Integrate Linux Commands into Windows with PowerShell and the Windows Subsystem for Linux
The result of these shortcomings is that Linux commands feel like second-class citizens to Windows and are harder to use than they should be. For a command to feel like a native Windows command, we’ll need to address these issues.
PowerShell Function Wrappers
We can remove the need to prefix commands with wsl , handle the translation of Windows paths to WSL paths, and support command completion with PowerShell function wrappers. The basic requirements of the wrappers are:
- There should be one function wrapper per Linux command with the same name as the command
- The wrapper should recognize Windows paths passed as arguments and translate them to WSL paths
- The wrapper should invoke wsl with the corresponding Linux command, piping in any pipeline input and passing on any command line arguments passed to the function
Since this template can be applied to any command, we can abstract the definition of these wrappers and generate them dynamically from a list of commands to import.
# The commands to import. $commands = "awk", "emacs", "grep", "head", "less", "ls", "man", "sed", "seq", "ssh", "tail", "vim" # Register a function for each command. $commands | ForEach-Object < Invoke-Expression @" Remove-Alias $_ -Force -ErrorAction Ignore function global:$_() < for (`$i = 0; `$i -lt `$args.Count; `$i++) < # If a path is absolute with a qualifier (e.g. C:), run it through wslpath to map it to the appropriate mount point. if (Split-Path `$args[`$i] -IsAbsolute -ErrorAction Ignore) < `$args[`$i] = Format-WslArgument (wsl.exe wslpath (`$args[`$i] -replace "\\", "/")) # If a path is relative, the current working directory will be translated to an appropriate mount point, so just format it. >elseif (Test-Path `$args[`$i] -ErrorAction Ignore) < `$args[`$i] = Format-WslArgument (`$args[`$i] -replace "\\", "/") >> if (`$input.MoveNext()) < `$input.Reset() `$input | wsl.exe $_ (`$args -split ' ') >else < wsl.exe $_ (`$args -split ' ') >> "@ > The $command list defines the commands to import. Then we dynamically generate the function wrapper for each using the Invoke-Expression command (first removing any aliases that would conflict with the function).
The function loops through the command line arguments, identifies Windows paths using the Split-Path and Test-Path commands, then converts those paths to WSL paths. We run the paths through a helper function we’ll define later called Format-WslArgument that escapes special characters like spaces and parentheses that would otherwise be misinterpreted.
Finally, we pass on pipeline input and any command line arguments through to wsl .
With these function wrappers in place, we can now call our favorite Linux commands in a more natural way without having to prefix them with wsl or worry about how Windows paths are translated to WSL paths:
- man bash
- less -i $profile.CurrentUserAllHosts
- ls -Al C:\Windows\ | less
- grep -Ein error *.log
- tail -f *.log
A starter set of commands is shown here, but you can generate a wrapper for any Linux command simply by adding it to the list. If you add this code to your PowerShell profile, these commands will be available to you in every PowerShell session just like native commands!
Default Parameters
It is common in Linux to define aliases and/or environment variables within login profiles to set default parameters for commands you use frequently (e.g. alias ls=ls -AFh or export LESS=-i ). One of the drawbacks of proxying through a non-interactive shell via wsl.exe is that login profiles are not loaded, so these default parameters are not available (i.e. ls within WSL and wsl ls would behave differently with the alias defined above).
PowerShell provides $PSDefaultParameterValues , a standard mechanism to define default parameter values, but only for cmdlets and advanced functions. Turning our function wrappers into advanced functions is possible but introduces complications (e.g. PowerShell matches partial parameter names (like matching -a for -ArgumentList ) which will conflict with Linux commands that accept the partial names as arguments), and the syntax for defining default values would be less than ideal for this scenario (requiring the name of a parameter in the key for defining the default arguments as opposed to just the command name).
With a small change to our function wrappers, we can introduce a model similar to $PSDefaultParameterValues and enable default parameters for Linux commands!
By passing $WslDefaultParameterValues down into the command line we send through wsl.exe , you can now add statements like below to your PowerShell profile to configure default parameters!
$WslDefaultParameterValues["grep"] = "-E" $WslDefaultParameterValues["less"] = "-i" $WslDefaultParameterValues["ls"] = "-AFh --group-directories-first" Since this is modeled after $PSDefaultParameterValues , you can temporarily disable them easily by setting the «Disabled» key to $true . A separate hash table has the additional benefit of being able to disable $WslDefaultParameterValues separately from $PSDefaultParameterValues .
Argument Completion
PowerShell allows you to register argument completers with the Register-ArgumentCompleter command. Bash has powerful programmable completion facilities. WSL lets you call into bash from PowerShell. If we can register argument completers for our PowerShell function wrappers and call through to bash to generate the completions, we can get rich argument completion with the same fidelity as within bash itself!
# Register an ArgumentCompleter that shims bash's programmable completion. Register-ArgumentCompleter -CommandName $commands -ScriptBlock < param($wordToComplete, $commandAst, $cursorPosition) # Map the command to the appropriate bash completion function. $F = switch ($commandAst.CommandElements[0].Value) < < "_longopt" break >"man" < "_man" break >"ssh" < "_ssh" break >Default < "_minimal" break >> # Populate bash programmable completion variables. $COMP_LINE = "`"$commandAst`"" $COMP_WORDS = "('$($commandAst.CommandElements.Extent.Text -join "' '")')" -replace "''", "'" for ($i = 1; $i -lt $commandAst.CommandElements.Count; $i++) < $extent = $commandAst.CommandElements[$i].Extent if ($cursorPosition -lt $extent.EndColumnNumber) < # The cursor is in the middle of a word to complete. $previousWord = $commandAst.CommandElements[$i - 1].Extent.Text $COMP_CWORD = $i break >elseif ($cursorPosition -eq $extent.EndColumnNumber) < # The cursor is immediately after the current word. $previousWord = $extent.Text $COMP_CWORD = $i + 1 break >elseif ($cursorPosition -lt $extent.StartColumnNumber) < # The cursor is within whitespace between the previous and current words. $previousWord = $commandAst.CommandElements[$i - 1].Extent.Text $COMP_CWORD = $i break >elseif ($i -eq $commandAst.CommandElements.Count - 1 -and $cursorPosition -gt $extent.EndColumnNumber) < # The cursor is within whitespace at the end of the line. $previousWord = $extent.Text $COMP_CWORD = $i + 1 break >> # Repopulate bash programmable completion variables for scenarios like '/mnt/c/Program Files'/ where should continue completing the quoted path. $currentExtent = $commandAst.CommandElements[$COMP_CWORD].Extent $previousExtent = $commandAst.CommandElements[$COMP_CWORD - 1].Extent if ($currentExtent.Text -like "/*" -and $currentExtent.StartColumnNumber -eq $previousExtent.EndColumnNumber) < $COMP_LINE = $COMP_LINE -replace "$($previousExtent.Text)$($currentExtent.Text)", $wordToComplete $COMP_WORDS = $COMP_WORDS -replace "$($previousExtent.Text) '$($currentExtent.Text)'", $wordToComplete $previousWord = $commandAst.CommandElements[$COMP_CWORD - 2].Extent.Text $COMP_CWORD -= 1 ># Build the command to pass to WSL. $command = $commandAst.CommandElements[0].Value $bashCompletion = ". /usr/share/bash-completion/bash_completion 2> /dev/null" $commandCompletion = ". /usr/share/bash-completion/completions/$command 2> /dev/null" $COMPINPUT = "COMP_LINE=$COMP_LINE; COMP_WORDS=$COMP_WORDS; COMP_CWORD=$COMP_CWORD; COMP_POINT=$cursorPosition" $COMPGEN = "bind `"set completion-ignore-case on`" 2> /dev/null; $F `"$command`" `"$wordToComplete`" `"$previousWord`" 2> /dev/null" $COMPREPLY = "IFS=`$'\n'; echo `"`$`"" $commandLine = "$bashCompletion; $commandCompletion; $COMPINPUT; $COMPGEN; $COMPREPLY" -split ' ' # Invoke bash completion and return CompletionResults. $previousCompletionText = "" (wsl.exe $commandLine) -split '\n' | Sort-Object -Unique -CaseSensitive | ForEach-Object < if ($wordToComplete -match "(.*=).*") < $completionText = Format-WslArgument ($Matches[1] + $_) $true $listItemText = $_ >else < $completionText = Format-WslArgument $_ $true $listItemText = $completionText >if ($completionText -eq $previousCompletionText) < # Differentiate completions that differ only by case otherwise PowerShell will view them as duplicate. $listItemText += ' ' >$previousCompletionText = $completionText [System.Management.Automation.CompletionResult]::new($completionText, $listItemText, 'ParameterName', $completionText) > > # Helper function to escape characters in arguments passed to WSL that would otherwise be misinterpreted. function global:Format-WslArgument([string]$arg, [bool]$interactive) < if ($interactive -and $arg.Contains(" ")) < return "'$arg'" >else < return ($arg -replace " ", "\ ") -replace "([()|])", ('\$1', '`$1')[$interactive] >> The code is a bit dense without an understanding of some bash internals, but basically:
- We register the argument completer for all of our function wrappers by passing the $commands list to the -CommandName parameter of Register-ArgumentCompleter
- We map each command to the shell function bash uses to complete for it ( $F which is named after complete -F used to define completion specs in bash)
- We convert PowerShell’s $wordToComplete , $commandAst , and $cursorPosition arguments into the format expected by bash completion functions per the bash programmable completion spec
- We build a command line that we can pass to wsl.exe that ensures the completion environment is set up correctly, invokes the appropriate completion function, then outputs a string containing the completion results separated by new lines
- We then invoke wsl with the command line, split the output string on the new line separator, then generate CompletionResults for each, sorting them, and escaping characters like spaces and parentheses that would otherwise be misinterpreted
The end result of this is now our Linux command wrappers will use the exact same completion that bash uses! For example:
Each completion will provide values specific to the argument before it, reading in configuration data like known hosts from within WSL!
will cycle through options. will show all available options.
Additionally, since bash completion is now in charge, you can resolve Linux paths directly within PowerShell!
In cases where bash completion doesn’t return any results, PowerShell falls back to its default completion which will resolve Windows paths, effectively enabling you to resolve both Linux paths and Windows paths at will.
Conclusion
With PowerShell and WSL, we can integrate Linux commands into Windows just as if they were native applications. No need to hunt around for Win32 builds of Linux utilities or be forced to interrupt your workflow to drop into a Linux shell. Just install WSL, set up your PowerShell profile, and list the commands you want to import! The rich argument completion shown here of both command options and Linux and Windows file paths is an experience even native Windows commands don’t provide today.
The complete source code described above as well as additional guidance for incorporating it into your workflow is available at https://github.com/mikebattista/PowerShell-WSL-Interop.
Which Linux commands do you find most useful? What other parts of your developer workflow do you find lacking on Windows?
Let us know in the comments below or over on GitHub!
How to run linux terminal commands on windows?
I’m not sure how to ask, but I’d like run the ‘bash’ command on windows 10 so that some linux commands run later. I’m using the framework Electron and the Child Process.
var os = require('os') var exec = require('child_process').exec if (os.platform() =='win32')< var cmd_win = 'bash' exec(cmd_win, function(error, stdout, stderr)< console.log(error) >); > The code snippet gives «Error: Command failed: bash». Does anyone know why? And can you help me? I hope you understood my question.
3 Answers 3
To initialize the WSL subsystem, you must launch a (hidden) Bash console window in the background, which doesn’t work if you execute bash.exe directly — it works with neither exec nor execFile .
The trick is to get the shell ( cmd ) process that Node.js spawns to launch bash.exe without blocking, which, unfortunately, isn’t easy to do: start cannot be used, because bash.exe is a console application and therefore makes start act synchronously.
The solution is to create an aux. VBScript file that launches bash.exe , which itself can be invoked asynchronously via wscript.exe . Note that the Bash console window is launched hidden:
var os = require('os') var exec = require('child_process').exec if (os.platform() === 'win32') < var cmd_win = '\ echo WScript.CreateObject("Shell.Application").\ ShellExecute "bash", "", "", "open", 0 >%temp%\launchBashHidden.vbs \ & wscript %temp%\launchBashHidden.vbs' exec(cmd_win, function(error, stdout, stderr)< if (error) console.error(error) >); > Note that the aux. VBScript file %temp%\launchBashHidden.vbs lingers between invocations. Cleaning it up after every run would require more work (you can’t just delete it right away, because wscript , due to running asynchronously, may not have loaded it yet).