WIndows API中的“L”和“LPCWSTR”

Question

I've found that

NetUserChangePassword(0, 0, L"ab", L"cd");

changes the user password from ab to cd. However,

NetUserChangePassword(0, 0, (LPCWSTR) "ab", (LPCWSTR) "cd");

doesn't work. The returned value indicates invalid password.

I need to pass const char* as last two parameters for this function call. How can I do that? For example,

NetUserChangePassword(0, 0, (LPCWSTR) vs[0].c_str(), (LPCWSTR) vs[1].c_str());

Where vs is std::vector<std::string> .

Answer 1

You have two problems.

The first is the practical problem - how to do this. You are confusing wide and narrow strings and casting from one to the other. A string with an L prefix is a wide string, where each character is two bytes (a wchar_t ). A string without the L is a single byte (a char ). You cannot cast from one to the other using the C-style cast (LPCWSTR) "ab" because you have an array of char s, and are casting it to a pointer to wide chars. It is simply changing the pointer type, not the underlying data.

To convert from a narrow string to a wide string, you would normally use MultiByteToWideChar . You don't mention what code page your narrow strings are in; you would probably pass in CP_ACP for the first parameter. However, since you are converting between a string and a wstring, you might be interested in other ways to convert ( one , two ). This will give you a wstring with your characters, not a string , and a wstring 's .c_str() method returns a pointer to wchar_t s.

The second is the following misunderstanding:

I need to pass const char* as last two parameters for this function call. How can I do that?

No you don't. You need to pass a wide string, which you got above. Your approach to this (casting the pointer) indicates you probably don't know about different string types and character encodings, and this is something every software developer should know. So, on the assumption you're interested, hopefully you'll find the following references handy:

• The Absolute Minimum Every Software Developer Absolutely, Positively Must Know About Unicode and Character Sets (No Excuses!)
• The Unicode FAQ (covers things like 'What is Unicode?')
• MSDN introduction to wide characters .

I'd recommend you investigate recompiling your application with UNICODE and using wide strings. Many APIs are defined in both narrow and wide versions, and normally this would mean you access the narrow version by default (you can access either the ANSI (narrow) or Wide versions of these APIs by directly calling the A or W version - they have A or W appended to their name, such as CreateWindowW - see the bottom of that page for the two names. You normally don't need to worry about this.) As far as I can tell, this API is always available as-is regardless of UNICODE , it's just it's only prototyped as wide.

Answer 2

Those are two totally different L 's. The first is a part of the C++ language syntax. Prefix a string literal with L and it becomes a wide string literal; instead of an array of char , you get an array of wchar_t .

The L in LPCWSTR doesn't describe the width of the characters, though. Instead, it describes the size of the pointer . Or, at least, it used to . The L abbreviation on type names is a relic of 16-bit Windows, when there were two kinds of pointers. There were near pointers, where the address was somewhere within the current 64 KB segment, and there were far , or long pointers, which could point beyond the current segment. The OS required callers to provide the latter to its APIs, so all the pointer-type names use LP . Nowadays, there's only one type of pointer; Microsoft keeps the same type names so that old code continues to compile.

The part of LPCWSTR that specifies wide characters is the W . But merely type-casting a char string literal to LPCWSTR is not sufficient to transform those characters into wide characters. Instead, what happens is the type-cast tells the compiler that what you wrote really is a pointer to a wide string, even though it really isn't. The compiler trusts you. Don't type-cast unless you really know better than the compiler what the real types are.

If you really need to pass a const char* , then you don't need to type-cast anything, and you don't need any L prefix. A plain old string literal is sufficient. (If you really want to cast to a Windows type, use LPCSTR — no W .) But it looks like what you really need to pass in is a const wchar_t* . As we learned above, you can get that with the L prefix on the string literal.

In a real program, you probably don't have a string literal. The user will provide a password, or you'll read a password from some other external source. Ideally, you would store that password in a std::wstring , which is like std::string but for wchar_t instead of char . The c_str() method of that type returns a const wchar_t* . If you don't have a wstring , a plain array of wchar_t might be sufficient.

But if you're storing the password in a std::string , then you'll need to convert it into wide characters some other way. To do a conversion, you need to know what code page the std::string characters use. The "current ANSI code page" is usually a safe bet; it's represented by the constant CP_ACP . You'll use that when calling MultiByteToWideString to have the OS convert from the password's code page into Unicode.

int required_size = MultiByteToWideChar(CP_ACP, 0, vs[0].c_str(), vs[0].size(), NULL, 0);
if (required_size == 0)
  ERROR;
// We'll be storing the Unicode password in this vector. Reserve at
// least enough space for all the characters plus a null character
// at the end.
std::vector<wchar_t> wv(required_size);
int result = MultiByteToWideChar(CP_ACP, 0, vs[0].c_str(), vs[0].size(), &wv[0], required_size);
if (result != required_size - 1)
  ERROR;

Now, when you need a wchar_t* , just use a pointer to the first element of that vector: &wv[0] . If you need it in a wstring , you can construct it from the vector in a few ways:

// The vector is null-terminated, so use "const wchar_t*" constructor
std::wstring ws1 = &wv[0];
// Use iterator constructor. The vector is null-terminated, so omit
// the final character from the iterator range.
std::wstring ws2(wv.begin(), wv.end() - 1);
// Use pointer/length constructor.
std::wstring ws3(&wv[0], wv.size() - 1);

Answer 3

C style casts as you've used here are a very blunt instrument. They assume you know exactly what you're doing.

You'll need to convert your ASCII or multi-byte strings into Unicode strings for the API. There might be a NetUserChangePasswordA function that takes the char * types you're trying to pass, try that first.

Answer 4

LPWSTR定义为wchar_t *（whcar_T是2字节char类型），其解释方式与正常的1字节字符不同。

Answer 5

LPCWSTR means you need to pass a wchar_t*.

if you change vs to std::vector<std::wstring> that will give you a wide char when you pass vs[0].c_str()

if you look at the example at http://msdn.microsoft.com/en-us/library/aa370650(v=vs.85).aspx you can see that they define UNICODE which is why they use the wchar_t.