You don't have a loop around the /g match, so it only matches once and uses the first matching alternative, "a".

The following loops over the string and outputs %+ whenever it matches something:

> perl -wE"while ('abcde'=~m[(?<a>a)|(?<b>b)|(?<c>c)|(?<d>d)]g){say'--
+';say for keys%+}"
--
a
--
b
--
c
--
d
[download]

I don't think that %+ will ever be filled with keys from more than one part of an alternation, so you'll never find more than one key in your example. To get all keys you'll likely need to introduce an explicit ordering or enumerate the alternatives.

So, each iteration of the implicit /g loop will overwrite %+. Ie. they don't accumulate?

---

Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error.

"Science is about questioning the status quo. Questioning authority".

In the absence of evidence, opinion is indistinguishable from prejudice.

RIP an inspiration; A true Folk's Guy

Scalar /g is an iterator (setting pos as a side effect).

Why should an iterator accumulate anything?

you may be interested in global regex returning a list of arrays?

Cheers Rolf

Note that if you had done the match in list context, again only one key in %+ would be set ("d" in that case). After all, you wouldn't expect $16 to be set either, would you?

Okay. What is going on here then?

perl -wE"'abcde' =~ m[(?<a>a)|(?<b>b)|(?<c>c)|(?<d>d)|(?<e>e)]g and do
+{ say for keys %+ }"
a

perl -wE"()='abcde' =~ m[(?<a>a)|(?<b>b)|(?<c>c)|(?<d>d)|(?<e>e)]g and
+ do{ say for keys %+ }"
e

perl -wE"my@a = 'abcde' =~ m[(?<a>a)|(?<b>b)|(?<c>c)|(?<d>d)|(?<e>e)]g
+ and do{ say for keys %+ }"
e

perl -wE"my@a = 'abcde' =~ m[(?<a>a)|(?<b>b)|(?<c>c)]g and do{ say for
+ keys %+ }"
## big fat nothing?
[download]

---

Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error.

"Science is about questioning the status quo. Questioning authority".

In the absence of evidence, opinion is indistinguishable from prejudice.

RIP an inspiration; A true Folk's Guy

perl -wE"'abcde' =~ m[(?<a>a)|(?<b>b)|(?<c>c)|(?<d>d)|(?<e>e)]g and do
+{ say for keys %+ }"
a
[download]

/g in scalar context matches once, %+ contains the first value, a.

perl -wE"()='abcde' =~ m[(?<a>a)|(?<b>b)|(?<c>c)|(?<d>d)|(?<e>e)]g and
+ do{ say for keys %+ }"
e
[download]

/g in list context matches all locations, and after having matched all, %+ contains the result of the last match, e.

perl -wE"my@a = 'abcde' =~ m[(?<a>a)|(?<b>b)|(?<c>c)|(?<d>d)|(?<e>e)]g
+ and do{ say for keys %+ }"
e
[download]

Same as above, except that you store the intermediate values.

perl -wE"my@a = 'abcde' =~ m[(?<a>a)|(?<b>b)|(?<c>c)]g and do{ say for
+ keys %+ }"
## big fat nothing?
[download]

You have removed [de] from the matching class. %+ is a global variable and it (seems it) is cleared by the last failing match.

It might or might not be nice to have %+ retain the value(s) of the last successful match, but maybe you can work around this by padding your RE with .*$re.* to convert it into a truely floating RE.