Windows PowerShell: Peering Through the Pipeline
Okay, PowerShell enthusiast. You say you're ready to embrace the "secret sauce" that gives Windows PowerShell its power? I've found that once an IT pro begins to understand the pipeline, his or her effectiveness with PowerShell makes dramatic gains.
By the time you finish reading this article, you should have the following PowerShell skills under your belt:
- Listing and using the properties and methods of PowerShell objects
- Intelligently piping the output of one PowerShell command to another
- Troubleshooting when PowerShell gives you unexpected output
Let's begin!
The 'Pipeline' in Cmd.exe
The Windows command processor, Cmd.exe, provides a tremendously rudimentary pipeline—specifically, what's called "standard in" (STDIN) and "standard out" (STDOUT). There's also a STDERR output, but we won't worry about that here.
In a limited number of cases, we can redirect the current command output (STDIN stream) to another output in addition to the screen (STDOUT stream):
dir "C:\Windows\System32" > "C:\sys32contents.txt"
The preceding example redirects a directory listing from the console to a text file. Perhaps you've done the following to display long directory listings one screen at a time:
dir "C:\Windows\System32" | more
While redirection like these examples can be useful, it's important to understand that in both Windows and Linux pipelines, you're dealing only with text—nothing more. To be sure, you can construct some complex pipelines in Linux. Imagine being able to do the following:
Generate a table showing the top five processes, ordered by CPU consumption in descending order, listing only the process name and CPU value.
Believe it or not, performing this task in Linux is quite difficult, involving a bunch of totally separate command-line utilities like ls, grep, awk, and sed. Even so, all you have in the pipeline is "dumb" text. To do the above in Windows without bringing VBScript to the table is inconceivable.
In PowerShell, Everything Is an Object
I can't stress enough the importance of understanding PowerShell's object orientation. We can look at an object as a "three-dimensional" data structure that has descriptive attributes (called properties), as well as actions that the object can perform (called methods).
In PowerShell, we pipe output to Get-Member to view the members of an object. Check this out:
Get-Service | Get-Member TypeName: System.ServiceProcess.ServiceController Name MemberType Definition ---- ---------- ---------- Name AliasProperty Name = ServiceName Close Method void Close() Pause Method void Pause() Refresh Method void Refresh() Start Method void Start(), void Start(string[] args) Stop Method void Stop() DependentServices Property System.ServiceProcess.ServiceControl... DisplayName Property string DisplayName {get;set;} MachineName Property string MachineName {get;set;} ServiceHandle Property System.Runtime.InteropServices.SafeH... ServiceName Property string ServiceName {get;set;} Status Property System.ServiceProcess.ServiceControl... ToString ScriptMethod System.Object ToString();
Don't be blown away by the member list. What you see above is actually a truncation of the full output. I just want you to get the feel for this now.
Let me draw your attention to the object data type that's produced by Get-Service:
System.ServiceProcess.ServiceController
IT pros tend to use the last part of an object name. Thus, we can say that Get-Service spits out ServiceController objects. Under the hood, ServiceController is a .NET Framework class. You can look it up on the Microsoft Developer Network (MSDN) website.
When you run Get-Service normally, you see a data table like the following (again, I'm truncating output to save space and spare your eyes):
Get-Service Status Name DisplayName ------ ---- ----------- Running AdobeARMservice Adobe Acrobat Update Service Stopped AeLookupSvc Application Experience Stopped ALG Application Layer Gateway Service Running AMD External Ev... AMD External Events Utility Stopped AppIDSvc Application Identity Running Appinfo Application Information Running Apple Mobile De... Apple Mobile Device Service
This is a list of ServiceController objects that the Microsoft folks call a collection. The three columns (Status, Name, and DisplayName) are selected properties from each ServiceController object. The output shows seven ServiceController instances from the entire collection. Do you see how the vocabulary fits together?
'Bookend' Cmdlets
In my opinion, the easiest way to start understanding the PowerShell pipeline is to combine what I call "bookend" cmdlets, which operate directly on the same kinds of objects. For instance, let's say that we want to stop the Spooler service on our computer. We can do this with one line of PowerShell code:
Get-Service -Name spooler | Stop-Service
Likewise, we can start the service:
Get-Service -Name spooler | Start-Service
Interesting. In this case, we already know what object type is produced by Get-Service, but if you aren't sure, you can use the GetType() built-in method:
$spool = Get-Service -Name spooler $spool.GetType() IsPublic IsSerial Name BaseType -------- -------- ---- -------- True False ServiceController System.ComponentM...
Let's turn to the Stop-Service help to investigate how this cmdlet might accept inbound ServiceController objects:
Get-Help Stop-Service -Parameter InputObject -InputObject <ServiceController[]> Specifies ServiceController objects representing the services to be stopped. Enter a variable that contains the objects, or type a command or expression that gets the objects. Required? true Position? 1 Default value Accept pipeline input? true (ByValue) Accept wildcard characters? false
Admittedly, I'm "cherry picking" my attributes. I knew in advance that Stop-Service had a parameter named InputObject that accepts ServiceController objects. But the point is the same.
I want you to notice this line:
-InputObject <ServiceController[]>
What this tells us is that the Stop-Service parameter named InputObject expects ServiceController objects as input. Now look at the following line in the parameter's attribute table:
Accept pipeline input? true (ByValue)
This is a crucial parameter attribute to understand when we face the PowerShell pipeline. We read that InputObject can in fact accept input (not all parameters can), and it does so ByValue, which really means "by object type." Thus, we see that feeding ServiceController objects to Get-Service fits hand-in-glove with the Stop-Service and Start-Service cmdlets by virtual of the target cmdlet's InputType parameter.
More About Parameter Binding
If you understand what I've explained thus far, you are most of the way toward mastering the Windows PowerShell pipeline. We've learned that PowerShell uses parameter binding to take input from the previous cmdlet and feed it to the next cmdlet in the pipeline.
PowerShell takes care of the binding itself, although you can "help it along" manually if necessary. Although I've found that the ByValue binding type is by far the most common, some cmdlet parameters can accept input ByPropertyName.
I've discovered that PowerShell tends to fall back on PropertyName binding when or if it encounters problems in doing so by value. When I say that PowerShell attempts a bind by property name, I mean that the receiving cmdlet only looks for an output data field (property) with a name that's understandable. Yes, it's that simple!
Let's try a practical example. Create a simple comma-separated value (CSV) file named spooler.csv that contains the following data:
SvcName,SvcStatus spooler,running
Now let's import the CSV and feed our spooler service data to Stop-Service:
Import-Csv -Path D:\spooler.csv | Stop-Service -PassThru
What happened? You should have seen what Windows PowerShell MVP Jason Helmick calls "blood on the screen." In other words, PowerShell first looked to bind on a ServiceController object; failing that, it looked for an output object property called "Name." No go there, either, so we are hosed.
Let's look at the -Name parameter of the Stop-Service cmdlet:
Get-Help Stop-Service -Parameter Name -Name <String[]> Specifies the service names of the services to be stopped. Wildcards are permitted. The parameter name is optional. You can use "Name" or its alias, "ServiceName", or you can omit the parameter name. Required? true Position? 1 Default value Accept pipeline input? true (ByPropertyName, ByValue) Accept wildcard characters? true
By investigating the previous output, we see that the Stop-Service -Name parameter a) expects to receive string values; and b) can accept pipeline input either by property name or by value.
I can tell you that the Import-Csv import gives you a PSCustomObject collection that consists of (drum roll, please) string data. Therefore, to fix the problem, we simply need to rename the column heading in our CSV file:
Name,SvcStatus spooler,running
Done and done.
Troubleshooting and Conclusion
I'm rapidly running out of "white space," so I need to wrap this up. PowerShell MVP Don Jones gave me the best advice concerning how to troubleshoot unexpected pipeline output. He said that you should a) remove the last element from the pipeline; and b) run Get-Member to see what kind of object you're getting.
For example, the following pipeline doesn't give me the CSV I want:
Get-Service | Format-Table -Property ServiceName, Status | Export-Csv -Path "D:\services.csv"
When I open the CSV, all I see is a bunch of formatting codes. Blech! Let's remove the last part of the pipeline and run a Get-Member. Now I get two .NET object classes:
Microsoft.PowerShell.Commands.Internal.Format.GroupEndData Microsoft.PowerShell.Commands.Internal.Format.FormatEndData
Aha! So this is why Don Jones taught me to "filter left, format right"!
I hope you found this article informative. Happy PowerShelling!