Extracting all overlapping substrings between nested matching parentheses with a .NET regex

Question

I'm trying to parse mathematical expressions with nested brackets:

(1 * (2 - 3)) + 4

I want to get every expression in brackets, like this:

(1 * (2 - 3))
(2 - 3)

Using this expression: (.*?\))(?=($|[^(]+)) I'm getting this result:

(1 * (2 - 3)

)

And using this expression: $.*?$ I'm getting this result:

(1 * (2 - 3)

But nothing works correctly. How can I loop an expression inside?

score 2 · Accepted Answer · answered Apr 19 '21 at 21:07

You can use

(?=(\((?>[^()]+|(?<c>)\(|(?<-c>)\))*(?(c)(?!))\)))

See the regex demo. Details:

(?= - a positive lookahead:
- ($(?>[^()]+|(?<c>)\(|(?<-c>)$)*(?(c)(?!))\))) - Group 1:
  - $ - a ( char
  - (?>[^()]+|(?<c>)\(|(?<-c>)$)* - zero or more repetitions of any one or more chars other than ( and ), or a ( char (with a value pushed onto Group "c" stack), or a ) char (with a value popped from the Group "c" stack)
  - (?(c)(?!)) - if Group "c" stack is not empty, fail and backtrack
  - \) - a ) char.

See the C# demo:

var text = "(1 * (2 - 3)) + 4";
var pattern = @"(?=(\((?>[^()]+|(?<c>)\(|(?<-c>)\))*(?(c)(?!))\)))";
var results = Regex.Matches(text, pattern)
    .Cast<Match>()
    .Select(m => m.Groups[1].Value)
    .ToList();
Console.WriteLine(String.Join(", ", results));
// => (1 * (2 - 3)), (2 - 3)

interesting this works well when I run it in Linqpad, but in RegexBuddy it fails to match even though it compiles fine — gordy, Apr 19 '21 at 21:52
@gordy Do not use it in Regex*, use it in C#. This is a .NET compliant regex. — Wiktor Stribiżew, Apr 19 '21 at 21:56

gordy · Answer 2 · 2021-04-19T21:31:13.907

-1

The usual way would be to use a recursive regular expression but unfortunately this capability is not supported by C#'s Regex. Alternatively, you can manually parse the string (and there is C# code provided in this PAQ to do that).

edited Apr 19 '21 at 21:31

answered Apr 19 '21 at 21:05

gordy

9,360
1
31
43

Please don't link to existing answers like this. If this question is a duplicate of the other, then mark it as so. – Enigmativity Apr 19 '21 at 21:10
@Enigmativity is there a community guideline for this? it's not a duplicate question since they asked specifically how to do this via regex but the answer is probably relevant - I thought linking would be better than copying and pasting the answer here since the context is probably also useful – gordy Apr 19 '21 at 21:14
Unless you have something substantive in your answer (i.e. it can stand alone) then it should be a comment. Does "C#'s Regex doesn't support recursive expressions, I suggest parsing this manually" answer this question? – Enigmativity Apr 19 '21 at 21:17
https://meta.stackoverflow.com/questions/265552/when-to-flag-an-answer-as-not-an-answer – Enigmativity Apr 19 '21 at 21:19

Extracting all overlapping substrings between nested matching parentheses with a .NET regex

2 Answers2