Question

I was trying to remove all comments and empty lines in a file with the help of a macro. Now I came up with this solution which deletes the comments(there is some bug described below) but is not able to delete the blank lines in between –

Sub CleanCode()
    Dim regexComment As String = "(REM [\d\D]*?[\r\n])|(?<SL>\'[\d\D]*?[\r\n])"
    Dim regexBlank As String = "^[\s|\t]*$\n"
    Dim replace As String = ""

    Dim selection As EnvDTE.TextSelection = DTE.ActiveDocument.Selection
    Dim editPoint As EnvDTE.EditPoint

    selection.StartOfDocument()
    selection.EndOfDocument(True)

    DTE.UndoContext.Open("Custom regex replace")
    Try
        Dim content As String = selection.Text
        Dim resultComment As String = System.Text.RegularExpressions.Regex.Replace(content, regexComment, replace)
        Dim resultBlank As String = System.Text.RegularExpressions.Regex.Replace(resultComment, regexBlank, replace)
        selection.Delete()
        selection.Collapse()
        Dim ed As EditPoint = selection.TopPoint.CreateEditPoint()
        ed.Insert(resultBlank)
    Catch ex As Exception
        DTE.StatusBar.Text = "Regex Find/Replace could not complete"
    Finally
        DTE.UndoContext.Close()
        DTE.StatusBar.Text = "Regex Find/Replace complete"
    End Try
End Sub

So, here is what it should looks like before and after running the macro.

BEFORE

Public Class Class1
    Public Sub New()
        ''asdasdas
        Dim a As String = "" ''asdasd
        ''' asd ad asd
    End Sub


    Public Sub New(ByVal strg As String)

        Dim a As String = ""

    End Sub


End Class

AFTER

Public Class Class1
    Public Sub New()
        Dim a As String = ""
    End Sub
    Public Sub New(ByVal strg As String)
        Dim a As String = ""
    End Sub
End Class

There are mainly two main problems with the macro

• It cannot delete the blank lines in between.
• If there is a piece of code which goes like this

Dim a as String = "Name='Soham'"

Then After running the macro it becomes

Dim a as String = "Name='"

Answer 1

To get rid of a line that contains whitespace or nothing, you can use this regex:

(?m)^[ \t]*[\r\n]+

Your regex, ^[\s|\t]*$\n would work if you specified Multiline mode ((?m)), but it’s still incorrect. For one thing, the | matches a literal |; there’s no need to specify “or” in a character class. For another, \s matches any whitespace character, including TAB (\t), carriage-return (\r), and linefeed (\n), making it needlessly redundant and inefficient. For example, at the first blank line (after the end of the first Sub), the ^[\s|\t]* will initially try to match everything before the word Public, then it will back off to the end of the previous line, where the $\n can match.

But a blank line, in addition to being empty or containing only horizontal whitespace (spaces or TABs), may also contain a comment. I choose to treat these “comment-only” lines as blank lines because it’s relatively easy to do, and it simplifies the task of matching comments in non-blank lines, which is much harder. Here’s my regex:

^[ \t]*(?:(?:REM|')[^\r\n]*)?[\r\n]+

After consuming any leading horizontal whitespace, if I see a REM or ' signifying a comment, I consume that and everything after it until the next line separator. Notice that the only thing that’s required to be present is the line separator itself. Also notice the absence of the end anchor, $. It’s never necessary to use that when you’re explicitly matching the line separators, and in this case it would break the regex. In Multiline mode, $ matches only before a linefeed (\n), not before a carriage-return (\r). (This behavior of the .NET flavor is incorrect and rather surprising, given Microsoft’s longstanding preference for \r\n as a line separator.)

Matching the remaining comments is a fundamentally different task. As you’ve discovered, simply searching for REM or ' is no good because you might find it in a string literal, where it does not signify the start of a comment. What you have to do is start from the beginning of the line, consuming and capturing anything that’s not the beginning of a comment or a string literal. If you find a double-quote, go ahead and consume the string literal. If you find a REM or ', stop capturing and go ahead and consume the rest of the line. Then you replace the whole line with just the captured portion–i.e., everything before the comment. Here’s the regex:

(?mn)^(?<line>[^\r\n"R']*(("[^"]*"|(?!REM)R)[^\r\n"R']*)*)(REM|')[^\r\n]*

Or, more readably:

(?mn)             # Multiline and ExplicitCapture modes
^                 # beginning of line
(?<line>          # capture in group "line"
  [^\r\n"R']*     # any number of "safe" characters
  (
    (
      "[^"]*"     # a string literal
      |
      (?!REM)R    # 'R' if it's not the beginning of 'REM'
    )
    [^\r\n"R']*   # more "safe" characters
  )*
)                 # stop capturing
(?:REM|')         # a comment sigil
[^\r\n]*          # consume the rest of the line

The replacement string would be "${line}". Some other notes:

• Notice that this regex does not end with [\r\n]+ to consume the line separator, like the “blank lines” regex does.
• It doesn’t end with $ either, for the same reason as before. The [^\r\n]* will greedily consume everything before the line separator, so the anchor isn’t needed.
• The only thing that’s required to be present is the REM or '; we don’t bother matching any line that doesn’t contain a comment.
• ExplicitCapture mode means I can use (...) instead of (?:...) for all the groups I don’t want to capture, but the named group, (?<line>...), still works.
• Gnarly as it is, this regex would be a lot worse if VB supported multiline comments, or if its string literals supported backslash escapes.

I don’t do VB, but here’s a demo in C#.