fix: remove use locale from Filters.pm for Unicode correctness (GH #137)

lib/Template/Filters.pm

@@ -21,7 +21,6 @@ package Template::Filters;
 
 use strict;
 use warnings;
-use locale;
 use base 'Template::Base';
 use Template::Constants;
 use Scalar::Util 'blessed';
@@ -52,10 +51,10 @@ our $FILTERS = {
     'xml'             => \&xml_filter,
     'uri'             => \&uri_filter,
     'url'             => \&url_filter,
-    'upper'           => sub { uc $_[0] },
-    'lower'           => sub { lc $_[0] },
-    'ucfirst'         => sub { ucfirst $_[0] },
-    'lcfirst'         => sub { lcfirst $_[0] },
+    'upper'           => sub { utf8::upgrade(my $s = $_[0]); uc $s },
+    'lower'           => sub { utf8::upgrade(my $s = $_[0]); lc $s },
+    'ucfirst'         => sub { utf8::upgrade(my $s = $_[0]); ucfirst $s },
+    'lcfirst'         => sub { utf8::upgrade(my $s = $_[0]); lcfirst $s },
     'stderr'          => sub { print STDERR @_; return '' },
     'trim'            => sub { for ($_[0]) { s/^\s+//; s/\s+$// }; $_[0] },
     'null'            => sub { return '' },

t/unicode_filters.t

@@ -0,0 +1,117 @@
+#============================================================= -*-perl-*-
+#
+# t/unicode_filters.t
+#
+# Test that case-changing filters (upper, lower, ucfirst, lcfirst) work
+# correctly with Unicode characters and don't produce wide character
+# warnings.  See GH #137.
+#
+#========================================================================
+
+use strict;
+use warnings;
+use lib qw( ./lib ../lib );
+use Test::More tests => 10;
+use Template;
+
+my $tt = Template->new();
+
+# Latin-1 characters (e.g. accented letters)
+{
+    my $out;
+    $tt->process(
+        \"[% FILTER upper %][% s %][% END %]",
+        { s => "caf\x{e9}" },
+        \$out
+    );
+    chomp $out;
+    is($out, "CAF\x{c9}", 'upper filter handles Latin-1 e-acute');
+}
+
+{
+    my $out;
+    $tt->process(
+        \"[% FILTER lower %][% s %][% END %]",
+        { s => "CAF\x{c9}" },
+        \$out
+    );
+    chomp $out;
+    is($out, "caf\x{e9}", 'lower filter handles Latin-1 E-acute');
+}
+
+{
+    my $out;
+    $tt->process(
+        \"[% s | ucfirst %]",
+        { s => "\x{e9}cole" },
+        \$out
+    );
+    chomp $out;
+    is($out, "\x{c9}cole", 'ucfirst handles Latin-1 e-acute');
+}
+
+{
+    my $out;
+    $tt->process(
+        \"[% s | lcfirst %]",
+        { s => "\x{c9}COLE" },
+        \$out
+    );
+    chomp $out;
+    is($out, "\x{e9}COLE", 'lcfirst handles Latin-1 E-acute');
+}
+
+# Wide characters (codepoints > 0xFF) — these triggered the warning in GH #137
+{
+    my @warnings;
+    local $SIG{__WARN__} = sub { push @warnings, $_[0] };
+
+    my $out;
+    # Cyrillic де (U+0434 U+0435)
+    $tt->process(
+        \"[% FILTER upper %][% s %][% END %]",
+        { s => "\x{434}\x{435}" },
+        \$out
+    );
+    chomp $out;
+    is($out, "\x{414}\x{415}", 'upper filter handles Cyrillic');
+    is(scalar @warnings, 0, 'no wide character warnings from upper filter');
+}
+
+{
+    my @warnings;
+    local $SIG{__WARN__} = sub { push @warnings, $_[0] };
+
+    my $out;
+    $tt->process(
+        \"[% FILTER lower %][% s %][% END %]",
+        { s => "\x{414}\x{415}" },
+        \$out
+    );
+    chomp $out;
+    is($out, "\x{434}\x{435}", 'lower filter handles Cyrillic');
+    is(scalar @warnings, 0, 'no wide character warnings from lower filter');
+}
+
+# ASCII still works as before
+{
+    my $out;
+    $tt->process(
+        \"[% FILTER upper %]hello world[% END %]",
+        {},
+        \$out
+    );
+    chomp $out;
+    is($out, 'HELLO WORLD', 'upper filter still works for ASCII');
+}
+
+{
+    my $out;
+    $tt->process(
+        \"[% FILTER lower %]HELLO WORLD[% END %]",
+        {},
+        \$out
+    );
+    chomp $out;
+    is($out, 'hello world', 'lower filter still works for ASCII');
+}