[coliveira]

I believe some people used that monster version because VB had simple libraries that didn't allow many options. Anything more complex required you to go down the Win32 rabbit hole. And that was the main weakness of VB, it was such a simple language that you didn't have many options to do something more complex than going straight to Win32.

[AshamedCaptain]

It is a very different thing to claim "the language doesn't allow simple things such as checking for a file's existence" (which is false, the language allows many ways to do so) than to claim "the language doesn't have many options, but allows you to interop with the entire win32 API if you need to". The first is a valid if only false complain; the second reads to me not as a complain, but as a feature.

[jasode]

>to claim "the language doesn't allow simple things such as checking for a file's existence" (which is false,

That's not what I wrote. I said there's no _built-in_ way to check existence of the file. A reasonable interpretation of "built-in" is for an obvious named function such as "File.exists()" _without_ cobbling together extra workarounds and hacks -- whether those workarounds include re-purposing other VB built-in functions with On Error ... or using Win32 interop.

> in VB, [...] would at worse be comprised of a exception handler (on error ...)

That's an example of what I meant of not being built-in to VB. I should have been more clear. We were using different meanings of "builtin".

In any case, the following "pure" VB code to check for file existence is not found on Google or Bing search but I still have it from my 1995 archives and copied it here. It has many lines of code (instead of a simple "one-liner") because it tries to cover all the edge cases. And yet with all that defensive code, there's still a hidden timebomb of a bug in it. Can anyone spot it? The code is unmodified from Visual Basic Programmer's Journal. In contrast, the alternative using Win32 interop with OF_EXISTS doesn't have the bug.

  Dim strTestFilename As String
  strFilename = Trim(strFilename)

  ' To avoid problem with known bug, don't pass root directory to
  ' the Name function
    Select Case Right$(strFilename, 1)
        Case "\": strExistsFilename = "Root": Exit Function
        Case "\.": strExistsFilename = "Root": Exit Function
        Case "\..": strExistsFilename = "Root": Exit Function
    End Select

    On Local Error Resume Next      ' Enable error trapping for Name function
    Name strFilename As strFilename ' A trick to see if filename/path is good

  ' Now check the resulting err code of the Name function
    Select Case Err
        Case 5  ' Illegal function call
        Case 53 ' File not found
        Case 58         ' File exists (maybe)
            strTestFilename = Dir$(strFilename)
            If Len(strTestFilename) Then
                strExistsFilename = ""      ' It's a file
            Exit Function
            Else
                strExistsFilename = "Dir"   ' It's a directory
                Exit Function
            End If
        Case 64 ' Bad filename
        Case 68 ' Device unavailable
        Case 71 ' Disk not ready
        Case 75 ' Access denied
        Case 76 ' Path not found
    End Select

My point was that C# builtin File.Exists() is a lot simpler than that.

[AshamedCaptain]

So, what is so wrong about simply calling FileLen() to check if the file exists, or if you also want to distinguish between Files/Directories using GetAttr to check the directory bit ? Why do you have to go all the way to this monster of code which even hardcodes lists of errors (why do you care which error it is?) ?

Your code sample renames the file, then globs anyway (when dir$ by itself can also be used to check existence). It looks too convoluted, and I cannot imagine why it needs to be so.

What is exactly the corner case that you want to avoid -- and if such case really exists, why are you so sure the File.Exists() implementation would not fall into the same pit trap, considering it also simply uses the metadata?

Disclaimer: I'm only familiar with VBA, not VB itself. But these file metadata functions definitely already exist in 95's VB5. (EDIT: Actually 97's. Maybe the problem is these functions did not exist in earlier VBs? But I would find it hard to imagine you could not check a file's size).

The sample code smells even in the trivial-case checks. E.g. Right$(_, 1) will only return a 1-char string at most, yet it tries to compare it with 2, 3 len strings. I didn't check any further than that.

[AshamedCaptain]

To elaborate.. just what is wrong with simply doing the following?

    Function FileExists(Filename) As Boolean
      On Error Resume Next
      FileExists = (GetAttr(Filename) And vbDirectory) = 0
    End Function

It's not 2 lines, but it's 4, including function header. All functions used are available at least since 16-bit VB3. Is this what you call "a workaround and a hack?". It's practically the same thing .NET is doing for File.Exists() (or at least the current one -- dunno what .NET 1 was doing), so I cannot think of any gotchas that would affect this but not File.Exists.

Another way, use Dir by itself. This is even mentioned in the VB docs...

    FileExists = Dir(Filename) <> ""

3 lines. It also ignores directories. Sure, this has problems with wildcards, but so does your example, and that's another oneliner to fix, using Replace$ (as wildcards are not legal characters in filenames in win32, either way). And if Dir("\") returns true for some reason, I guess that's a runtime bug, which is pretty valid criticism, even if a bit of a corner case (who really wants to check if "\" exists?). I would prefer the first version anyway.

I still stand by my original point that this is a very poor example, since there are a million ways check for a file's existence, many using only builtin functions of the language. You can certainly find nigh-overcomplicated ways to do so, apparently even on books, but ... why? Not a fan of cargo-culting like this; it tends to create Cherteston's fences.

Many languages also lack a direct "File_Exists" function (e.g. Lua comes to mind) since it's about the most trivial thing to implement (and a magnet for TOCTOU issues). Or even if they do come with such function, it does not distinguish directories from files (e.g. Tcl). I hardly think the lack of a "file_exists" function discriminates anything in language design, modern or otherwise.

[blacksqr]

> does not distinguish directories from files (e.g. Tcl)

Tcl has the built-in functions [file exists], [file isdirectory] and [file isfile] to serve your existence-checking needs.

[AshamedCaptain]

I know. So does VB. This is precisely my point.

GP complains that there is no direct function for "file exist". And technically there is not on Tcl either because file exists cmd will return true for directories. You can trivially build your proc that specifically checks for files (it will likely use isfile) but that is exactly the point of my comment. I have shown how do that for VB and it is not far from a one liner.

Presence/absence of a file exists function is a very poor example. If the language didn't allow you to check e.g. if a file was a directory or not, then GP might had have a point. But this is not the case.