Using PowerShell to write a file in UTF-8 without the BOM

Question

Out-File seems to force the BOM when using UTF-8:

$MyFile = Get-Content $MyPath
$MyFile | Out-File -Encoding "UTF8" $MyPath

How can I write a file in UTF-8 with no BOM using PowerShell?

Update 2021

PowerShell has changed a bit since I wrote this question 10 years ago. Check multiple answers below, they have a lot of good information!

Solution 1

Using .NET's UTF8Encoding class and passing $False to the constructor seems to work:

$MyRawString = Get-Content -Raw $MyPath
$Utf8NoBomEncoding = New-Object System.Text.UTF8Encoding $False
[System.IO.File]::WriteAllLines($MyPath, $MyRawString, $Utf8NoBomEncoding)

Solution 2

The proper way as of now is to use a solution recommended by @Roman Kuzmin in comments to @M. Dudley answer:

[IO.File]::WriteAllLines($filename, $content)

(I've also shortened it a bit by stripping unnecessary System namespace clarification - it will be substituted automatically by default.)

Solution 3

I figured this wouldn't be UTF, but I just found a pretty simple solution that seems to work...

Get-Content path/to/file.ext | out-file -encoding ASCII targetFile.ext

For me this results in a utf-8 without bom file regardless of the source format.

Solution 4

Note: This answer applies to Windows PowerShell; by contrast, in the cross-platform PowerShell Core edition (v6+), UTF-8 without BOM is the default encoding, across all cmdlets.

• In other words: If you're using PowerShell [Core] version 6 or higher, you get BOM-less UTF-8 files by default (which you can also explicitly request with -Encoding utf8 / -Encoding utf8NoBOM, whereas you get with-BOM encoding with -utf8BOM).

• If you're running Windows 10 and you're willing to switch to BOM-less UTF-8 encoding system-wide - which can have side effects - even Windows PowerShell can be made to use BOM-less UTF-8 consistently - see this answer.

---

To complement M. Dudley's own simple and pragmatic answer (and ForNeVeR's more concise reformulation):

For convenience, here's advanced function Out-FileUtf8NoBom, a pipeline-based alternative that mimics Out-File, which means:

• you can use it just like Out-File in a pipeline.
• input objects that aren't strings are formatted as they would be if you sent them to the console, just like with Out-File.
• an additional -UseLF switch allows you transform Windows-style CRLF newlines to Unix-style LF-only newlines.

Example:

(Get-Content $MyPath) | Out-FileUtf8NoBom $MyPath # Add -UseLF for Unix newlines

^{Note how (Get-Content $MyPath) is enclosed in (...), which ensures that the entire file is opened, read in full, and closed before sending the result through the pipeline. This is necessary in order to be able to write back to the same file (update it in place).
Generally, though, this technique is not advisable for 2 reasons: (a) the whole file must fit into memory and (b) if the command is interrupted, data will be lost.}

A note on memory use:

• M. Dudley's own answer requires that the entire file contents be built up in memory first, which can be problematic with large files.
• The function below improves on this only slightly: all input objects are still buffered first, but their string representations are then generated and written to the output file one by one.

---

Source code of function Out-FileUtf8NoBom:

Note: The function is also available as an MIT-licensed Gist, and only it will be maintained going forward.

You can install it directly with the following command (while I can personally assure you that doing so is safe, you should always check the content of a script before directly executing it this way):

# Download and define the function.
irm https://gist.github.com/mklement0/8689b9b5123a9ba11df7214f82a673be/raw/Out-FileUtf8NoBom.ps1 | iex

function Out-FileUtf8NoBom {
<#
.SYNOPSIS
  Outputs to a UTF-8-encoded file *without a BOM* (byte-order mark).
.DESCRIPTION
  Mimics the most important aspects of Out-File:
    * Input objects are sent to Out-String first.
    * -Append allows you to append to an existing file, -NoClobber prevents
      overwriting of an existing file.
    * -Width allows you to specify the line width for the text representations
       of input objects that aren't strings.
  However, it is not a complete implementation of all Out-File parameters:
    * Only a literal output path is supported, and only as a parameter.
    * -Force is not supported.
    * Conversely, an extra -UseLF switch is supported for using LF-only newlines.
  Caveat: *All* pipeline input is buffered before writing output starts,
          but the string representations are generated and written to the target
          file one by one.
.NOTES
  The raison d'être for this advanced function is that Windows PowerShell
  lacks the ability to write UTF-8 files without a BOM: using -Encoding UTF8 
  invariably prepends a BOM.
  Copyright (c) 2017, 2020 Michael Klement <[email protected]> (http://same2u.net), 
  released under the [MIT license](https://spdx.org/licenses/MIT#licenseText).
#>

  [CmdletBinding()]
  param(
    [Parameter(Mandatory, Position=0)] [string] $LiteralPath,
    [switch] $Append,
    [switch] $NoClobber,
    [AllowNull()] [int] $Width,
    [switch] $UseLF,
    [Parameter(ValueFromPipeline)] $InputObject
  )

  #requires -version 3

  # Convert the input path to a full one, since .NET's working dir. usually
  # differs from PowerShell's.
  $dir = Split-Path -LiteralPath $LiteralPath
  if ($dir) { $dir = Convert-Path -ErrorAction Stop -LiteralPath $dir } else { $dir = $pwd.ProviderPath}
  $LiteralPath = [IO.Path]::Combine($dir, [IO.Path]::GetFileName($LiteralPath))

  # If -NoClobber was specified, throw an exception if the target file already
  # exists.
  if ($NoClobber -and (Test-Path $LiteralPath)) {
    Throw [IO.IOException] "The file '$LiteralPath' already exists."
  }

  # Create a StreamWriter object.
  # Note that we take advantage of the fact that the StreamWriter class by default:
  # - uses UTF-8 encoding
  # - without a BOM.
  $sw = New-Object System.IO.StreamWriter $LiteralPath, $Append

  $htOutStringArgs = @{}
  if ($Width) {
    $htOutStringArgs += @{ Width = $Width }
  }

  # Note: By not using begin / process / end blocks, we're effectively running
  #       in the end block, which means that all pipeline input has already
  #       been collected in automatic variable $Input.
  #       We must use this approach, because using | Out-String individually
  #       in each iteration of a process block would format each input object
  #       with an indvidual header.
  try {
    $Input | Out-String -Stream @htOutStringArgs | % { 
      if ($UseLf) {
        $sw.Write($_ + "`n") 
      }
      else {
        $sw.WriteLine($_) 
      }
    }
  } finally {
    $sw.Dispose()
  }

}

Solution 5

Starting from version 6 powershell supports the UTF8NoBOM encoding both for set-content and out-file and even uses this as default encoding.

So in the above example it should simply be like this:

$MyFile | Out-File -Encoding UTF8NoBOM $MyPath

View more solutions

Comments

[Signal15]

BOM = Byte-Order Mark. Three chars placed at the beginning of a file (0xEF,0xBB,0xBF) that look like ""

[MichaelGG]

This is incredibly frustrating. Even third party modules get polluted, like trying to upload a file over SSH? BOM! "Yeah, let's corrupt every single file; that sounds like a good idea." -Microsoft.

[Paul Shiryaev]

The default encoding is UTF8NoBOM starting with Powershell version 6.0 docs.microsoft.com/en-us/powershell/module/…

[Dragas]

Talk about breaking backwards compatibility...

[Scott Muc]

Ugh, I hope that's not the only way.

[Roman Kuzmin]

One line [System.IO.File]::WriteAllLines($MyPath, $MyFile) is enough. This WriteAllLines overload writes exactly UTF8 without BOM.

[darksoulsong]

This one fails without any warning. What version of powershell should I use to run it?

[Groostav]

Created an MSDN feature request here: connect.microsoft.com/PowerShell/feedbackdetail/view/1137121‌​/…

[BermudaLamb]

The WriteAllLines solution works great for small files. However, I need a solution for larger files. Every time I try to use this with a larger file I'm getting an OutOfMemory error.

[sourcenouveau]

Per the Out-File documentation specifying the Default encoding will use the system's current ANSI code page, which is not UTF-8, as I required.

[ForNeVeR]

No, it will convert the output to current ANSI codepage (cp1251 or cp1252, for example). It is not UTF-8 at all!

[Greg]

Thanks Robin. This may not have worked for writing a UTF-8 file without the BOM but the -Encoding ASCII option removed the BOM. That way I could generate a bat file for gvim. The .bat file was tripping up on the BOM.

[mklement0]

@ForNeVeR: You're correct that encoding ASCII is not UTF-8, but it's als not the current ANSI codepage - you're thinking of Default; ASCII truly is 7-bit ASCII encoding, with codepoints >= 128 getting converted to literal ? instances.

[ForNeVeR]

@mklement0 AFAIK ASCII really mean the default single-byte encoding in this API and generally in Windows. Yes, it is not in sync with the official ASCII definition, but is just a historical legacy.

[mklement0]

@ForNeVeR: You're probably thinking of "ANSI" or "extended ASCII". Try this to verify that -Encoding ASCII is indeed 7-bit ASCII only: 'äb' | out-file ($f = [IO.Path]::GetTempFilename()) -encoding ASCII; '?b' -eq $(Get-Content $f; Remove-Item $f) - the ä has been transliterated to a ?. By contrast, -Encoding Default ("ANSI") would correctly preserve it.

[Liam]

This (for whatever reason) did not remove the BOM for me, where as the accepted answer did

[ForNeVeR]

@Liam, probably some old version of PowerShell or .NET?

[eythort]

This does seem to work for me, at least for Export-CSV. If you open the resulting file in a proper editor, the file encoding is UTF-8 without BOM, and not Western Latin ISO 9 as I would have expected with ASCII

[TNT]

@rob This is the perfect answer for everybody who just doesn't need utf-8 or anything else that is different to ASCII and is not interested in understanding encodings and the purpose of unicode. You can use it as utf-8 because the equivalent utf-8 characters to all ASCII characters are identical (means converting an ASCII-file to an utf-8-file results in an identical file (if it gets no BOM)). For all who have non-ASCII characters in their text this answer is just false and misleading.

[sschuberth]

Note that WriteAllLines seems to require $MyPath to be absolute.

[Just Rudy]

This worked for me, except I used -encoding utf8 for my requirement.

[codewario]

@sschuberth I just tried WriteAllLines with a relative path, works fine for me. Does it give you an error with a relative path?

[sschuberth]

@AlexanderMiles It "works", but the file ends up being in some weird directory (not relative to the current working directory). IIRC it was the path of the PowerShell interpreter binary.

[codewario]

I believe older versions of the .NET WriteAllLines function did write the BOM by default. So it could be a version issue.

[ForNeVeR]

@AlexanderMiles best I can tell from .NET 2.0 documentation, it still uses BOMless UTF-8 there.

[xdhmoore]

For me, it seems to write the file to my Desktop even if I'm currently in another directory.

[user1529294]

Thank you very much. I am working with dump logs of a tool - which had tabs inside it. UTF-8 was not working. ASCII solved the problem. Thanks.

[mklement0]

Yes, -Encoding ASCII avoids the BOM problem, but you obviously only get 7-bit ASCII characters. Given that ASCII is a subset of UTF-8, the resulting file is technically also a valid UTF-8 file, but all non-ASCII characters in your input will be converted to literal ? characters.

[Rosberg Linhares]

If you don't want an extra new line in the end of the file, you can do this: [IO.File]::WriteAllText($MyPath, $MyFile).

[emptyother]

Many editors open the file as UTF-8 if they can't detect the encoding.

[BobHy]

Can confirm this writes UTF8 no BOM on Win10 / .Net 4.6. But still needs an absolute path .

[Shayan Toqraee]

@xdhmoore WriteAllLines gets the current directory from [System.Environment]::CurrentDirectory. If you open PowerShell and then change your current directory (using cd or Set-Location), then [System.Environment]::CurrentDirectory will not be changed and the file will end up being in the wrong directory. You can work around this by [System.Environment]::CurrentDirectory = (Get-Location).Path.

[chazbot7]

Confirmed with writes with a BOM in Powershell 3, but without a BOM in Powershell 4. I had to use M. Dudley's original answer.

[Johny Skovdal]

So it works on Windows 10 where it's installed by default. :) Also, suggested improvement: [IO.File]::WriteAllLines(($filename | Resolve-Path), $content)

[mklement0]

Good pointers; suggestions/: the simpler alternative to $ExecutionContext.SessionState.Path.GetUnresolvedProviderPat‌​hFromPSPath($MyPath) is Convert-Path $MyPath; if you want to ensure a trailing CRLF, simply use [System.IO.File]::WriteAllLines() even with a single input string (no need for Out-String).

[mklement0]

Yes, -Encoding ASCII avoids the BOM problem, but you obviously only get support for 7-bit ASCII characters. Given that ASCII is a subset of UTF-8, the resulting file is technically also a valid UTF-8 file, but all non-ASCII characters in your input will be converted to literal ? characters.

[Amit Naidu]

This answer needs more votes. The sqlplus incompatibility with BOM is a cause of many headaches.

[watery]

This looks to be the solution still in 2018 with Out-File from PowerShell 6; but Notepad++ states the file has no encoding, any hint?

[DoubleOZ]

$MyFile variable does not have to be object that is created by a Get-Content. It can also be a plain string, i.e. $MyFile = "utf8 string of some kind..."

[pholpar]

Instead of New-Object System.Text.UTF8Encoding $False you can use simply New-Object System.Text.UTF8Encoding, since "This constructor creates an instance that does not provide a Unicode byte order mark", see docs.microsoft.com/en-us/dotnet/api/…

[PolarBear]

As @RosbergLinhares noted, WriteAllLines adds an extra new line at the end of a file. But to make WriteAllText work you have to use -Raw parameter for Get-Content, otherwise all text will be squashed into a single line. $fileContent = Get-Content -Raw "$fileFullName"; [System.IO.File]::WriteAllText($fileFullName, $fileContent)

[KCD]

Nice. FYI check version with $PSVersionTable.PSVersion

[mklement0]

Worth noting that in PowerShell [Core] v6+ -Encoding UTF8NoBOM is never required, because it is the default encoding.

[mklement0]

Nice - works great with strings (which may be all that is needed and certainly meets the requirements of the question). In case you need to take advantage of the formatting that Out-File, unlike Set-Content, provides, pipe to Out-String first; e.g., $MyFile = Get-ChildItem | Out-String

[mklement0]

To spell it out: This is a system-wide setting that makes Windows PowerShell default to BOM-less UTF-8 across all cmdlets, which may or may not be desired, not least because the feature is still in beta (as of this writing) and can break legacy console applications - see this answer for background information.

[duct_tape_coder]

In powershell 5.1 on Win 10: [IO.File]::WriteAllLines("c:\users\user\file.txt", $content) gives me Cannot find an overload for "WriteAllLines" and the argument count "2"

[Joel Coehoorn]

@AmitNaidu No, this is the wrong answer, because it won't work if the text has any non-ascii characters: any accents, umlauts, oriental/cryllic, etc.

[Erik Anderson]

@JoelCoehoorn This is a correct answer according to what the user asked. Since the user asked for a way to "force", they're not expecting any issues or don't care probably because the source doesn't use any non-ASCII characters. For those who do care about the preservation of those characters, this will not work.

[ygoe]

Warning: Definitely not. This deletes all non-ASCII characters and replaces them with question marks. Don't do this or you will lose data! (Tried with PS 5.1 on Windows 10)