I think this might be close, I'm not exactly sure what output you want for those test cases. You might have to tweak the return statement slightly in
fix_xref. Usual warnings about parsing apply, is there optional white space, can the link text contain a '>', etc.
sub fix_xref {
my ($parts, $text, $ref, $action, $link_text) = @_;
$parts =~ s@(\d+(?:\.\d+)?(?:\([\w\d]+\))*)@<JL:ref$ref,"$action $
+1">s. $ref</JL>@g;
return qq@$parts $text <JL:ref$ref,"$action">$link_text</JL>@;
}
while (<DATA>) {
s@s\. (.+?) (of(?: the)) <JL:ref(\d+),"([^"]+)">([^<]+)</JL>@fix_x
+ref($1,$2,$3,$4,$5)@eg;
print;
print "\n";
}
__DATA__
s. 3 and 4 of the <JL:ref2,"Interpretation Act">Interpretation Act</JL
+>
s. 3 or 5 of the <JL:ref2,"Interpretation Act">Interpretation Act</JL>
+
s. 3, 4, 5, 6, 7 or 8 of the <JL:ref2,"Interpretation Act">Interpretat
+ion Act</JL>
gav^