Passing a string from C# to cpp with COM

Question

I have a C# COM server which is consumed by a cpp client.

One of the C# methods returns a string.

In cpp the returned string is represented in Unicode (UTF-16), at least according to the memory view.

Is this always the case with COM strings?
Is there a way to use UTF-8 instead?
I saw some code where strings were passed between cpp and c# as byte arrays. Is there any benefit in this?

This thread has turned into an unattractive downvoting fest with conflicting answers. I recommend you look up the definitions for BSTR and SysAllocString in the MSDN Library and draw your own conclusions. — Hans Passant
– Hans Passant, Commented Apr 26, 2010 at 16:04

Hans Passant · Accepted Answer · 2010-04-24 22:13:14Z

1

Yes. The standard COM string type is BSTR. It is a Unicode string encoded in UTF16, just like Windows' native string type.
No, a COM method isn't going to understand a UTF8 string, it will turn it into Chinese. UTF8 is a good encoding for a text file, not for programs manipulating strings in memory. UTF8 requires anywhere between 1 and 4 bytes to encode a Unicode codepoint. Very incompatible with basic string manipulations like getting the size or indexing a character.
C and C++ programs tend to use 8-bit encodings, compatible with the "char" type. That's an old practice, dating back from an era before Unicode was around. There's nothing attractive about it, there are many 8-bit encodings. The typical problem is that data entered as text can only be interpreted correctly if it is read by a program that uses the same 8-bit encoding. In other words, when the computers are less than 1000 miles apart. Less in Europe.

answered Apr 24, 2010 at 21:29

Hans Passant

946k151 gold badges1.8k silver badges2.6k bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Sounds to me like you've got it backward. He's calling into a C# COM component from C++.

@sblom: yes, your answer mystified me. COM looks the same way on both ends. Automation has always been Unicode enabled.

sblom · Accepted Answer · 2010-04-24 23:06:52Z

0

No.
Yes. Put the attribute [return: MarshalAs(UnmanagedType.LPStr)] before the method definition in C# if you'd like to return the string as an ANSI string instead of Unicode.
Yeah--the author may have done that to maintain very fine-grained control on the encoding of the contents of the string by side-stepping the default marshalling behavior.

answered Apr 24, 2010 at 16:59

sblom

27.5k4 gold badges74 silver badges96 bronze badges