FIX: CString::TrimLeft Fails in a UNICODE Application (129770)

SYMPTOMS

In a UNICODE application, CString::TrimLeft creates a string that is the same length as the original with the first half of the trimmed string replacing the first few characters of the original string.

The following are examples that show the result of using TrimLeft in a UNICODE application. The dash (-) character represents white space.

CString::TrimLeft()

Before      After
-------------------
--A         A-A
--AB        ABAB
--ABC       ABABC
--ABCD      ABCBCD
--ABCDE     ABCBCDE

CAUSE

The CString::TrimLeft function trims all leading white space characters from a string when using ASCII, which is a single-byte character set. However, when using UNICODE, which is a double-byte character set, the CString::TrimLeft function copies only half of the non-white space characters over the white spaces. In addition, because the null character is not in the first half of the non-white space characters, you end up with a new string that is a combination of the copied characters and the original string. This is due to the following line of code in STREX.CPP, which you'll find in MSVC20\MFC\SRC:

   memmove(m_pchData, lpsz, nDataLength+1);

This line of code moves (nDataLength+1) bytes instead of characters.

RESOLUTION

The easiest solution is to create a global TrimLeft function to, in this case, take a CString as a parameter. An implementation for this global function is listed in the "Sample Code to Workaround Problem" section below.

While it is possible to override the CString class and modify the functionality of TrimLeft so that characters are moved instead of bytes, Microsoft doesn't recommend it because you would also need to override the constructors, assignment operators, and destructors for your CString derived class. To override CString::TrimLeft, copy the functionality from CString::TrimLeft in STREX.CPP line 390 to your function, and then change the problem line of code to this:

   memmove(m_pchData, lpsz, (nDataLength+1)*sizeof(TCHAR));

Both solutions will work with both single-byte and double-byte character sets.

MORE INFORMATION

Sample Code to Workaround Problem

/* Compile options needed:
This change does not require special options.
*/ 

void TrimLeft(CString& str){

   // Get the cstring length and buffer pointer:

      int    str_length = str.GetLength();
      LPTSTR pstr_Data =  str.GetBuffer(str_length);

  // Find the first non white-space character:

     LPCTSTR pstr_non_space = pstr_Data;
     while (_istspace(*pstr_non_space))
             pstr_non_space = _tcsinc(pstr_non_space);

  // Find the new length in characters:

     int nDataLength = str_length - (pstr_non_space - pstr_Data);

  // Replace this line:

     // memmove(m_pchData, lpsz, nDataLength + 1);
     // WITH THE FOLLOWING TO MOVE CHARACTERS INSTEAD OF BYTES:

  memmove(pstr_Data,pstr_non_space, (nDataLength+1)*sizeof(TCHAR));

  // Release the buffer pointer, and adjust the cstring length:

     str.ReleaseBuffer(nDataLength);
}