The $replace function can replace matches of a regular expression within a string with another string. The replacement string uses the syntax $N to refer to matching groups of the regex, where N is the position of the matching group.

For example, this expression drops the suffix from ordinal numbers and prefixes the resulting cardinal numbers with a hash sign.

$replace("1st, 2nd, 3rd and 4th", /([0-9]+)(?:st|nd|rd|th)\b/, "#$1")

returns:

"#1, #2, #3 and #4"

The "$1" in the replacement string represents the first submatch of the regex (a run of one or more digits in this case).

There's room for ambiguity in this syntax. An expression like "$10" could either refer to the tenth submatch of a regex, or the first submatch followed by a zero. In fact, the same expression behaves differently depending on the number of submatches present. With ten or more, "$10" evaluates to the tenth submatch. With fewer than ten, "$10" evaluates to the first submatch and a zero. Note that it's not possible to achieve the latter configuration with ten or more submatches.

Go's regular expression library has some similar functionality but with a modification that avoids this issue. In the Go version, "$10" always refers to the tenth submatch, regardless of how many submatches actually exist. To get the first (or any) submatch followed by a digit, you wrap it in braces, e.g.

${1}0

With this approach the intent of the expression is clear and there are no unrepresentable configurations. Perhaps something similar could be implemented in JSONata.