8

I have many scripts. After making changes, I like to run them all to see if I broke anything. I wrote a script to loop through each, running it on fresh data.

Inside my loop I'm currently running powershell.exe -command <path to script>. I don't know if that's the best way to do this, or if the two instances are totally separate from each other.

What's the preferred way to run a script in a clean instance of PowerShell? Or should I be saying "session"?

Improve this question
3
  • What does "the best way to do this" mean? Without this information your approach with powershell.exe looks the best. It does exactly what you need, according to the question. Commented Nov 10, 2012 at 6:29
  • I assumed there was best practice. Your comment implies not, unless someone else knows of one. The powershell.exe way doesn't allow me to get the script's location via $MyInvocation.MyCommand.ScriptBlock.File, unless I was doing something wrong. Commented Nov 10, 2012 at 21:44
  • Actually, the MyInvocation thing might work... I could swear it didn't before. Now I'm just pulling my hair out trying to get it to accept paths with space AND not launch in a new window. I've put in 10+ hours on this, so any help is welcome. Commented Nov 10, 2012 at 22:01

4 Answers 4

Reset to default
8
+50

Using powershell.exe seems to be a good approach but with its pros and cons, of course.

Pros:

  • Each script is invoked in a separate clean session.
  • Even crashes do not stop the whole testing process.

Cons:

  • Invoking powershell.exe is somewhat slow.
  • Testing depends on exit codes but 0 does not always mean success.

None of the cons is mentioned is a question as a potential problem.

The demo script is below. It has been tested with PS v2 and v3. Script names may include special characters like spaces, apostrophes, brackets, backticks, dollars. One mentioned in comments requirement is ability to get script paths in their code. With the proposed approach scripts can get their own path as $MyInvocation.MyCommand.Path

# make a script list, use the full paths or explicit relative paths
$scripts = @(
    '.\test1.ps1' # good name
    '.\test 2.ps1' # with a space
    ".\test '3'.ps1" # with apostrophes
    ".\test [4].ps1" # with brackets
    '.\test `5`.ps1' # with backticks
    '.\test $6.ps1' # with a dollar
    '.\test ''3'' [4] `5` $6.ps1' # all specials
)

# process each script in the list
foreach($script in $scripts) {
    # make a command; mind &, ' around the path, and escaping '
    $command = "& '" + $script.Replace("'", "''") + "'"

    # invoke the command, i.e. the script in a separate process
    powershell.exe -command $command

    # check for the exit code (assuming 0 is for success)
    if ($LastExitCode) {
        # in this demo just write a warning
        Write-Warning "Script $script failed."
    }
    else {
        Write-Host "Script $script succeeded."
    }
}
Improve this answer
Sign up to request clarification or add additional context in comments.

3 Comments

This seems to work, even in my actual code! Thanks, Roman. I was escaping my quotes with back-ticks. I see you double-single-quoted. Is that being interpreted by PowerShell? Also, a note for future readers, if the "subscript" is a FileInfo object then you'll have to add .FullName before the Replace. PowerShell usually converts it automagically, but not in this case.
My first version before the edit was with backticks, too, so that it is possible as well. But the version with single quotes seems to be simpler. In single quoted strings single quotes themselves are escaped by doubling. See also help about_Quoting_Rules.
Ah, somehow I never knew about the quoting in single-quotes. I'll read up!
6

If you're on PowerShell 2.0 or higher, you can use jobs to do this. Each job runs in a separate PowerShell process e.g.:

$scripts = ".\script1.ps1", ".\script2.ps1"

$jobs = @()
foreach ($script in $scripts)
{
    $jobs += Start-Job -FilePath $script
}

Wait-Job $jobs

foreach ($job in $jobs)
{
    "*" * 60
    "Status of '$($job.Command)' is $($job.State)"
    "Script output:"
    Receive-Job $job
}

Also, check out the PowerShell Community Extensions. It has a Test-Script command that can detect syntax errors in a script file. Of course, it won't catch runtime errors.

Improve this answer

2 Comments

I like this answer, but { & $script } causes errors later, when running Receive-Job: "The expression after '&' in a pipeline element produced a object that was not valid. It must result in a command name, script block or CommandInfo object". I don't know what's wrong with the script block. It looks good to me. I discovered the -FilePath parameter for Start-Job, which seems to work: Start-Job -FilePath $script. I'll update the code example and mark the answer. But if you have any idea why the script block approach didn't work, I'm curious.
Also, my production scripts query their own paths and then use relative paths to find each other. I do this with Split-Path -Parent $MyInvocation.MyCommand.ScriptBlock.File. I think that's causing the error I'm now getting: "Cannot bind argument to parameter 'Path' because it is null". It's not clear because I get no filenames or line-numbers in errors from the jobs. I'm starting to think Jobs might not be the right tool for running arbitrary scripts. Let me know if you have any ideas.
3

One tip for PowerShell V3 users: we (the PowerShell team) added a new API on the Runspace class called ResetRunspace(). This API resets the global variable table back to the initial state for that runspace (as well as cleaning up a few other things). What it doesn't do is clean out function definitions, types and format files or unload modules. This allows the API to be much faster. Also note that the Runspace has to have been created using an InitialSessionState object, not a RunspaceConfiguration instance. ResetRunspace() was added as part of the Workflow feature in V3 to support parallel execution efficiently in a script.

Improve this answer

1 Comment

Interesting information. How do I use it? E.g. if I call [System.Management.Automation.Runspaces.Runspace]::DefaultRunspace.ResetRunspaceState() then it fails: Exception calling "ResetRunspaceState" with "0" argument(s): "Pipeline not run because a pipeline is already running. Pipelines cannot be run concurrently.
2

The two instances are totally separate, because they are two different processes. Generally, it is not the most efficient way to start a Powershell process for every script run. Depending on the number of scripts and how often you re-run them, it may be affecting your overall performance. If it's not, I would leave everything AS IS.

Another option would be to run in the same runspace (this is a correct word for it), but clean everything up every time. See this answer for a way to do it. Or use below extract:

$sysvars = get-variable | select -Expand name
function remove-uservars {
 get-variable |
   where {$sysvars -notcontains $_.name} |
     remove-variable
}
Improve this answer

7 Comments

JohnB@: question about your bounty - how does my solution don't seem to work for arbitrary scripts?
I think you suggested two things. Doing it the way I have been, which doesn't handle paths with spaces (I've been wrestling with parsing modes, script blocks, Start-Process, etc. and it's horrendous) and keeps launching a new window, making it hard to see the output in one stream or capture to file. The other way is to delete variables, but that doesn't cover functions, types, modules, and who knows what else. I could do some research to figure out what all can be globally created and then track it by hand, but that's a lot of work and it'll possibly change between versions.
Any help getting the first idea working would be greatly appreciated. I got the spaces thing working (I think) but the newly run PowerShell processes aren't outputting to screen.
@JohnB: Why does it not handle paths with spaces? Did you forget to include the path into double quotes by any chance? Another thing to try would be using Invoke-Command or Invoke-Expression. I don't think you need to account for anything but variables when you start another script. You will probably have 99% of test cases covered anyway. Things like functions should get out of scope, so no need to dispose of them.
Sorry, I edited my comment just now. I thought I had spaces working, but I didn't. Quotes don't work. Nested, escaped quotes don't work. Lots of others online have trouble with this, too. How should I use Invoke-Command and Invoke-Expression? There are global functions, which will invalidate testing. What if a subsequent script is missing an Import-Module? It'd work fine if a previous script loaded it, which is wrong. I'm looking for a general solution because this is useful stuff for people.
|

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.