Best way to determine if two path reference to same file in C/C++

Question

How would I compare 2 strings to determine if they refer to the same path in Win32 using C/C++?

While this will handle a lot of cases it misses some things:

_tcsicmp(szPath1, szPath2) == 0

For example:

• forward slashes / backslashes

• relative / absolute paths.

[Edit] Title changed to match an existing C# question.

Answer 1

A simple string comparison is not sufficient for comparing paths for equality. In windows it's quite possible for c:\foo\bar.txt and c:\temp\bar.txt to point to exactly the same file via symbolic and hard links in the file system.

Comparing paths properly essentially forces you to open both files and compare low level handle information. Any other method is going to have flaky results.

Check out this excellent post Lucian made on the subject. The code is in VB but it's pretty translatable to C/C++ as he PInvoke'd most of the methods.

http://blogs.msdn.com/vbteam/archive/2008/09/22/to-compare-two-filenames-lucian-wischik.aspx

Answer 2

use the GetFullPathName from kernel32.dll, this will give you the absolute path of the file. Then compare it against the other path that you have using a simple string compare

edit: code TCHAR buffer1[1000]; TCHAR buffer2[1000]; TCHAR buffer3[1000]; TCHAR buffer4[1000];

GetFullPathName(TEXT("C:\\Temp\\..\\autoexec.bat"),1000,buffer1,NULL);
GetFullPathName(TEXT("C:\\autoexec.bat"),1000,buffer2,NULL);
GetFullPathName(TEXT("\\autoexec.bat"),1000,buffer3,NULL);
GetFullPathName(TEXT("C:/autoexec.bat"),1000,buffer4,NULL);
_tprintf(TEXT("Path1: %s\n"), buffer1);
_tprintf(TEXT("Path2: %s\n"), buffer2);
_tprintf(TEXT("Path3: %s\n"), buffer3);
_tprintf(TEXT("Path4: %s\n"), buffer4);

the code above will print the same path for all three path representations.. you might want to do a case insensitive search after that

Answer 3

See this question: Best way to determine if two path reference to same file in C#

The question is about C#, but the answer is just the Win32 API call GetFileInformationByHandle.

Answer 4

Comparing the actual path strings will not produce accurate results if you refer to UNC or Canonical paths (i.e. anything other than a local path).

shlwapi.h has some Path Functions that may be of use to you in determing if your paths are the same.

It contains functions like PathIsRoot that could be used in a function of greater scope.

Answer 5

Open both files with CreateFile, call GetFileInformationByHandle for both, and compare dwVolumeSerialNumber, nFileIndexLow, nFileIndexHigh. If all three are equal they both point to the same file:

GetFileInformationByHandle function

BY_HANDLE_FILE_INFORMATION Structure

Answer 6

Based on answers about GetFileInformationByHandle(), here is the code.

Note: This will only work if the file already exists...

//Determine if 2 paths point ot the same file...
//Note: This only works if the file exists
static bool IsSameFile(LPCWSTR szPath1, LPCWSTR szPath2)
{
    //Validate the input
    _ASSERT(szPath1 != NULL);
    _ASSERT(szPath2 != NULL);

    //Get file handles
    HANDLE handle1 = ::CreateFileW(szPath1, 0, FILE_SHARE_DELETE | FILE_SHARE_READ | FILE_SHARE_WRITE, NULL, OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL, NULL); 
    HANDLE handle2 = ::CreateFileW(szPath2, 0, FILE_SHARE_DELETE | FILE_SHARE_READ | FILE_SHARE_WRITE, NULL, OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL, NULL); 

    bool bResult = false;

    //if we could open both paths...
    if (handle1 != NULL && handle2 != NULL)
    {
        BY_HANDLE_FILE_INFORMATION fileInfo1;
        BY_HANDLE_FILE_INFORMATION fileInfo2;
        if (::GetFileInformationByHandle(handle1, &fileInfo1) && ::GetFileInformationByHandle(handle2, &fileInfo2))
        {
            //the paths are the same if they refer to the same file (fileindex) on the same volume (volume serial number)
            bResult = fileInfo1.dwVolumeSerialNumber == fileInfo2.dwVolumeSerialNumber &&
                      fileInfo1.nFileIndexHigh == fileInfo2.nFileIndexHigh &&
                      fileInfo1.nFileIndexLow == fileInfo2.nFileIndexLow;
        }
    }

    //free the handles
    if (handle1 != NULL)
    {
        ::CloseHandle(handle1);
    }

    if (handle2 != NULL)
    {
        ::CloseHandle(handle2);
    }

    //return the result
    return bResult;
}

Answer 7

What you need to do is get the canonical path.

For each path you have ask the file system to convert to a canonical path or give you an identifier the uniquely identifies the file (such as the iNode).

Then compare the canonical path or the unique identifier.

Note: Do not try and figure out the conical path yourself the File System can do things with symbolic links etc that that are not easily tractable unless you are very familiar with the filesystem.

Answer 8

If the files exist and you can deal with the potential race condition and performance hit from opening the files, an imperfect solution that should work on any platform is to open one file for writing by itself, close it, and then open it for writing again after opening the other file for writing. Since write access should only be allowed to be exclusive, if you were able to open the first file for writing the first time but not the second time then chances are you blocked your own request when you tried to open both files.

(chances, of course, are also that some other part of the system has one of your files open)