Using Regex to replace part of the entire string/expression

Question

Regex are simple yet complex at times. Stuck to replace an expression having variables, assuming variable is of the following pattern:

\w+(\.\w+)*

I want to replace all the occurrences of my variable replacing dot (.) because i have to eventually tokenize the expression where tokenizer do not recognize variable having dots. So i thought to replace them with underscore before parsing. After tokenizing however i want to get the variable token with original value.

Expression:

(x1.y2.z3 + 9.99) + y2_z1 - x1.y2.z3

Three Variables:

x1.y2.z3

y2_z1

x1.y2.z3

Desired Output:

(x1_y2_z3 + 9.99) + y2_z1 - x1_y2_z3

• Question 1: How to use Regex replace in this case?

• Question 2: Is there any better way to address above mentioned problem because variable can have underscore so replacing dot with underscore is not a viable solution to get the original variable back in tokens?

Answer 1

This regex pattern seems to work: [a-zA-Z]+\\d+\\S+

To replace a dot found only in a match you use MatchEvaluator:

    private static char charToReplaceWith = '_';
    static void Main(string[] args)
    {
        string s = "(x1.y2.z3 + 9.99) + y2_z1 - x1.y2.z3";
        Console.WriteLine(Regex.Replace(s, @"[a-zA-Z]+\d+\S+", new MatchEvaluator(ReplaceDotWithCharInMatch)));
        Console.Read();
    }

    private static string ReplaceDotWithCharInMatch(Match m)
    {
        return m.Value.Replace('.', charToReplaceWith);
    }

Which gives this output: (x1_y2_z3 + 9.99) + y2_z1 - x1_y2_z3

I don't fully understand your second question and how to deal with tokenizing variables that already have underscores, but you should be able to choose a character to replace with (ie, if (string.Contains('_')) is true then you choose a different character to replace with, but probably have to maintain a dictionary that says "I replaced all dots with underscores, and all underscores with ^ , etc..).

Answer 2

Try this:

        string input = "(x1.y2.z3 + 9.99) + y2_z1 - x1.y2.z3";
        string output = Regex.Replace(input, "\\.(?<![a-z])", "_");

This will replace only periods which are followed by a letter (az).

Answer 3

Use Regex' negative lookahead by making a group that starts with (?!

A dot followed by something non-numeric would be as simple as this:

// matches any dot NOT followed by a character in the range 0-9
String output = Regex.Replace(input, "\\.(?![0-9])", "_");

This has the advantage that while the [0-9] is part of the expression, it is only checked as being behind the match, but is not actually part of the match.