PowerShell的推荐编码风格是什么?

时间:2010-01-08 06:55:53

标签: powershell coding-style

有没有推荐的编码风格如何编写PowerShell脚本?

关于如何构造代码(多少函数,如果使用模块,......)。它是关于'如何编写代码以便它可读'。

在编程语言中有一些推荐的编码样式缩进的内容,如何缩进 - 空格/制表符,在哪里制作新行 >,在哪里放大括号,...),但我还没有看到任何有关PowerShell的建议。

我特别感兴趣:


如何编写参数

function New-XYZItem
  ( [string] $ItemName
  , [scriptblock] $definition
  ) { ...

(我发现它更像是'V1'语法)

function New-PSClass  {
  param([string] $ClassName
       ,[scriptblock] $definition
  )...

或(为什么要添加空属性?)

function New-PSClass  {
  param([Parameter()][string] $ClassName
       ,[Parameter()][scriptblock] $definition
  )...

或(我在Jaykul的代码中看到的其他格式)

function New-PSClass {
  param(
        [Parameter()]
        [string]
        $ClassName
        ,
        [Parameter()]
        [scriptblock]
        $definition
  )...

或......?


如何编写复杂的管道

Get-SomeData -param1 abc -param2 xyz | % {
    $temp1 = $_
    1..100 | % {
      Process-somehow $temp1 $_
    }
  } | % {
    Process-Again $_
  } |
  Sort-Object -desc

或(新行上的cmdlet名称)

Get-SomeData -param1 abc -param2 xyz |
  % {
    $temp1 = $_
    1..100 |
      % {
        Process-somehow $temp1 $_
      }
  } |
  % {
    Process-Again $_
  } |
  Sort-Object -desc |

如果有-begin-process-end参数怎么办?如何使其最具可读性?

Get-SomeData -param1 abc -param2 xyz |
  % -begin {
     init
  } -process {
     Process-somehow2 ...
  } -end {
     Process-somehow3 ...
  } |
  % -begin {
  } ....

Get-SomeData -param1 abc -param2 xyz |
  %  `
    -begin {
      init
    } `
    -process {
      Process-somehow2 ...
    } `
    -end {
      Process-somehow3 ...
    } |
  % -begin {
  } ....

缩进在这里很重要,并且在新行上放置了什么元素。


我只是经常提到我脑海中浮现的问题。还有其他一些,但我想保持这个Stack Overflow问题“简短”。

欢迎任何其他建议。

3 个答案:

答案 0 :(得分:86)

在花了几年时间深入研究PowerShell v2.0之后,我已经解决了这个问题:

<#
.SYNOPSIS
Cmdlet help is awesome.  Autogenerate via a template so I never forget.

.DESCRIPTION
.PARAMETER
.PARAMETER
.INPUTS
.OUTPUTS
.EXAMPLE
.EXAMPLE
.LINK
#>
function Get-Widget
{
    [CmdletBinding()]
    param (
        # Think about which parameters users might loop over.  If there is a clear
        # favorite (80/20 rule), make it ValueFromPipeline and name it InputObject.
        [parameter(ValueFromPipeline=$True)]
        [alias("Server")]
        [string]$InputObject,

        # All other loop candidates are marked pipeline-able by property name.  Use Aliases to ensure the most 
        # common objects users want to feed in will "just work".
        [parameter(Mandatory=$true, Position=0, ValueFromPipelineByPropertyName=$True)]
        [alias("FullName")]
        [alias("Path")]
        [string[]]$Name,

        # Provide and document defaults for optional parameters whenever possible.
        [parameter(Position=1)]
        [int]$Minimum = 0,

        [parameter(Position=2)]
        [int]$ComputerName = "localhost",

        # Stick to standardized parameter names when possible.  *Especially* with switches.  Use Aliases to support 
        # domain-specific terminology and/or when you want to expose the parameter name of the .Net API you're wrapping.
        [parameter()]
        [Alias("IncludeFlibbles")]
        [switch]$All,
    )

    # The three main function blocks use this format if & only if they are short one-liners    
    begin { $buf = new-list string }

    # Otherwise they use spacing comparable to a C# method
    process    
    {
        # Likewise, control flow statements have a special style for one-liners
        try
        {
            # Side Note: internal variables (which may be inherited from a parent scope)  
            # are lowerCamelCase.  Direct parameters are UpperCamelCase.
            if ($All)
                { $flibbles = $Name | Get-Flibble }   
            elseif ($Minimum -eq 0)          
                { $flibbles = @() }
            else
                { return }                       

            $path = $Name |
                ? { $_.Length -gt $Minimum } |
                % { $InputObject.InvokeGetAPI($_, $flibbles) } |
                ConvertTo-FullPath
        }
        finally { Cleanup }

        # In general, though, control flow statements also stick to the C# style guidelines
        while($true)
        {
            Do-Something
            if ($true)
            {
                try
                {
                    Do-Something
                    Do-Something
                    $buf.Add("abc")
                }
                catch
                {
                    Do-Something
                    Do-Something
                }
            }            
        }    
    }    
}

<# 
Pipelines are a form of control flow, of course, and in my opinion the most important.  Let's go 
into more detail.

I find my code looks more consistent when I use the pipeline to nudge all of PowerShell's supported 
language constructs (within reason) toward an "infix" style, regardless of their legacy origin.  At the 
same time, I get really strict about avoiding complexity within each line.  My style encourages a long,
consistent "flow" of command-to-command-to-command, so we can ensure ample whitespace while remaining
quite compact for a .NET language. 

Note - from here on out I use aliases for the most common pipeline-aware cmdlets in my stable of 
tools.  Quick extract from my "meta-script" module definition:
sal ?? Invoke-Coalescing
sal ?: Invoke-Ternary
sal im Invoke-Method
sal gpv Get-PropertyValue
sal spv Set-PropertyValue
sal tp Test-Path2
sal so Select-Object2        
sal eo Expand-Object        

% and ? are your familiar friends.
Anything else that begins with a ? is a pseudo-infix operator autogenerated from the Posh syntax reference.
#>        
function PipelineExamples
{
    # Only the very simplest pipes get to be one-liners:
    $profileInfo = dir $profile | so @{Path="fullname"; KBs={$_.length/1kb}}
    $notNull = $someString | ?? ""        
    $type = $InputObject -is [Type] | ?: $InputObject $InputObject.GetType()        
    $ComObject | spv Enabled $true
    $foo | im PrivateAPI($param1, $param2)
    if ($path | tp -Unc)
        { Do-Something }

    # Any time the LHS is a collection (i.e. we're going to loop), the pipe character ends the line, even 
    # when the expression looks simple.
    $verySlowConcat = ""            
    $buf |
        % { $verySlowConcat += $_ }
    # Always put a comment on pipelines that have uncaptured output [destined for the caller's pipeline]
    $buf |
        ? { $_ -like "*a*" }


    # Multi-line blocks inside a pipeline:
    $orders |
        ? { 
            $_.SaleDate -gt $thisQuarter -and
            ($_ | Get-Customer | Test-Profitable) -and
            $_.TastesGreat -and
            $_.LessFilling
        } |
        so Widgets |        
        % {                
            if ($ReviewCompetition)
            {
                $otherFirms |
                    Get-Factory |
                    Get-ManufactureHistory -Filter $_ |
                    so HistoryEntry.Items.Widgets                     
            }
            else
            {
                $_
            }
        } |            
        Publish-WidgetReport -Format HTML


    # Mix COM, reflection, native commands, etc. seamlessly
    $flibble = Get-WmiObject SomethingReallyOpaque |
        spv AuthFlags 0xf -PassThru |
        im Put() -PassThru |
        gpv Flibbles |
        select -first 1

    # The coalescing operator is particularly well suited to this sort of thing
    $initializeMe = $OptionalParam |
        ?? $MandatoryParam.PropertyThatMightBeNullOrEmpty |
        ?? { pwd | Get-Something -Mode Expensive } |
        ?? { throw "Unable to determine your blahblah" }           
    $uncFolderPath = $someInput |
        Convert-Path -ea 0 |
        ?? $fallback { tp -Unc -Folder }

    # String manipulation        
    $myName = "First{0}   Last{1}   " |
        ?+ "Suffix{2}" |
        ?replace "{", ": {" |
        ?f {eo richard berg jr | im ToUpper}            

    # Math algorithms written in this style start to approach the elegance of functional languages
    $weightedAvg = $values |
        Linq-Zip $weights {$args[0] * $args[1]} |
        Linq-Sum |
        ?/ ($weights | Linq-Sum)
}

# Don't be afraid to define helper functions.  Thanks to the script:Name syntax, you don't have to cram them into 
# the begin{} block or anything like that.  Name, parameters, etc don't always need to follow the cmdlet guidelines.
# Note that variables from outer scopes are automatically available.  (even if we're in another file!)
function script:Cleanup { $buf.Clear() }

# In these small helpers where the logic is straightforward and the correct behavior well known, I occasionally 
# condense the indentation to something in between the "one liner" and "Microsoft C# guideline" styles
filter script:FixComputerName
{
    if ($ComputerName -and $_) {            
        # Handle UNC paths 
        if ($_[1] -eq "\") {   
            $uncHost = ($_ -split "\\")[2]
            $_.Replace($uncHost, $ComputerName)
        } else {
            $drive = $_[0]
            $pathUnderDrive = $_.Remove(0,3)            
            "\\$ComputerName\$drive`$\$pathUnderDrive"
        }
    } else {
        $_
    }
}

Stack Overflow的语法高亮显示器完全放弃了我。将其粘贴到ISE中。

答案 1 :(得分:13)

我相信PowerShell最全面的编码风格资源仍然是The PowerShell Best Practices and Style Guide

从他们的介绍:

  

与英语拼写和语法规则一样,PowerShell编程最佳实践和样式规则几乎总是有例外,但我们正在记录代码结构,命令设计,编程,格式甚至样式的基线,这将有助于您避免常见问题,并帮助您编写更多可重用,可读的代码 - 因为不需要重写可重用代码,并且可以维护可读代码。

他们还提供了这些GitBook链接:

答案 2 :(得分:8)

我最近遇到了an excellent point about indent style in PowerShell。如链接注释所述,请观察这些相同语法之间的区别:

1..10 | Sort-Object
{
    -$_
}

1..10 | Sort-Object {
    -$_
}

虽然我倾向于“像罗马人那样做”并且使用标准的C#缩进样式(Allman,或多或少),但我对这个例外以及与之类似的其他例子有疑问。

这使我个人倾向于使用我喜欢的1TBS,但我可以说服不然。出于好奇,你是怎么解决的?