Marshal::StringToHGlobalAnsi and the corresponding overhead

Thread view:

Enable EMail Alerts

Start New Thread

Thread rating:

DaTurk - 27 Mar 2007 16:22 GMT

Hi,

I have a c# GUI that needs to untimately send its data down to
unmanaged c++. I've decided to do this by having a ref struct in the
CLI layer and have the c# populate this, and then pass it back to the
CLI layer. The CLI layer will then populate the unmanged struct where
the data needs to end up.

Now, my question, I just realized that I'm going to have alot of
strings. I need to have these things end up as char*, but they start
as String^. I'm just thinking about overhead, I was thinking about
Marshalling the strings via Marshal::StringToHGlobalAnsi, but I'm a
little worried about the overhead. Any ideas?

Thanks in advance.

Reply to this Message

William DePalo [MVP VC++] - 27 Mar 2007 17:57 GMT

> Now, my question, I just realized that I'm going to have alot of
> strings. I need to have these things end up as char*, but they start
> as String^. I'm just thinking about overhead, I was thinking about
> Marshalling the strings via Marshal::StringToHGlobalAnsi, but I'm a
> little worried about the overhead. Any ideas?

If you don't plan to modify the strings you can use PtrToStringChars():

http://blogs.msdn.com/slippman/archive/2004/06/02/147090.aspx

IMO, it's only by profiling that you'll be able to determine if your worries
are warranted or if one method is materially better than another.

Regards,
Will

Reply to this Message

Ben Voigt - 28 Mar 2007 21:47 GMT

> Hi,
>
[quoted text clipped - 9 lines]
> Marshalling the strings via Marshal::StringToHGlobalAnsi, but I'm a
> little worried about the overhead. Any ideas?

Most of the overhead there is in Unicode->ANSI conversion. If the client
can use byte[] (cli::array<System::Byte>^) instead, it will be faster.
Depends on whether it acts like character strings or raw byte data.

> Thanks in advance.

Reply to this Message

DaTurk - 30 Mar 2007 20:46 GMT

> > Hi,
>
[quoted text clipped - 17 lines]
>
> - Show quoted text -

That's interesting, I didn't think about that. There will still have
to be someconversion at some point as the information I'm going to be
using is going to be coming from a WinForm, so it will be of the type
String^ at the presentation layer. SO I huess the question then
becomes, would it be more costly to convert a String^ -> a byte[] at
the presentation layer, or to convert a String^ -> a char* down lower
in the program?

Reply to this Message

Ben Voigt - 30 Mar 2007 20:56 GMT

> That's interesting, I didn't think about that. There will still have
> to be someconversion at some point as the information I'm going to be
[quoted text clipped - 3 lines]
> the presentation layer, or to convert a String^ -> a char* down lower
> in the program?

I think the Unicode->ANSI conversion may be implemented differently in .NET
vs native, so that could be a consideration too. Otherwise, I don't see why
either path would be any faster. I would probably hand a String^ to C++ to
keep the presentation layer free from implementation details.

My basic suggestion was that not all char* start life as strings, some are
more "magic numbers". E.g. in SMTP, the "MAIL TO:" command isn't a string
subject to localization, it has to be those exact bytes. Similarly for HTML
tags and the filetype codes used by a lot of programs (GIF does this).
Yours is user-entered though, so you do need the conversion.

Reply to this Message

Rate this thread:

Marshal::StringToHGlobalAnsi and the corresponding overhead

Free Magazines