimeOffset, as there is no default constructor and $null or "" won't convert. Arguably, the fix there is to just explicitly construct a value using whatever the type does offer ([DateTimeOffset] $d = [DateTimeOffset]::Now, [DateTimeOffset] $d = [DateTimeOffset]::MinValue, [DateTimeOffset] $d = "0001-01-01 00:00Z"). Only in the very rare case that you have a dynamic type, and you'd like to get what C# would give you with default, would you need some special code. You can do it in pure PowerShell (well, almost, we need to call a method available since .NET 1.0):
Function Get-Default([Type] $t) { [Array]::CreateInstance($t, 1)[0] }
And then [DateTimeOffset] $d = Get-Default DateTimeOffset works (there is no way to infer the type in this case, though you are of course free to omit it from the variable).
Of course this does create a garbage array on every invocation; it does not invoke any constructors of the type itself, however. There are more involved approaches that avoid array creation, but they all involve getting complicated with generic methods (relying on LINQ) or explicitly compiling C# and aren't really worth demonstrating as they're less general. Obviously, even the function above should be used only in the unusual case where it might be needed and not as a general way of initializing variables -- typically you know the type and how to initialize it, or you don't care about the type in the first place.
I am trying to figure out how to get the name of a powershell variable from the object, itself.
I'm doing this because I'm making changes to an object passed by reference into a function, so I don't know what the object will be and I am using the Set-Variable cmdlet to change that variable to read only.
# .__NEEDTOGETVARNAMEASSTRING is a placeholder because I don't know how to do that.
function Set-ToReadOnly{
param([ref]$inputVar)
$varName = $inputVar.__NEEDTOGETVARNAMEASSTRING
Set-Variable -Name $varName -Option ReadOnly
}
$testVar = 'foo'
Set-ToReadOnly $testVar
I've looked through a lot of similar questions and can't find anything that answers this specifically. I want to work with the variable entirely inside of the function--I don't want to rely on passing in additional information.
Also, while there may be easier/better ways of setting read-only, I have been wanting to know how to reliably pull the variable name from a variable for a long time, so please focus solving that problem, not my application of it in this example.
Mathias R. Jessen's helpful answer explains why the originating variable cannot be reliably determined if you only pass its value.
The only robust solution to your problem is to pass a variable object rather than its value as an argument:
function Set-ToReadOnly {
param([psvariable] $inputVar) # note the parameter type
$inputVar.Options += 'ReadOnly'
}
$testVar = 'foo'
Set-ToReadOnly (Get-Variable testVar) # pass the variable *object*
If your function is defined in the same scope as the calling code - which is not true if you the function is defined in a (different) module - you can more simply pass just the variable name and retrieve the variable from the parent / an ancestral scope:
# Works ONLY when called from the SAME SCOPE / MODULE
function Set-ToReadOnly {
param([string] $inputVarName)
# Retrieve the variable object via Get-Variable.
# This will implicitly look up the chain of ancestral scopes until
# a variable by that name is found.
$inputVar = Get-Variable $inputVarName
$inputVar.Options += 'ReadOnly'
}
$testVar = 'foo'
Set-ToReadOnly testVar # pass the variable *name*
As noted in this answer to a similar question, what you're asking (resolving the identity of a variable based on its value) can not be done reliably:
The simple reason being that contextual information about a variable
being referenced as a parameter argument will have been stripped away
by the time you can actually inspect the parameter value inside the
function.
Long before the function is actually called, the parser will have
evaluated the value of every single parameter argument, and
(optionally) coerced the type of said value to whatever type is
expected by the parameter it's bound to.
So the thing that is ultimately passed as an argument to the function
is not the variable $myVariable, but the (potentially coerced) value
of $myVariable.
What you could do for reference types is simply go through all variables in the calling scope and check if they have the same value:
function Set-ReadOnlyVariable {
param(
[Parameter(Mandatory=$true)]
[ValidateScript({ -not $_.GetType().IsValueType })]
$value
)
foreach($variable in Get-Variable -Scope 1 |Where-Object {$_.Value -ne $null -and $_.Value.Equals($value)}){
$variable.Options = $variable.Options -bor [System.Management.Automation.ScopedItemOptions]::ReadOnly
}
}
But this will set every single variable in the callers scope with that value to readonly, not just the variable you referenced, and I'd strongly recommend against this kind of thing - you're most likely doing something horribly wrong if you need to do this
I've got some fairly complex functions that I'm writing for a library module, with lots of different ways it can be called. However, it is in fact possible to default all of them, but when I try to call my function with no parameters the call fails because the parameter set cannot be determined.
I would like to define a parameter set that contains no parameters whatsoever, such that calls with no parameters succeed. This is difficult to do since ParameterSetName is a property of the Parameter attribute though, and it's not possible to attribute nothing.
I experimented with the DefaultParameterSet property of the CmdletBinding attribute that is placed on the param block, however nothing seemed to work. It seems that the parameter set name defined there must actually exist in order for Powershell to default to it.
Currently my best approximation of this use case is to define one of the parameter sets to have no Mandatory parameters, however these fail when empty strings or nulls are piped in, and I would like for this not to be the case.
Is this possible?
Sure is. Just specify a default parameter set name that isn't used otherwise:
function Foo {
[CmdletBinding(DefaultParameterSetName='x')]
Param(
[Parameter(Mandatory=$true, ParameterSetName='y')]$a,
[Parameter(Mandatory=$false, ParameterSetName='y')]$b,
[Parameter(Mandatory=$true, ParameterSetName='z')]$c,
[Parameter(Mandatory=$false)]$d
)
"`$a: $a"
"`$b: $b"
"`$c: $c"
"`$d: $d"
}
I am sure I read somewhere that there is an easy way to pass named parameters from a calling function to a called function without explicitly naming and specifying each parameter.
This is more than just reusing the position; I'm interested in the case where the name of the passed parameters is the same in some cases, but not in others.
I also think there is a way that is not dependent on position.
function called-func {
param([string]$foo, [string]$baz, [string]$bar)
write-debug $baz
write-host $foo,$bar
}
function calling-func {
param([int]$rep = 1, [string]$foo, [string]$bar)
1..$rep | %{
called-func -foo $foo -bar $bar -baz $rep ## <---- Should this be simpler?
}
}
calling-func -rep 10 -foo "Hello" -bar "World"
What would the method be, and is there a link?
I thought it might have been Jeffrey Snover, but I'm not sure.
In PowerShell v2 (which admittedly you may not be ready to move to yet) allows you to pass along parameters without knowing about them ahead of time:
called-func $PSBoundParameters
PSBoundParameters is a dictionary of all the parameters that were actually provided to your function. You can remove parameters you don't want (or add I suppose).
Well, I think I was confusing a blog post I read about switch parameters. As far as I can tell the best way is to just reuse the parameters like so:
called-func -foo:$foo -bar:$bar
How about
called-func $foo $bar