CODEWITHSUNDEEP
c++visual-studiocharacter-encoding
user236520
user236520

Reputation:

Why doesn't this character conversion work?

Visual Studio 2008

Project compiled as multibyte character set

LPWSTR lpName[1] = {(WCHAR*)_T("Setup")};

After this conversion, lpName[0] contains garbage (at least when previewed in VS)

LPWSTR is typedef'd as follows:
typedef __nullterminated WCHAR *NWPSTR, *LPWSTR, *PWSTR;

Upvotes: 0

Views: 130

Answers (2)

n. m. could be an AI
n. m. could be an AI

Reputation: 120079

It's an expanded version of my comment above.

The code shown casts a pointer of type A to a pointer of type B. This is a low-vevel, machine-dependent operation. It almost never works as a conversion of an object of type A to an object of type B, especially if one type is a regular character type and the other is wide characters.

Imagine that you take a French book, and read it aloud as if it was written in English. 

FRENCH* book;
readaloud ((ENGLISH*) book);

You will mostly hear gibberish. The letters used in the two languages are the same (or similar, at any rate), but the rules of the two languages are are totally different. The representation is the same for both languages, but the meaning is not.

This is very similar to what we have here. Whatever type you have, bits and bytes are the same, but the rules are totally different. You take bits laid out according to regular character rules, and try to interpret them according to wide character rules. It doesn't work. The representation is the same in both cases, but the meaning is not.

To convert one character flavor to another, you in general need a lookup table or some other means to convert each character from one type to the other — change representation, but keep the meaning. Likewise, to convert a French book into an English book, you need to use a big lookup table a.k.a. dictionary... well, the analogy breaks here, because there's no formal set of conversion rules, you need to be creative! But you get the idea.

The rules of C++ actually prohibit such casts. You can only cast an object type poiner to void*, and only use the result to cast it back to the original object type. Everything else is a no-no (unless you are willing to venture in the realm of undefined behavior).

So what should you do?

  1. Pick one character variant and stick to it.
  2. If you must convert between flavors, do so with a library function.
  3. Try to avoid pointer casts, they almost always signal trouble.

Upvotes: 1

Ken Wayne VanderLinde
Ken Wayne VanderLinde

Reputation: 19347

I think what you're looking for is

LPTSTR lpName[1] = {_T("Setup")};

The various typedefs with a T in them (e.g. TSTR, LPTSTR) are dependant on whether you use unicode or multi-byte or whatever else. By using these, you should be able to write code that work in whatever encoding you are using (i.e., tomorrow you could switch to ascii, and a large portion of your code should still work).

Edit

If you are in situation where you really must convert between encodings, then there are various conversion functions available, such as wcstombs (or microsoft's documentation) and mbstowcs. These are defined in <cstdlib>

Upvotes: 0

Related Questions