In C#, as in many other programming language, the regex engine supports capturing groups, that are submatches, parts of substrings that match a whole regex pattern, defined in a regex pattern with the help of parentheses (e.g. 1([0-9])3
will match 123
and save the value of 2
into a capture group 1 buffer). Captured texts are accessed via Match.Groups[n].Value
where n is the index of the capture group inside the pattern.
Capturing is much more effecient that lookarounds. Whenever there is no need for complex conditions, capturing groups are much better alternatives.
See my regex speed test performed at regexhero.net:

Now, how can we get the substring inside curly braces?
- if there is no other curly braces inside, with a negated character class:
{([^{}]*)
- if there can be nested curly brackets:
{((?>[^{}]+|{(?<c>)|}(?<-c>))*(?(c)(?!)))
In both cases, we match an opening {
, and then match (1) any character other than {
or }
, or (2) any characters up to the first paired }
.
Here is sample code:
var matches = Regex.Matches("Test {Token1} {Token 2}", @"{([^{}]*)");
var results = matches.Cast<Match>().Select(m => m.Groups[1].Value).Distinct().ToList();
Console.WriteLine(String.Join(", ", results));
matches = Regex.Matches("Test {Token1} {Token {2}}", @"{((?>[^{}]+|{(?<c>)|}(?<-c>))*(?(c)(?!)))");
results = matches.Cast<Match>().Select(m => m.Groups[1].Value).Distinct().ToList();
Console.WriteLine(String.Join(", ", results));
Result: Token1, Token 2
, Token1, Token {2}
.
Note that RegexOptions.IgnoreCase
is redundant when you have no literal letters that can have different case in the pattern.