Tips for golfing in PowerShell

Question

What general tips do you have for golfing in Windows PowerShell? I'm looking for ideas that can be applied to code golf problems in general that are at least somewhat specific to PowerShell (e.g. "remove comments" is not an answer). Please post one tip per answer.

—taken nearly verbatim from marcog's question.

Answer 1

Powers of 10 literals with scientific notation:

4e6 = 4000000

Powers of 2 literals:

4KB = 4096
4MB = 4194304
4GB = 4294967296

# TB and PB suffixes also exist, but less useful for golf.

Could come in handy.

Answer 2

If you need to run a loop, and you know exactly how many times it needs to run every time, consider piping an array of contiguous integers into ForEach-Object via the % alias instead of using for.

for($x=1;$x-le10;$x++){...}

vs

1..10|%{...}

Answer 3

You can skip spaces a lot in PowerShell. If it feels like it might not be needed, it quite possibly isn't. This is particularly useful in comparisons.

Example:

$x-eq$i-and$x-ne9

vs.

$x -eq $i -and $x -ne 9

---

If you need to branch your script based on the result of a single test which may have multiple outcomes, switch can sometimes match or beat an if statement.

Example (switch vs. if/else - tie):

if($x%2-eq0){'even'}else{'odd'}

vs.

switch($x%2){0{'even'}1{'odd'}}

Or (switch vs. if/elseif/else - switch wins by 15):

if($x%2-eq0){'even'}elseif($x%2-eq1){'odd'}else{'error'}

vs.

switch($x%2){0{'even'}1{'odd'}2{'error'}}

If the switch is actually based on certain math results, like the modulo operation above, you can replace the switch entirely with an array. Here, it saves another 13 characters and is even shorter than the original two-option if/else statement. (Thanks to Danko Durbic for this bit.)

('even','odd','error')[$x%2]

If you will be using a particular command a lot, especially one without a pre-existing short-hand alias, set up a single-character alias early on.

Example:

nal g Get-Random;g 10;g 10;g 10

vs.

Get-Random 10;Get-Random 10;Get-Random 10

Answer 4

Encapsulating the command that defines a variable in parenthesis allows you to feed the variable's definition directly to other commands.

For example, you can set $x and then set $y based on the value of $x in one shot with this:

$y=($x=1)+1

Instead of this:

$x=1;$y=$x+1

You can set $h and output it with this:

($h='Hello World!')

Instead of this:

$h='Hello World!';$h

Answer 5

A switch can act like a loop, when given an array. For example:

$FooBarMeh='a','b','c'
switch ($FooBarMeh)
{
    'a'{'FOO'}
    'b'{'BAR'}
    default{'MEH'}
}

Will output:

FOO
BAR
MEH

I'm not totally sure where this will be useful, but I expect it will be handy for someone some time.

Answer 6

Replace [math]::pow with multiplication. Instead of

[math]::pow($a,$b)

you can write

"$a*"*$b+1|iex

This works for integer exponents >= 0.

Answer 7

Comparison operators work on collections of values by returning matching values:

1..5 -gt 2

will yield 3, 4 and 5. In some cases this can help to save an otherwise longer |?{$_...}.

-match is a comparison operator too.

Answer 8

Use aliases whenever possible. There are a bunch of useful ones:

?        Where-Object
%        ForEach-Object
gu       Get-Unique
sort     Sort-Object
iex      Invoke-Expression

Answer 9

Want to find the maximum or minimum of a collection of values? Tried

(...|measure -ma).Maximum

or

(...|measure -mi).Minimum

already?

Just sort and use the last or first item:

(...|sort)[-1]  # maximum
(...|sort)[0]   # minimum

Answer 10

Shortening Property Names

Sadly, unlike parameters, properties/methods (anything accessed with a dot .) cannot usually be shortened down to its unambiguous form.

But certain cmdlets can operate on property names and take wildcards, and there are little-known parameter sets of % and ? that can be useful.

Usually we pass in a scriptblock and refer to the item with $_, but there's another form of these that takes a property name, and it accepts a wildcard.

$o|select Le*  
$o|%{$_.Length}

With a property like .Length we can't use the v3 magic that would normally work on an array because Length is a property of the array itself, so the above two could be used to get the lengths of the individual members. The select comes in a little bit shorter.

But % can take a property name directly and return that value:

$a|% Length

Which can be shortened with wildcards. The wildcard must resolve to a single property (or method, more on that later), so it will throw a helpful error if it doesn't, indicating exactly which members it could resolve to.

In the case of Length, Le* is typically the shortest. Even on a single string, this method is 1 byte shorter than just using the property.

$a.Length                # 9   #(doesn't work on array)
$a|%{$_.Length}          # 15
$a|% Le*                 # 8

But depending on what you're doing with this, this can be worse. You can do $a.Length*5 but to do it with the pipeline expression you'd have to wrap it ($a|% Le*)*5; might still be worth it if it's against an array, but the point is it's not always appropriate as a straight substitution.

It works with methods too, and you can leave off the () which makes a full name the same length, but same restriction as above about sometimes having to wrap it. The method must have an overload that takes no parameters (you can pass arguments by placing them after the method name, which is really nice):

$a.ToUpper()             # 12
$a|% *per                #  9

With arguments:

'gaga'-replace'g','r'    # 21
'gaga'|% *ce g r         # 16

These aren't strictly the same in that the -replace operator does a regex replace, but if you're just doing a string replace, it can (now) be shorter to use the method; it helps that the strings are cmdlet arguments instead of method arguments so they don't need to be quoted.

Where-Object Properties

? can take (partial) property names as well, and apply an "operator" to it (in the form of switch parameters). Again this can be shorter than using the standard Where-Object scriptblock approach if the property name is sufficiently long and unique.

$a|?{$_.Length-gt5}      # 19
$a|? Le* -GT 5           # 14

($a|% Le*)-gt5           # 14 - Lengths, not objs

Answer 11

Finding a sum the long way:

(...|measure -s).Sum

A shorter way:

...|%{$s+=$_};$s

And even shorter:

...-join'+'|iex

Answer 12

The Get verb is implied. This can shorten any Get-Frob to just Frob. Frequent contenders are date or random.

Note that this won't work properly in some cases because you might have GNU utilities in your path (or other native programs that clash). Order of command lookup in that case seems to prefer the native program before it considers cmdlets with the Get- removed:

PS Home:\> date

Freitag, 15. November 2013 07:13:45


PS Home:\> $Env:Path += ';D:\Users\Joey\Apps\GnuWin32\bin'
PS Home:\> date
Fr Nov 15 07:14:13 W. Europe Standard Time 2013

Answer 13

Semicolons and line breaks are interchangeable. Golfed code is often more readable if not jammed into a single line. And the length is still the same (provided you use U+000A as line break which PowerShell handles without problems).

Answer 14

for loops can have anything between 0 and three statements in their header:

Endless loop:

for(){}

Loop with initialization:

for($n=0){}

Loop with initialization and end condition:

for($n=0;$n-lt7){}

In such cases the additional semicolons at the end may be omitted (it's explicitly stated in the language specification, so it's not an implementation detail) in contrast to C-like languages which always require exactly three statements.

This also makes while a bit shorter. Compare

while(...){}

and

for(;...){}

With the added bonus that you can stick in a previous line (if there is one) into the for as well without extra cost (and even saving a character).

Answer 15

Casting to string:

[string]$x

vs.

"$x"

Casting to string like this can also be used to flatten an array of strings, instead of joining it:

$a = @('a','b','c')
$a -join ' '

vs.

$a = @('a','b','c')
"$a"

Casting a string to a numeric type:

[int]$x     [float]$x

vs.

+$x

Also very useful to know that PowerShell always takes the type of the left operand to determine the final type of an expression and conversions to apply:

'1'+2    -> '12'
1+'2'    -> 3

which can help determining where needless casts are.

Answer 16

If you are assigning an array that you know will only have two values, don't use indexing.

Something like this:

$a="apple","orange"
$a[0] # apple
$a[1] # orange

Can easily be turned into this:

$a,$o="apple","orange"
$a # apple
$o # orange

This can also be useful for if you just need to the first element of an array:

$a,$b=1..10
$a # 1
$b # 2..10

Answer 17

Don't forget that you don't always need to provide the full name of a parameter, and some parameters are positional.

Get-Random -InputObject (0..10)

...can be trimmed to...

Get-Random -I (0..10)

...because "I", in this case, is enough to uniquely identify InputObject from the other valid parameters for this command.

You could trim it further to...

Get-Random (0..10)

...because InputObject is a positional parameter.

Piping is usually shorter than feeding objects as a parameter, especially when it can remove the need for parenthesis. Let's trim our random number generator further...

0..10|Get-Random

Also be on the lookout for other ways to accomplish the same thing, even if you can't change the command. For the above case, you could do this:

Get-Random 11

Or, incorporating another suggestion*:

Random 11

**Note: Omitting Get- from a command name can bloat the run time by about 50,000%. Not bad if you only need the command once, but be careful using it in long loops.*

And that's how can knock a simple command down to a third of its size.

Answer 18

When using a number as an argument to an operator that would otherwise require a string, you can use the number directly. Compare

...-join'0'

vs.

...-join0

Works with -split as well. The argument is always converted to a string first.

Answer 19

Fake ternary operator. You can assign straight from an if statement:

$z=if($x-eq$y){"truth"}else{"false"}

But you can use a 2-element array and use the test to index into it. $falsey results get element 0, $truthy results take element 1:

$z=("false","true")[$x-eq$y]

NB. that this is really doing array indexing, and if the test results in a value which can be cast to an integer, you'll ask for an item outside the bounds of the array and get $null back, and will need to do !(test) to force cast the result to a bool, with the options reversed.

Answer 20

Absolute value

With

$n=-123

Instead of

[math]::abs($n)

use

$n-replace'-'

Of course, the savings are cancelled if parentheses are needed.

Answer 21

If you need to silence errors, the obvious variant would be

try{ <# something that throws errors #> }catch{}

However, this is way too long. A shorter variant is to run the try block as a script block and just redirect the errors into an unset variable ($null would be the usual one, but that's still too long):

.{ <# something that throws errors #> }2>$x

This saves five valuable bytes (if not the world).

Answer 22

Use the $ofs special variable to change the Output Field Separator used when stringifying an array. Useful if you're needing to transform arrays to strings multiple times.

For example:

$a=1,2,3,4
$a-join',';$a-join','
$ofs=',';"$a";"$a"

Saves 2+n characters on the 2nd -join, where n is the length of the separator, and saves an additional 5+n for the 3rd -join and each thereafter.

Answer 23

Automatic variables have booleans for True and False as $true and $false but you can get similar results using the logical not operator ! and the integers 0 and 1( or any non-zero integer.)

PS C:\Users\matt> !1
False

PS C:\Users\matt> !0
True

Near all PowerShell expressions can be evaluated as booleans. So as long as you are aware of how certain data is evaluated you can get booleans and never need to explicitly cast them. Be aware of the LHS value when doing these.

• Integer 0 is false and non-zero integers are evaluated to true.
• non-zero length strings are true and empty or null (and nulls themselves) strings are false.

There are other examples but you can easily test by doing a cast

PS C:\Users\matt> [bool]@(0)
False

Answer 24

Consider storing repeated script blocks in variables, instead of using functions.

I was going to use this to save some characters in my Rock, Paper, Scissors implementation before I realized that re-writing the function as a variable made even the variable unnecessary. This could still be useful for other scripts though, where you're actually running the same code multiple times.

function Hi{echo 'Hello, World!'};Hi

vs.

$Hi={echo 'Hello, World!'};&$Hi

Answer 25

You can use $s|% t*y instead [char[]]$s to split a string to char array. Given from TessellatingHeckler's answer: % t*y expands to | ForEach-Object -Method ToCharArray equiv. of "$args".ToCharArray()

For example, compare

$s|% t*y

and

[char[]]$s

and

$s.ToCharArray()

It's useful with $args especially: $args|% t*y

Answer 26

Invoke-Expression and Get-Random can also get pipeline input instead of arguments.

For iex this allows to save parentheses on some expressions:

iex 1..5-join'+'   # won't work
iex(1..5-join'+')  # does work, but has extra parentheses
1..5-join'+'|iex   # doesn't need the parentheses

In case of random this allows a common case to be optimized a bit:

random -mi 10 31   # gets a random integer between 10 and 30
10..30|random      # much better :-)
(random 21)+10     # if needed in another expression that doesn't require extra
                   # parentheses

The latter way of using it simply selects an item from a list. The -c argument can be given to allow more than a single selection.

Answer 27

Reversing an array

Comes in handy in a lot of ascii-art challenges where the output is mirrored in some fashion.

Suppose you have

$a=1,2,3,4,5

The traditional reversal method is long, boring, and doesn't leave the array on the pipeline for immediate use.

[array]::reverse($a)

Indexing directly into the array in reverse order saves a few bytes, since it leaves the results on the pipeline, but is still rather long:

$a[($a.count-1)..0]

Instead, try a loop

$a|%{$a[-++$i]}

Answer 28

Starting with PowerShell Core 6, you can also use ranges for characters:

'a'..'z'

which can replace the much more cumbersome

0..25|%{[char]($_+97)}

Answer 29

Use variables to store .NET names

As outlined by cogumel0 in this answer, you can use variables to store .NET type names. For example, changing

param($a)$a-[math]::pow(2,[math]::floor([math]::log($a,2)))

into

param($a)$a-($m=[math])::pow(2,$m::floor($m::log($a,2)))

saved 3 bytes in this example.

Answer 30

A simple one but control-flow statements (as opposed to pipelines), such as if, for, and while, do not need a semi-colon after their braces.

E.g.

# Control-flow statements (All are valid)
if(...){...}"foo"
for(...){...}"foo"
while(...){...}"foo"
switch(...){...}"foo"

# Pipelines
1..2|%{...}"bar" #This will throw an error.
1..2|%{...};"bar" #This will work.

A free byte that's easy to miss.

Answer 31

Converting floating-point numbers to integers in PowerShell is a bit of a minefield. By default the conversion does Bankers Rounding which doesn't always trim off the decimal and leave the smaller whole number, or always round .5 up to the next number like people do casually, it rounds evens one way and odds another way - this can be surprising, e.g.

PS C:\> [int]1.5
2

PS C:\> [int]2.5
2

and break codegolf calculations. Many other common languages do truncation-rounding, therefore golf questions often require truncation. You might reach for [Math]::Floor() as the next best thing, but beware this only behaves the same as truncation for positive numbers, but it takes negative numbers lower - further away from zero. [Math]::Truncate() is what you need to bring PS behaviour in line with other language's default rounding, but it's a lot of characters.

Regex replacing digits after the decimal point can help save a couple of characters:

[Math]::Truncate(1.5)
[Math]::Floor(1.5)
1.5-replace'\..*'

[Math]::Truncate($a)
[Math]::Floor($a)
$a-replace'\..*'
$a.split('.')[0]        # literal character split, not regex pattern split

Answer 32

Use Boolean logic in place of if-counters in a loop

Suppose you're adding all even numbers from 1 to 10 ... 2+4+6+8+10=30

1..10|%{if($_%2-eq0){$o+=$_}};$o

You could use Boolean negation to shorten it to

1..10|%{if(!($_%2)){$o+=$_}};$o

to save a byte, but how about instead use implicit casting of Booleans to ints, and roll the conditional into the accumulator

1..10|%{$o+=$_*!($_%2)};$o

Saves 6 bytes in this example.

Answer 33

SLS instead -match

You can use an sls cmdlet (alias for Select-String) with |% M* as shortcut for property Matches instead a -match operator with a $Matches automatic variable. Compare:

$args|sls 'pattern'|% M*

vs.

if("$args"-match'pattern'){$Matches}

---

You can also use sls with -a parameter (-AllMatches) to find all matches:

$args|sls 'pattern'-a|% M*

---

Also, you can find all matched lines in a multiline string without explicitly line breaking. It need to use a regexp option ?(m). Compare:

$args|sls '(?m)^pattern$'-a|% M*

vs.

$args-split"``n"|%{$_-match'^pattern$';$Matches}

Answer 34

When declaring a hard-coded list of strings, depending on the size of the strings and the number of them, it's often shorter to split a single string up by spaces, since space-delimited splitting doesn't require a delimiter provided as a parameter

For example,

Instead of

$numbers='one','two','three','four','five','six','seven','eight','nine','ten'

Do

$numbers=-split'one two three four five six seven eight nine ten'

You can further delimit with another character and split twice to hold shortened hard-coded multidimensional arrays

Answer 35

If you need an if/else at the end of the program (maybe to handle a special case differently), then instead of

if(foo){a}else{b}

do

if(foo){a;exit}b

which saves a character.

Answer 36

Declaring an anonymous function should come up alot. There are multiple variations of this that are covered here https://stackoverflow.com/questions/10995667/lambda-expression-in-powershell involving scripts blocks. I use a similar one to this

$a={iex([int[]][char[]]$args[0]-join"+")}
&$a 'abcd'

This would save a few character from declaring a function with a name and goes better if used more than once. The one above converts a string into a char array then into an int array. Then creates a string joined with + so that Invoke-Expression will add all the values.

Answer 37

Use $x-shr$k if you want to get the result of \$ \big\lfloor{x\over2^k}\big\rfloor \$.

The biggest application of this is getting the truncated int result of halving a positive number without having to deal with Banker's rounding.

Works on ints and floats, however it will cast floats to ints before shifting, resulting in the same banker problem as before. Whether or not this is acceptable will obviously depend. Below is a sample script that shows the results of -5..10. 3/2 and 7/2 demonstrate the difference in return value.

Try it online!

^{Inspired by this answer}

Answer 38

Get Length of elements of an array

You can write the name immediately after the point to get the value of the property or method.

$a=gci
$a.fullName

compare to:

$a|%{$_.fullName}

Both expressions returns array like this:

C:\Archive
C:\PerfLogs
C:\Pictures
C:\Program Files
C:\Program Files (x86)
C:\Users
C:\Windows

There is a property Length which is defined for the array itself. So the Powershell expression $a.Length returns the number of elements in an array. You can use the Shortening Names to get length of elements:

$s=@('22','55555','7777777')
$s|% Length       # 11 bytes

or shortcut:

$s|% Le*          # 8 bytes.

Compare to:

$s|%{$_.Length}   # 15 bytes

Result is the array 2,5,7, not the number 3.

Answer 39

Use Uppercase over lowercase when dealing with ASCII values

When given the choice, it's usually better to deal entirely with Uppercase. This is because their range is 65..90 vs lowercase's 97..122. For example, the one-liner to generate ABCD...Z in PS v5

-join([char[]](65..90))
-join([char[]](97..122))

This still applies when normalizing mix-case because using the |% short-property trick, ToUpper and ToLower shorten to *per and *wer, respectively.

"Lower to upper: "+-join([char[]](97..122))|% *per
"Upper to lower: "+-join([char[]](65..90))|% *wer

Try it online!

Answer 40

Use the new() method instead of New-Object

For example, instead of doing the following (37 bytes):

New-Object Drawing.Bitmap(1024,1024)

Try (32 bytes):

[Drawing.Bitmap]::new(1024,1024)

Kind of an edge case, but you can do this pretty much anytime you initialize a new .NET object. Sadly, there's no equivalent for COM objects so you still have to use New-Object for those :P

Answer 41

Short way to convert bits to integer:

if you trust that $args contains int only (see Joey's comments below):

$args|%{$r+=$_+$r};$r

if you assume that $args is an array of int or string representation:

$args|%{$r+=+$r+$_};$r

$args is an array of char 48, 49:

$args|%{$r+=+$r+"$_"};$r

A sliding byte containing 8 bits from the array of int:

$args|%{($r=2*$r%256+$_)}

Compare to:

[Convert]::ToInt64($args,2)

Answer 42

If you're calling a command with a string argument, you don't need quotes unless there are spaces:

gci MyPath
gci 'My Path'

But that space cost you 2 extra bytes. You can save a single byte by escaping the space without quotes:

gci My` Path

This is only worth it when there's a single space, because there's no advantage with 2 or more.

Answer 43

If a path contains more than 1 space, you can do :

gci C:\My*Example*Path

instead of

gci 'C:\My Example Path'

or

gci C:\My` Example` Path