Regular expression for splitting by char which might be escaped

Question

I need to split a string similar to a path, which is delimited by dots. The tricky part is that the each subentry may also contain dots, which are escaped by another dot. Each entry may otherwise contain basically anything (including special characters such as space or :;/\|(), etc..)

Two examples:

"Root.Subpath.last/entry:with special;chars" -> [0] Root [1] SubPath [2] last/entry:with special;chars

"Root.Subpath..with..dots.Username" -> [0] Root [1] SubPath.with.dots [2] Username

Currently I am not using a regular expression for this, instead I am replacing any ".." with something else before running a split, and adding them back after the split. This works fine and everything, but its not super clean. However, mostly I am curious (or maybe annoyed about?) how to create a Regex for Regex.Split that does the same thing, as this was my first idea of approach. I provide my current solution to show what output I expect.

Split(path.Replace("..", REP_STR), ".") _
  .Select(Function(s as string) s.Replace(REP_STR, ".")).ToArray

I am using VB.NET.

Answer 1

I don't know if VB.NET supports non-capturing groups, but in Java I would use this regular expression to split your string:

(?<=[^\.])\.(?=[^\.]|$)

Answer 2

I can't test VB.NET at home so this code is not test but I think it should work.

Dim Temp = ""
Dim aTempMaker as New RegEx("([^\.])\.([^\.])")
Dim aDeEscaper as New RegEx("\.\.")
Dim aSpliter   as New RegEx(Temp)

aStrs    = aSpliter.Split(aTempMaker.Replace(Text, "$1"+Temp+"$2"))
aResults = New String(aStrs.Length)

i = 0
For Each aStr In aStrs
    aResults(i) = aDeEscaper.Replace(aStr, ".")
Next

Answer 3

I would not use a regular expression for matching items and returning them. Even if you make the perfect regular expression, you'll still need to replace the double dots with single dots afterwards.

You could use a regex such as (?<!\.)\.(?!\.) for splitting, but I would probably just stick with your current method as it is more efficient. Alternatively, write your own splitting function that will do the "de-dotting" at the same time.

Here's a custom function that might look long, but is probably still more efficient than replacing, splitting then replacing again (and more efficient than a regex too):

And yes, it's C#, because I don't know VB.NET, but for the most part the two languages are interchangeable.

public static string[] SplitPath(string path)
{
    List<string> pieces = new List<string>();

    int index = -1, last = 0;
    // Keep looping as long as there are dots.
    while ((index = path.IndexOf('.', index + 1)) >= 0)
    {
        // Don't do more checking on last character.
        if (index == path.Length - 1) break;

        // If next character is also a dot, skip.
        if (path[index + 1] == '.')
        {
            index++;
            continue;
        }

        // Add current piece.
        pieces.Add(path.Substring(last, index - last).Replace("..", "."));

        // Store start of next piece.
        last = index + 1;
    }

    // Add final piece, unless it is empty.
    if (last < path.Length - 1) pieces.Add(path.Substring(last).Replace("..", "."));

    return pieces.ToArray();
}

Answer 4

The regex

(?<!\.)\.(?!\.)

will match a dot only if it is neither preceded nor followed by another dot.

Answer 5

Here’s another regular expression that is a little more efficient since the look-behind assertion is only tested if a dot has already been found:

\.(?<!\.\.)(?!\.)

Answer 6

This will match the first dot in a odd-lengthed sequence of dots.

\.(?<!\.\.)(?=(\.\.)*[^.])

An example of splitting at this pattern:

// input
'Foo.Bar..Baz...Bop....Quux'

// becomes
0 => 'Foo'
1 => 'Bar..Baz'
2 => '.Bop....Quux'

Slightly confusing, but it works. It should be possible to split at the last dot in the sequence as well using a variable-width lookbehind, however they are not widely supported in regular expression libraries.