RT #130907: Fix the Unicode Bug in split " "

lib/feature.pm

@@ -5,7 +5,7 @@
 
 package feature;
 
-our $VERSION = '1.47';
+our $VERSION = '1.48';
 
 our %feature = (
     fc              => 'feature_fc',
@@ -175,8 +175,9 @@ C<use feature 'unicode_strings'> subpragma is B<strongly> recommended.
 
 This feature is available starting with Perl 5.12; was almost fully
 implemented in Perl 5.14; and extended in Perl 5.16 to cover C<quotemeta>;
-and extended further in Perl 5.26 to cover L<the range
-operator|perlop/Range Operators>.
+was extended further in Perl 5.26 to cover L<the range
+operator|perlop/Range Operators>; and was extended again in Perl 5.28 to
+cover L<special-cased whitespace splitting|perlfunc/split>.
 
 =head2 The 'unicode_eval' and 'evalbytes' features
 

pod/perldelta.pod

@@ -343,6 +343,15 @@ expression had no named captures.  The same applies to access to any
 hash tied with L<Tie::Hash::NamedCapture> and C<< all => 1 >>. [perl
 #130822]
 
+=item *
+
+C<split ' '> now handles the argument being split correctly when in the
+scope of the L<< C<unicode_strings>|feature/"The 'unicode_strings' feature"
+>> feature. Previously, when a string using the single-byte internal
+representation contained characters that are whitespace by Unicode rules but
+not by ASCII rules, it treated those characters as part of fields rather
+than as field separators. This resolves [perl #130907].
+
 =back
 
 =head1 Known Problems

pod/perlfunc.pod

@@ -7601,6 +7601,14 @@ special case was restricted to the use of a plain S<C<" ">> as the
 pattern argument to split; in Perl 5.18.0 and later this special case is
 triggered by any expression which evaluates to the simple string S<C<" ">>.
 
+As of Perl 5.28, this special-cased whitespace splitting works as expected in
+the scope of L<< S<C<"use feature 'unicode_strings">>|feature/The
+'unicode_strings' feature >>. In previous versions, and outside the scope of
+that feature, it exhibits L<perlunicode/The "Unicode Bug">: characters that are
+whitespace according to Unicode rules but not according to ASCII rules can be
+treated as part of fields rather than as field separators, depending on the
+string's internal encoding.
+
 If omitted, PATTERN defaults to a single space, S<C<" ">>, triggering
 the previously described I<awk> emulation.
 

pod/perlunicode.pod

@@ -1824,6 +1824,17 @@ outside its scope, it could produce strings whose length in characters
 exceeded that of the right-hand side, where the right-hand side took up more
 bytes than the correct range endpoint.
 
+=item *
+
+In L<< C<split>'s special-case whitespace splitting|perlfunc/split >>.
+
+Starting in Perl 5.28.0, the C<split> function with a pattern specified as
+a string containing a single space handles whitespace characters consistently
+within the scope of of C<unicode_strings>. Prior to that, or outside its scope,
+characters that are whitespace according to Unicode rules but not according to
+ASCII rules were treated as field contents rather than field separators when
+they appear in byte-encoded strings.
+
 =back
 
 You can see from the above that the effect of C<unicode_strings>

pod/perluniintro.pod

@@ -151,11 +151,12 @@ serious Unicode work.  The maintenance release 5.6.1 fixed many of the
 problems of the initial Unicode implementation, but for example
 regular expressions still do not work with Unicode in 5.6.1.
 Perl v5.14.0 is the first release where Unicode support is
-(almost) seamlessly integrable without some gotchas. (There are two
+(almost) seamlessly integrable without some gotchas. (There are a few
 exceptions. Firstly, some differences in L<quotemeta|perlfunc/quotemeta>
 were fixed starting in Perl 5.16.0. Secondly, some differences in
 L<the range operator|perlop/Range Operators> were fixed starting in
-Perl 5.26.0.)
+Perl 5.26.0. Thirdly, some differences in L<split|perlfunc/split> were fixed
+started in Perl 5.28.0.)
 
 To enable this
 seamless support, you should C<use feature 'unicode_strings'> (which is

pp.c

@@ -5716,6 +5716,7 @@ PP(pp_split)
     STRLEN len;
     const char *s = SvPV_const(sv, len);
     const bool do_utf8 = DO_UTF8(sv);
+    const bool in_uni_8_bit = IN_UNI_8_BIT;
     const char *strend = s + len;
     PMOP *pm = cPMOPx(PL_op);
     REGEXP *rx;
@@ -5802,6 +5803,10 @@ PP(pp_split)
 	    while (s < strend && isSPACE_LC(*s))
 		s++;
 	}
+        else if (in_uni_8_bit) {
+            while (s < strend && isSPACE_L1(*s))
+                s++;
+        }
 	else {
 	    while (s < strend && isSPACE(*s))
 		s++;
@@ -5833,6 +5838,10 @@ PP(pp_split)
             {
 	        while (m < strend && !isSPACE_LC(*m))
 		    ++m;
+            }
+            else if (in_uni_8_bit) {
+                while (m < strend && !isSPACE_L1(*m))
+                    ++m;
             } else {
                 while (m < strend && !isSPACE(*m))
                     ++m;
@@ -5867,6 +5876,10 @@ PP(pp_split)
             {
 	        while (s < strend && isSPACE_LC(*s))
 		    ++s;
+            }
+            else if (in_uni_8_bit) {
+                while (s < strend && isSPACE_L1(*s))
+                    ++s;
             } else {
                 while (s < strend && isSPACE(*s))
                     ++s;

regen/feature.pl

@@ -367,7 +367,7 @@ sub longest {
 __END__
 package feature;
 
-our $VERSION = '1.47';
+our $VERSION = '1.48';
 
 FEATURES
 
@@ -485,8 +485,9 @@ =head2 The 'unicode_strings' feature
 
 This feature is available starting with Perl 5.12; was almost fully
 implemented in Perl 5.14; and extended in Perl 5.16 to cover C<quotemeta>;
-and extended further in Perl 5.26 to cover L<the range
-operator|perlop/Range Operators>.
+was extended further in Perl 5.26 to cover L<the range
+operator|perlop/Range Operators>; and was extended again in Perl 5.28 to
+cover L<special-cased whitespace splitting|perlfunc/split>.
 
 =head2 The 'unicode_eval' and 'evalbytes' features
 

t/op/split.t

@@ -7,7 +7,7 @@ BEGIN {
     set_up_inc('../lib');
 }
 
-plan tests => 163;
+plan tests => 172;
 
 $FS = ':';
 
@@ -480,6 +480,24 @@ is($cnt, scalar(@ary));
         qq{split(\$cond ? qr/ / : " ", "$exp") behaves as expected over repeated similar patterns};
 }
 
+SKIP: {
+    # RT #130907: unicode_strings feature doesn't work with split ' '
+
+    my ($sp) = grep /\s/u, map chr, reverse 128 .. 255 # prefer \xA0 over \x85
+        or skip 'no unicode whitespace found in high-8-bit range', 9;
+
+    for (["$sp$sp. /", "leading unicode whitespace"],
+         [".$sp$sp/",  "unicode whitespace separator"],
+         [". /$sp$sp", "trailing unicode whitespace"]) {
+        my ($str, $desc) = @$_;
+        use feature "unicode_strings";
+        my @got = split " ", $str;
+        is @got, 2, "whitespace split: $desc: field count";
+        is $got[0], '.', "whitespace split: $desc: field 0";
+        is $got[1], '/', "whitespace split: $desc: field 1";
+    }
+}
+
 {
     # 'RT #116086: split "\x20" does not work as documented';
     my @results;