5 # (feature name) => (internal name, used in %^H)
7 switch => 'feature_switch',
9 state => "feature_state",
10 unicode_strings => "feature_unicode",
13 # This gets set (for now) in $^H as well as in %^H,
14 # for runtime speed of the uc/lc/ucfirst/lcfirst functions.
15 our $hint_uni8bit = 0x00000800;
17 # NB. the latest bundle must be loaded by the -E switch (see toke.c)
19 my %feature_bundle = (
20 "5.10" => [qw(switch say state)],
21 "5.11" => [qw(switch say state unicode_strings)],
25 $feature_bundle{"5.9.5"} = $feature_bundle{"5.10"};
28 # - think about versioned features (use feature switch => 2)
32 feature - Perl pragma to enable new syntactic features
36 use feature qw(switch say);
38 when (1) { say "\$foo == 1" }
39 when ([2,3]) { say "\$foo == 2 || \$foo == 3" }
40 when (/^a[bc]d$/) { say "\$foo eq 'abd' || \$foo eq 'acd'" }
41 when ($_ > 100) { say "\$foo > 100" }
42 default { say "None of the above" }
45 use feature ':5.10'; # loads all features available in perl 5.10
49 It is usually impossible to add new syntax to Perl without breaking
50 some existing programs. This pragma provides a way to minimize that
51 risk. New syntactic constructs, or new semantic meanings to older
52 constructs, can be enabled by C<use feature 'foo'>, and will be parsed
53 only when the appropriate feature pragma is in scope.
57 Like other pragmas (C<use strict>, for example), features have a lexical
58 effect. C<use feature qw(foo)> will only make the feature "foo" available
59 from that point to the end of the enclosing block.
63 say "say is available here";
65 print "But not here.\n";
69 Features can also be turned off by using C<no feature "foo">. This too
73 say "say is available here";
76 print "But not here.\n";
78 say "Yet it is here.";
80 C<no feature> with no features specified will turn off all features.
82 =head2 The 'switch' feature
84 C<use feature 'switch'> tells the compiler to enable the Perl 6
87 See L<perlsyn/"Switch statements"> for details.
89 =head2 The 'say' feature
91 C<use feature 'say'> tells the compiler to enable the Perl 6
94 See L<perlfunc/say> for details.
96 =head2 the 'state' feature
98 C<use feature 'state'> tells the compiler to enable C<state>
101 See L<perlsub/"Persistent Private Variables"> for details.
103 =head2 the 'unicode_strings' feature
105 C<use feature 'unicode_strings'> tells the compiler to treat
106 strings with codepoints larger than 128 as Unicode. It is available
107 starting with Perl 5.11.3.
111 This feature modifies the semantics for the 128 characters on ASCII
112 systems that have the 8th bit set. (See L</EBCDIC platforms> below for
113 EBCDIC systems.) By default, unless C<S<use locale>> is specified, or the
114 scalar containing such a character is known by Perl to be encoded in UTF8,
115 the semantics are essentially that the characters have an ordinal number,
116 and that's it. They are caseless, and aren't anything: they're not
117 controls, not letters, not punctuation, ..., not anything.
119 This behavior stems from when Perl did not support Unicode, and ASCII was the
120 only known character set outside of C<S<use locale>>. In order to not
121 possibly break pre-Unicode programs, these characters have retained their old
122 non-meanings, except when it is clear to Perl that Unicode is what is meant,
123 for example by calling utf8::upgrade() on a scalar, or if the scalar also
124 contains characters that are only available in Unicode. Then these 128
125 characters take on their Unicode meanings.
127 The problem with this behavior is that a scalar that encodes these characters
128 has a different meaning depending on if it is stored as utf8 or not.
129 In general, the internal storage method should not affect the
132 The behavior is known to have effects on these areas:
138 Changing the case of a scalar, that is, using C<uc()>, C<ucfirst()>, C<lc()>,
139 and C<lcfirst()>, or C<\L>, C<\U>, C<\u> and C<\l> in regular expression
144 Using caseless (C</i>) regular expression matching
148 Matching a number of properties in regular expressions, such as C<\w>
152 User-defined case change mappings. You can create a C<ToUpper()> function, for
153 example, which overrides Perl's built-in case mappings. The scalar must be
154 encoded in utf8 for your function to actually be invoked.
158 B<This lack of semantics for these characters is currently the default,>
159 outside of C<use locale>. See below for EBCDIC.
161 To turn on B<case changing semantics only> for these characters, use
162 C<use feature "unicode_strings">.
164 The other old (legacy) behaviors regarding these characters are currently
165 unaffected by this pragma.
167 =head4 EBCDIC platforms
169 On EBCDIC platforms, the situation is somewhat different. The legacy
170 semantics are whatever the underlying semantics of the native C language
171 library are. Each of the three EBCDIC encodings currently known by Perl is an
172 isomorph of the Latin-1 character set. That means every character in Latin-1
173 has a corresponding EBCDIC equivalent, and vice-versa. Specifying C<S<no
174 legacy>> currently makes sure that all EBCDIC characters have the same
175 B<casing only> semantics as their corresponding Latin-1 characters.
177 =head1 FEATURE BUNDLES
179 It's possible to load a whole slew of features in one go, using
180 a I<feature bundle>. The name of a feature bundle is prefixed with
181 a colon, to distinguish it from an actual feature. At present, the
182 only feature bundle is C<use feature ":5.10"> which is equivalent
183 to C<use feature qw(switch say state)>.
185 Specifying sub-versions such as the C<0> in C<5.10.0> in feature bundles has
186 no effect: feature bundles are guaranteed to be the same for all sub-versions.
188 =head1 IMPLICIT LOADING
190 There are two ways to load the C<feature> pragma implicitly :
196 By using the C<-E> switch on the command-line instead of C<-e>. It enables
197 all available features in the main compilation unit (that is, the one-liner.)
201 By requiring explicitly a minimal Perl version number for your program, with
202 the C<use VERSION> construct, and when the version is higher than or equal to
211 and so on. Note how the trailing sub-version is automatically stripped from the
214 But to avoid portability warnings (see L<perlfunc/use>), you may prefer:
218 with the same effect.
227 croak("No features specified");
230 my $name = shift(@_);
231 if (substr($name, 0, 1) eq ":") {
232 my $v = substr($name, 1);
233 if (!exists $feature_bundle{$v}) {
234 $v =~ s/^([0-9]+)\.([0-9]+).[0-9]+$/$1.$2/;
235 if (!exists $feature_bundle{$v}) {
236 unknown_feature_bundle(substr($name, 1));
239 unshift @_, @{$feature_bundle{$v}};
242 if (!exists $feature{$name}) {
243 unknown_feature($name);
245 $^H{$feature{$name}} = 1;
246 $^H |= $hint_uni8bit if $name eq 'unicode_strings';
253 # A bare C<no feature> should disable *all* features
255 delete @^H{ values(%feature) };
256 $^H &= ~ $hint_uni8bit;
262 if (substr($name, 0, 1) eq ":") {
263 my $v = substr($name, 1);
264 if (!exists $feature_bundle{$v}) {
265 $v =~ s/^([0-9]+)\.([0-9]+).[0-9]+$/$1.$2/;
266 if (!exists $feature_bundle{$v}) {
267 unknown_feature_bundle(substr($name, 1));
270 unshift @_, @{$feature_bundle{$v}};
273 if (!exists($feature{$name})) {
274 unknown_feature($name);
277 delete $^H{$feature{$name}};
278 $^H &= ~ $hint_uni8bit if $name eq 'unicode_strings';
283 sub unknown_feature {
285 croak(sprintf('Feature "%s" is not supported by Perl %vd',
289 sub unknown_feature_bundle {
291 croak(sprintf('Feature bundle "%s" is not supported by Perl %vd',