The program inv.exe returns some console data based on parameters. It seems like a JSON/dictionary, but it's in text format (printed output). It works when I simply call it without trying to capture the output.

.\inv.exe getter segments
{28: 'Renda Fixa', 29: 'Renda Variável', ...

However, if I try to capture it, it doesn't work:

$segmentsjson = .\inv.exe getter segments
$segmentsjson
{28: 'Renda Fixa', 29: 'Renda Vari�vel'....

$segmentsjson = .\inv.exe getter segments | ConvertFrom-Json
$segmentsjson
{"28": "Renda Fixa", "29": "Renda Variável"...

What I tried:

1. chcp 65001

2. $OutputEncoding = [System.Text.Encoding]::UTF8

3. [Console]::OutputEncoding = [System.Text.Encoding]::UTF8

4. $OutputEncoding = [Console]::InputEncoding = [Console]::OutputEncoding = New-Object System.Text.UTF8Encoding

5. .\inv.exe getter segments > test.txt
$segmentsjson = Get-Content "test.txt" -Encoding UTF8

6. .\inv.exe getter segments | Out-File -FilePath "output_temp.txt" -Encoding UTF-8

7. cmd /c inv.exe getter segments > test.txt

The program inv.exe returns some console data based on parameters. It seems like a JSON/dictionary, but it's in text format (printed output). It works when I simply call it without trying to capture the output.

.\inv.exe getter segments
{28: 'Renda Fixa', 29: 'Renda Variável', ...

However, if I try to capture it, it doesn't work:

$segmentsjson = .\inv.exe getter segments
$segmentsjson
{28: 'Renda Fixa', 29: 'Renda Vari�vel'....

$segmentsjson = .\inv.exe getter segments | ConvertFrom-Json
$segmentsjson
{"28": "Renda Fixa", "29": "Renda Variável"...

What I tried:

1. chcp 65001

2. $OutputEncoding = [System.Text.Encoding]::UTF8

3. [Console]::OutputEncoding = [System.Text.Encoding]::UTF8

4. $OutputEncoding = [Console]::InputEncoding = [Console]::OutputEncoding = New-Object System.Text.UTF8Encoding

5. .\inv.exe getter segments > test.txt
$segmentsjson = Get-Content "test.txt" -Encoding UTF8

6. .\inv.exe getter segments | Out-File -FilePath "output_temp.txt" -Encoding UTF-8

7. cmd /c inv.exe getter segments > test.txt

• powershell
• printing
• utf-8
• output
• stdout

Share Improve this question Follow asked Nov 19, 2024 at 14:10 LegsNotHandLegsNotHand 2344 bronze badges 2

• If you have the chance to test with pwsh 7 probably wouldn't have this issue. A few things you can test, 1. change the encoding of your ps1 script file. 2. try with utf8 with a BOM ($OutputEncoding = [Console]::InputEncoding = [Console]::OutputEncoding = [System.Text.UTF8Encoding]::new($true) – Santiago Squarzon Commented Nov 19, 2024 at 14:30
• Some native programs behave differently when redirected. When you redirect to a file, using cmd shell, is the file actually encoded in UTF-8? inv.exe getter segments >file.txt. Then inspect the encoding using Notepad++ for instance. – zett42 Commented Nov 19, 2024 at 16:39

Add a comment  | 

1 Answer 1

Sorted by: Reset to default 2

Character-encoding problems may only surface if external-program output is captured or redirected in PowerShell on Windows, because some CLIs - including high-profile ones such as python.exe and node.exe - use the Unicode version of the WriteConsole WinAPI function when printing to the console, where all characters print as intended.^[1]

PowerShell indeed uses [Console]::OutputEncoding when decoding external-program output into .NET strings (System.String ([string], in PowerShell terms), which internally uses a Unicode encoding composed of in-memory UTF-16 code units (System.Char ([char])).

If [Console]::OutputEncoding = [System.Text.Encoding]::UTF8 doesn't help, the implication is that inv.exe's output isn't UTF-8.

Thus, you must (temporarily) set [Console]::OutputEncoding to match the actual character encoding inv.exe uses, which looks to be the legacy system locale's active ANSI encoding, presumably Windows-1252.

The following code temporarily sets [Console]::OutputEncoding to the active ANSI code page's, calls .inv.exe, then restores the original encoding:

$segmentsjson = 
  & {
    $prevEnc = [Console]::OutputEncoding
    # Set [Console]::OutputEncoding to that of the system's active ANSI code page.
    [Console]::OutputEncoding = 
      if ($IsCoreCLR) { [Text.Encoding]::GetEncoding([int] (Get-ItemPropertyValue registry::HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Nls\CodePage ACP)) } 
      else { [Text.Encoding]::Default }

    .\inv.exe getter segments

    [Console]::OutputEncoding = $prevEnc
  }

Note:

• Obtaining the encoding for the active ANSI code page is simpler in Windows PowerShell (the legacy, ships-with-Windows, Windows-only edition of PowerShell whose latest and last version is 5.1), via [Text.Encoding]::Default, compared to PowerShell (Core) 7, (detected via $IsCoreClr being $true) where [Text.Encoding]::Default reports UTF-8, and the registry must be consulted for the ANSI code page number).

• For additional information and helper functions, see this answer; with the Invoke-WithEncoding helper function from the linked answer, the solution would be as simple as:

  $segmentsJson = 
    Invoke-WithEncoding -Encoding ANSI { .\inv.exe getter segments }
  

• Alternatively, see if your CLI offers a parameter that would allow you to specify the desired output character encoding, such as UTF-8 (which is preferable, because it doesn't limit the set of characters that can be output).
  CLIs that use Python behind the scenes may switch to UTF-8 if $env:PYTHONUTF8 = 1 is in effect.

---

^{[1] Depending on the selected font, not all Unicode characters may render properly, but the console buffer does store them correctly, so you can copy and paste them without loss of information.}