Sort-Object not sorting correctly due to encoding issue

Question

When I run the following code:

'Windows Embedded Standard',
'Windows 7 Enterprise',
'Windows XP Professional',
'Windows Server 2003',
'Windows 7 Entreprise',
'',
'Windows 7 Professionnel',
'Windows 7 Professional',
'Windows 10 Enterprise',
'Windows Server 2008 R2 Standard',
'Windows Server 2012 Standard',
'Windows Server 2012 R2 Standard',
'unknown',
'Windows Server 2008 R2 Enterprise' | Sort-Object

It sorts as following:

unknown
Windows 10 Enterprise
Windows 7 Enterprise
Windows 7 Professional
Windows 7 Professionnel
Windows Embedded Standard
Windows Server 2003
Windows Server 2008 R2 Enterprise
Windows Server 2008 R2 Standard
Windows Server 2012 R2 Standard
Windows Server 2012 Standard
Windows XP Professional
Windows 7 Entreprise

I can't figure out why the last entry Windows 7 Entreprise is not correctly sorted. It should come in the 4th position. What am I missing here?

UPDATE

Thanks to the comments it became clear that the source data seems to be the problem:

$Computers = Get-ADComputer -SearchBase 'OU=EU,DC=domain,DC=net' -Filter * -Properties OperatingSystem
$Computers | Group-Object OperatingSystem | Sort-Object Name | Select-Object Count, Name | Format-Table -AutoSize

ANSI Values:

$Computers | Group-Object OperatingSystem | ForEach-Object{$_.Name;[int[]][char[]]$_.Name -join "|"}
Windows Server 2008 R2 Standard
87|105|110|100|111|119|115|32|83|101|114|118|101|114|32|50|48|48|56|32|82|50|32|83|116|97|110|100|97|114|100
Windows Server 2003
87|105|110|100|111|119|115|32|83|101|114|118|101|114|32|50|48|48|51
Windows Server 2012 Standard
87|105|110|100|111|119|115|32|83|101|114|118|101|114|32|50|48|49|50|32|83|116|97|110|100|97|114|100


Windows Server 2012 R2 Standard
87|105|110|100|111|119|115|32|83|101|114|118|101|114|32|50|48|49|50|32|82|50|32|83|116|97|110|100|97|114|100
unknown
117|110|107|110|111|119|110
Windows Server 2008 R2 Enterprise
87|105|110|100|111|119|115|32|83|101|114|118|101|114|32|50|48|48|56|32|82|50|32|69|110|116|101|114|112|114|105|115|101
Windows 7 Enterprise
87|105|110|100|111|119|115|32|55|32|69|110|116|101|114|112|114|105|115|101
Windows 7 Entreprise
87|105|110|100|111|119|115|160|55|32|69|110|116|114|101|112|114|105|115|101
Windows 7 Professionnel
87|105|110|100|111|119|115|32|55|32|80|114|111|102|101|115|115|105|111|110|110|101|108
Windows XP Professional
87|105|110|100|111|119|115|32|88|80|32|80|114|111|102|101|115|115|105|111|110|97|108
Windows 7 Professional
87|105|110|100|111|119|115|32|55|32|80|114|111|102|101|115|115|105|111|110|97|108
Windows 10 Enterprise
87|105|110|100|111|119|115|32|49|48|32|69|110|116|101|114|112|114|105|115|101
Windows Embedded Standard
87|105|110|100|111|119|115|32|69|109|98|101|100|100|101|100|32|83|116|97|110|100|97|114|100

Is there a way to encode it differently so PowerShell can sort this correctly?

Answer 1

Use the String.Normalize() function with Normalization Form KC or KD (documented here):

$Computers = Get-ADComputer -SearchBase 'OU=EU,DC=domain,DC=net' -Filter * -Properties OperatingSystem
$OSList = $Computers | Select-Object -ExpandProperty OperatingSystem | ForEach-Object {
    $_.ToString().Normalize([Text.NormalizationForm]::FormKC)
}

Note that the Normalize function will remote diacritics such as accents and graves. For example, á will beome a. In order to convert character 160 to character 32, I had to use one of the normalization forms that used "full compatibility decomposition," which is either the KC or KD form.