如何在 PowerShell 中读取 UTF-16 编码的 StdIn

Question

I am trying to pass (a large number of [1]) strings from a native Windows host application (C++/WinApi) to a PowerShell script, which the host application launches using CreateProcess .

I use an anonymous pipe in STARTUPINFO::hStdInput as the IPC mechanism. The data being written to the pipe consists of lines of UTF-16LE strings [2]. What is printed by a naive PowerShell script

foreach ($line in $input) {
    write-host  $line
}

however, looks like the data from StdIn is being interpreted in an ANSI code page (each UTF 16 code unit from the input shows up as a pair of letters in the output).

How can I make PowerShell to recognize the data from StdIn as UTF-16?

I have already tried to

• prepend a UTF-16 BOM before the rest of the data on the pipe
• play with PowerShell's $InputEncoding , $OutputEncoding and .Net's [Console]::InputEncoding

to no avail. Yes, I could write a large text file first and then read it in PowerShell but I would rather not do this.

[1] This is why I would like to use a pipe and leverage the stream processing capabilities of PowerShell.
[2] Translating the data to a non-Unicode code page is not an option.

Answer 1

Just to finally clean up this old question: Setting up the .Net console input encoding (that is what Powershell builds upon) correctly is a pretty nontrivial issue. I finally worked around the problem, because I didn't want to burden the Powershell-script developers with the input encoding setup. So I ended up

• Encoding the data as Common Language Infrastructure objects in CLIXML
• Prefixing the stream with a "#< CLIXML\\r\\n" marker to declare the format to Powershell
• and finally (*cringe*) xml-escape every character in the xml document outside the ASCII range to completely avoid any input encoding ambiguities

The final point turned out to be necessary, because the handling of clixml comes only after the text goes through the fragile console input decoding process.