clarify EReg.matched spec regarding capture groups

[nadako]

So currently the doc for EReg.matched says this:

The index n corresponds to the n-th set of parentheses in the pattern of this EReg. If no such sub-group exists, the result is unspecified.

But it's not immediately clear what "no such sub-group exists" means exactly. I assume it means that the group was not defined in the regex. But it's a bit confusing in case the group was defined but was not matched.

I.e. is it safe to rely on the following to consistently trace an empty string?

var r = ~/(a)?b/;
if (r.match("b")) {
    trace(r.matched(1));
}

[nadako]

is it safe to rely on the following to consistently trace an empty string?

Looks like it isn't because at least on JS that returns undefined. Maybe null could be a spec then? :)

[nadako]

And looks like we indeed test it for null: https://github.com/HaxeFoundation/haxe/blob/3352cab1d38992ef317d110151783c50fd138b41/tests/unit/src/unit/TestEReg.hx#L23

That makes it a documentation and type-hinting (Null<String> instead of String) issue.

[lublak]

I had the same question. Just when you want to test if a group applies: Example: https://regex101.com/r/RvD4XP/1. Would it be practical to test for zero. If most targets return zero anyway, a documentation about it would be really great.