=pod treated as syntax

[dk]

Hello,

I'm not 100% sure if this is a bug, but this looks funny:

$ cat a.pl
sub x { qw(y)

=pod

=cut
}

$ perl -c a.pl
Can't modify constant item in scalar assignment at a.pl line 6, near "cut
}"
/home/dk/a.pl had compilation errors.

This is very ancient and I traced this back to e438e50 from 2005, and is basically this diff:

diff --git a/toke.c b/toke.c
index afa32a54a5..9d566385ef 100644
--- a/toke.c
+++ b/toke.c
@@ -4899,6 +4899,7 @@ Perl_yylex(pTHX)
            s = scan_str(s,FALSE,FALSE);
            if (!s)
                missingterm((char*)0);
+           PL_expect = XOPERATOR;
            force_next(')');
            if (SvCUR(PL_lex_stuff)) {
                OP *words = Nullop;

I've applied the modified version of this to the toke.c from blead, like this:

diff --git a/toke.c b/toke.c
index 650284657c..f4a7598b29 100644
--- a/toke.c
+++ b/toke.c
@@ -5813,7 +5813,7 @@ yyl_qw(pTHX_ char *s, STRLEN len)
     s = S_scan_terminated(aTHX_ s, 0);
 
     COPLINE_SET_FROM_MULTI_END;
-    PL_expect = XOPERATOR;
+//    PL_expect = XOPERATOR;
     if (SvCUR(PL_lex_stuff)) {
         int warned_comma = !ckWARN(WARN_QW);
         int warned_comment = warned_comma;

and perl passed all tests.

I don't think I have the deep knowledge to decide whether that is a bug, a feature, or part of something bigger, but here's all the info you need if you would address that interesting behavior.

Regards,
Dmitry Karasik

[ilmari]

POD can't be used in the middle of a statement, only between statements, and because there's no semicolon after the qw(y), you're still inside that statement. This parses (due to the lack of use strict) as qw(y) = "pod" = "cut", and the error message is correct, since you can't assign to the string literal "pod".

Your change "fixes" this particular case, but only if the last expression in the statement the would-be POD is in was qw(...), not for any other expression.

The fact that this doesn't break any tests just means nothing tries to assign a bareword to a qw() expression in a manner that looks like POD (which would also fail).

[ilmari]

For reference, this is documented in perlpod:

You can place a Pod statement where perl expects the beginning of a new statement, but not within a statement, as that would result in an error.

[dk]

Fair enough, thanks for the quick clarification

[book]

My favourite use of POD for both documentation and data was shown to me by @Abigail:

# the data will show up in the documentation

my $raw_data = << '=cut';
=pod

...

=cut

# remove the POD syntax from the string to get the actual data

[Grinnz]

That looks somewhat nicer than the other hack I've seen (to show a block of code as a verbatim pod paragraph): https://metacpan.org/release/SPROUT/HTML-DOM-0.058/source/lib/HTML/DOM/Interface.pm#L91

[karenetheridge]

That looks somewhat nicer than the other hack I've seen

THIS! This is why people say Perl is nothing but line noise!!!!!

[book]

That looks somewhat nicer than the other hack I've seen

THIS! This is why people say Perl is nothing but line noise!!!!!

It took me literal minutes to understand what was going on there. Kept looking for a lone r and failed to see what the ; was for.