package feature;
-our $VERSION = '1.18';
+our $VERSION = '1.23';
# (feature name) => (internal name, used in %^H)
my %feature = (
+ say => 'feature_say',
+ state => 'feature_state',
switch => 'feature_switch',
- say => "feature_say",
- state => "feature_state",
- unicode_strings => "feature_unicode",
+ evalbytes => 'feature_evalbytes',
+ unicode_eval => 'feature_unieval',
+ unicode_strings => 'feature_unicode',
);
# This gets set (for now) in $^H as well as in %^H,
# NB. the latest bundle must be loaded by the -E switch (see toke.c)
my %feature_bundle = (
- "5.10" => [qw(switch say state)],
- "5.11" => [qw(switch say state unicode_strings)],
- "5.12" => [qw(switch say state unicode_strings)],
- "5.13" => [qw(switch say state unicode_strings)],
+ "5.10" => [qw(say state switch)],
+ "5.11" => [qw(say state switch unicode_strings)],
+ "5.12" => [qw(say state switch unicode_strings)],
+ "5.13" => [qw(say state switch unicode_strings)],
+ "5.14" => [qw(say state switch unicode_strings)],
+ "5.15" => [qw(say state switch unicode_strings unicode_eval
+ evalbytes)],
+ "5.16" => [qw(say state switch unicode_strings unicode_eval
+ evalbytes)],
);
# special case
=head1 SYNOPSIS
- use feature qw(switch say);
+ use feature qw(say switch);
given ($foo) {
when (1) { say "\$foo == 1" }
when ([2,3]) { say "\$foo == 2 || \$foo == 3" }
some existing programs. This pragma provides a way to minimize that
risk. New syntactic constructs, or new semantic meanings to older
constructs, can be enabled by C<use feature 'foo'>, and will be parsed
-only when the appropriate feature pragma is in scope.
+only when the appropriate feature pragma is in scope. (Nevertheless, the
+C<CORE::> prefix provides access to all Perl keywords, regardless of this
+pragma.)
=head2 Lexical effect
C<no feature> with no features specified will turn off all features.
-=head2 The 'switch' feature
-
-C<use feature 'switch'> tells the compiler to enable the Perl 6
-given/when construct.
-
-See L<perlsyn/"Switch statements"> for details.
-
=head2 The 'say' feature
C<use feature 'say'> tells the compiler to enable the Perl 6
See L<perlsub/"Persistent Private Variables"> for details.
+=head2 The 'switch' feature
+
+C<use feature 'switch'> tells the compiler to enable the Perl 6
+given/when construct.
+
+See L<perlsyn/"Switch statements"> for details.
+
=head2 the 'unicode_strings' feature
-C<use feature 'unicode_strings'> tells the compiler to treat
-all strings outside of C<use locale> and C<use bytes> as Unicode. It is
-available starting with Perl 5.11.3, but is not fully implemented.
+C<use feature 'unicode_strings'> tells the compiler to use Unicode semantics
+in all string operations executed within its scope (unless they are also
+within the scope of either C<use locale> or C<use bytes>). The same applies
+to all regular expressions compiled within the scope, even if executed outside
+it.
+
+C<no feature 'unicode_strings'> tells the compiler to use the traditional
+Perl semantics wherein the native character set semantics is used unless it is
+clear to Perl that Unicode is desired. This can lead to some surprises
+when the behavior suddenly changes. (See
+L<perlunicode/The "Unicode Bug"> for details.) For this reason, if you are
+potentially using Unicode in your program, the
+C<use feature 'unicode_strings'> subpragma is B<strongly> recommended.
+
+This subpragma is available starting with Perl 5.11.3, but was not fully
+implemented until 5.13.8.
+
+=head2 the 'unicode_eval' and 'evalbytes' features
+
+Under the C<unicode_eval> feature, Perl's C<eval> function, when passed a
+string, will evaluate it as a string of characters, ignoring any
+C<use utf8> declarations. C<use utf8> exists to declare the encoding of
+the script, which only makes sense for a stream of bytes, not a string of
+characters. Source filters are forbidden, as they also really only make
+sense on strings of bytes. Any attempt to activate a source filter will
+result in an error.
+
+The C<evalbytes> feature enables the C<evalbytes> keyword, which evaluates
+the argument passed to it as a string of bytes. It dies if the string
+contains any characters outside the 8-bit range. Source filters work
+within C<evalbytes>: they apply to the contents of the string being
+evaluated.
+
+Together, these two features are intended to replace the historical C<eval>
+function, which has (at least) two bugs in it, that cannot easily be fixed
+without breaking existing programs:
+
+=over
-See L<perlunicode/The "Unicode Bug"> for details.
+=item *
+
+C<eval> behaves differently depending on the internal encoding of the
+string, sometimes treating its argument as a string of bytes, and sometimes
+as a string of characters.
+
+=item *
+
+Source filters activated within C<eval> leak out into whichever I<file>
+scope is currently being compiled. To give an example with the CPAN module
+L<Semi::Semicolons>:
+
+ BEGIN { eval "use Semi::Semicolons; # not filtered here " }
+ # filtered here!
+
+C<evalbytes> fixes that to work the way one would expect:
+
+ use feature "evalbytes";
+ BEGIN { evalbytes "use Semi::Semicolons; # filtered " }
+ # not filtered
+
+=back
+
+These two features are available starting with Perl 5.16.
=head1 FEATURE BUNDLES
It's possible to load a whole slew of features in one go, using
a I<feature bundle>. The name of a feature bundle is prefixed with
a colon, to distinguish it from an actual feature. At present, the
-only feature bundle is C<use feature ":5.10"> which is equivalent
-to C<use feature qw(switch say state)>.
+only feature bundles correspond to Perl releases, e.g. C<use feature
+":5.10"> which is equivalent to C<use feature qw(switch say state)>.
+
+By convention, the feature bundle for any given Perl release includes
+the features of previous releases, down to and including 5.10, the
+first official release to provide this facility. Since Perl 5.12
+only provides one new feature, C<unicode_strings>, and Perl 5.14
+provides none, C<use feature ":5.14"> is equivalent to C<use feature
+qw(switch say state unicode_strings)>.
-Specifying sub-versions such as the C<0> in C<5.10.0> in feature bundles has
+Specifying sub-versions such as the C<0> in C<5.14.0> in feature bundles has
no effect: feature bundles are guaranteed to be the same for all sub-versions.
+Note that instead of using release-based feature bundles it is usually
+better, and shorter, to use implicit loading as described below.
+
=head1 IMPLICIT LOADING
There are two ways to load the C<feature> pragma implicitly :