Regex to match many times

Simon H Published at Dev

Simon H

I'm trying to match a type definition

def euro : t1 -> t2 -> t3 (and this pattern my repeat further in other examples)

I came up with this regex

^def ([^\s]*)\s:\s([^\s]*)(\s->\s[^\s]*)*

But while it matches euro and t1 it

then matches -> t2 rather than t2
fails to match anything with t3

I can't see what I am doing wrong, and my goal is to capture

euro t1 t2 t3

as four separate items, and what I currently get is

0: "def euro : t1 -> t2 -> t3"
1: "euro"
2: "t1"
3: " -> t3"

Wiktor Stribiżew

You can't use a repeated capturing group in JS regex, all but the last values will be "dropped", re-written upon each subsequent iteration.

When creating a regular expression that needs a capturing group to grab part of the text matched, a common mistake is to repeat the capturing group instead of capturing a repeated group. The difference is that the repeated capturing group will capture only the last iteration, while a group capturing another group that's repeated will capture all iterations.

The way out can be capturing the whole substring and then split it. Here is an example:

var s = "def euro : t1 -> t2 -> t3";
var rx = /^def (\S*)\s:\s(\S*)((?:\s->\s\S*)*)/;
var res = [];
var m = s.match(rx);
if (m) {
  res = [m[1], m[2]];
  for (var s of m[3].split(" -> ").filter(Boolean)) {
     res.push(s);
  }
}
console.log(res);

Pattern details

^ - start of string
def - a literal substring
(\S*) - Capturing group 1: 0+ non-whitespace chars
\s:\s - a : enclosed with single whitespaces
(\S*) - Capturing group 2: 0+ non-whitespace chars
- ((?:\s->\s\S*)*) - Capturing group 3: 0+ repetitions of the following pattern sequences:
- \s->\s - whitespace, ->, whitespace
- \S* - 0+ non-whitespace chars