How do I use regex to split only on commas not in angle brackets?

Question

I have the string DobuleGeneric<DoubleGeneric<int,string>,string>

I am trying to grab the 2 type arguments: DoubleGeneric<int,string> and string

Initially I was using a split on ','. This worked, but only if the generic args are not themeselves generic.

My Code:

string fullName = "DobuleGeneric<DoubleGeneric<int,string>,string>";
Regex regex = new Regex( @"([a-zA-Z\._]+)\<(.+)\>$" );
Match m = regex.Match( fullName );
string frontName = m.Groups[1].Value;
string[] innerTypes = m.Groups[2].Value.Split( ',' );

foreach( string strInnerType in innerTypes ) {
        Console.WriteLine( strInnerType );
}

Question: How do I do a regex split on commas that are not encapsulated in angle brackets?

Answer 1

Both commas are between angle brackets! Regex does a bad job when parsing a complex nested syntax. The question should be, how to find a comma, which is between angle brackets that are themselves not between angle brackets. I don't think that this can be done with regex.

If possible, try to work with Reflection. You might also use CS-Script to compile your code snippet and then use Reflection to retrieve the information you need.

Answer 2

To split the example you have given you can use the following. However, this is not generic; it could be made generic based on the other strings that you expect. Depending on the variation of the strings you have, this method could get complex; but I would suggest that the use of Roslyn here is overkill...

string fullName = "DobuleGeneric<DoubleGeneric<int,string>,string>"; 
Regex Reg = 
    new Regex(@"(?i)<\s*\p{L}+\s*<\s*\p{L}+\s*,\s*\p{L}+\s*>\s*,\s*\p{L}+\s*>");
Match m = Reg.Match(fullName);
string str = m.ToString().Trim(new char[] { '<', '>' });
Regex rr = new Regex(@"(?i),(?!.*>\s*)");
string[] strArr = rr.Split(str);

I hope this helps.

Answer 3

The answers are correct, using Regex is the wrong approach.

I ended up doing a linear pass, replacing items encapsulated in brackets with ~ s, and then doing a split.

static void Main( string[] args ) {

    string fullName = "Outer<blah<int,string>,int,blah<int,int>>";          

    Regex regex = new Regex( @"([a-zA-Z\._]+)\<(.+)\>$" );
    Match m = regex.Match( fullName );
    string frontName = m.Groups[1].Value;
    string inner = m.Groups[2].Value;

    var genArgs = ParseInnerGenericArgs( inner );

    foreach( string s in genArgs ) {
        Console.WriteLine(s);
    }
    Console.ReadKey();
}

private static IEnumerable<string> ParseInnerGenericArgs( string inner ) {
    List<string> pieces = new List<string>();
    int angleCount = 0;
    StringBuilder sb = new StringBuilder();
    for( int i = 0; i < inner.Length; i++ ) {
        string currChar = inner[i].ToString();
        if( currChar == ">" ) {
            angleCount--;
        }
        if( currChar == "<" ) {
            angleCount++;
        }
        if( currChar == ","  &&  angleCount > 0 ) {

            sb.Append( "~" );

        } else {
            sb.Append( currChar );
        }

    }
    foreach( string item in sb.ToString().Split( ',' ) ) {
        pieces.Add(item.Replace('~',','));
    }
    return pieces;
}

Answer 4

Here is the regex I will use:

\<(([\w\.]+)(\<.+\>)?)\,(([\w\.]+)(\<.+\>)?)$

([\\w\\.]+) matches "DoubleGeneric". (\\<.+\\>)? matches the possible generic args like DoubleGeneric <OtherGeneric<int, ...>>

The key point is that no matter how many nested generic args you have you will have only one ">," in the whole expression.

You can use m.Gruops[1] and m.Groups[4] to get the first and second Type.