qw(cleanup_eval_fail),
qw(hash_slice),
],
- utf8 => [qw(encode_utf8_structure)],
+ utf8 => [qw(encode_utf8_structure encode_utf8_safely),
+ qw(convert_to_utf8)],
date => [qw(secs_to_english)],
quit => [qw(quit)],
lock => [qw(filelock unfilelock lockpid)],
use Debbugs::MIME qw(decode_rfc1522);
use Mail::Address;
use Cwd qw(cwd);
-use Encode qw(encode_utf8 is_utf8);
+use Encode qw(encode_utf8 is_utf8 decode);
+use Text::Iconv;
use Storable qw(dclone);
use Params::Validate qw(validate_with :types);
$type //= 'address';
my $fh = IO::File->new($fn,'r') or
die "Unable to open $fn for reading: $!";
+ binmode($fh,':encoding(UTF-8)');
while (<$fh>) {
chomp;
next unless m/^(\S+)\s+(\S.*\S)\s*$/;
push @ret,$_;
}
else {
- push @ret,__encode_utf8($_);
+ push @ret,encode_utf8_safely($_);
}
}
--$depth;
return @ret;
}
-sub __encode_utf8 {
+=head2 encode_utf8_safely
+
+ $octets = encode_utf8_safely($string);
+
+Given a $string, returns the octet equivalent of $string if $string is
+in perl's internal encoding; otherwise returns $string.
+
+Silently returns REFs without encoding them. [If you want to deeply
+encode REFs, see encode_utf8_structure.]
+
+=cut
+
+
+sub encode_utf8_safely{
my @ret;
for my $r (@_) {
- if (not ref($r) and is_utf8($r)) {
+ if (not ref($r) and is_utf8($r)) {
$r = encode_utf8($r);
}
push @ret,$r;
}
- return @ret;
+ return wantarray ? @ret : (length @_ > 1 ? @ret : $_[0]);
+}
+
+=head2 convert_to_utf8
+
+ $utf8 = convert_to_utf8("text","charset");
+
+=cut
+
+our %iconv_converters;
+
+sub convert_to_utf8 {
+ my ($data,$charset) = @_;
+ if (is_utf8($data)) {
+ return encode_utf8($data);
+ }
+ $charset = uc($charset);
+ if (not defined $iconv_converters{$charset}) {
+ eval {
+ $iconv_converters{$charset} = Text::Iconv->new($charset,"UTF-8") or
+ die "Unable to create converter for '$charset'";
+ };
+ if ($@) {
+ warn $@;
+ # We weren't able to create the converter, so use Encode
+ # instead
+ return __fallback_convert_to_utf8($data,$charset);
+ }
+ # It shouldn't be necessary when converting to UTF8, but lets
+ # allow for transliteration and silent discarding of broken
+ # sequences
+ eval {
+ $iconv_converters{$charset}->set_attr("transliterate");
+ $iconv_converters{$charset}->set_attr("discard_ilseq");
+ };
+ # This shouldn't fail on Debian systems; we're warning here
+ # just in case we've made a mistake above. This warning should
+ # probably be disabled on non-GNU libc systems.
+ warn $@ if $@;
+ }
+ if (not defined $iconv_converters{$charset}) {
+ warn "The converter for $charset wasn't created properly somehow!";
+ return __fallback_convert_to_utf8($data,$charset);
+ }
+ my $converted_data = $iconv_converters{$charset}->convert($data);
+ # if the conversion failed, retval will be undefined or perhaps
+ # -1.
+ if (not defined $iconv_converters{$charset}->retval() or
+ $iconv_converters{$charset}->retval() < 0
+ ) {
+ # Fallback to encode, which will probably also fail.
+ return __fallback_convert_to_utf8($data,$charset);
+ }
+ return $converted_data;
+}
+
+# Bug #61342 et al.
+# we're switching this to return UTF8 octets instead of perl's internal
+# encoding
+sub __Fallback_convert_to_utf8 {
+ my ($data, $charset) = @_;
+ # raw data just gets returned (that's the charset WordDecorder
+ # uses when it doesn't know what to do)
+ return $data if $charset eq 'raw';
+ if (not defined $charset and not is_utf8($data)) {
+ warn ("Undefined charset, and string '$data' is not in perl's internal encoding");
+ return $data;
+ }
+ # lets assume everything that doesn't have a charset is utf8
+ $charset //= 'utf8';
+ my $result;
+ eval {
+ $result = decode($charset,$data) unless is_utf8($data);
+ $result = encode_utf8($result);
+ };
+ if ($@) {
+ warn "Unable to decode charset; '$charset' and '$data': $@";
+ return $data;
+ }
+ return $result;
}