Fixing utf8 errors

Rodster001 has asked for the wisdom of the Perl Monks concerning the following question:

Hi, I have this little routine that works great at finding and removing bad encoding (for my purposes):

sub _checkit {

        my $txt = shift;

        use warnings FATAL => 'utf8';

        1 until eval {
                print "$txt\n";
                1;

        } or do {

                my ($charcode) = $@ =~ /U\+(\S+)/ or die $@;
                $txt =~ s/\x{$charcode}//g;
                0; # Try again!
        };

        return ($txt);
}
[download]

I have two questions about this:

I think I understand, but I'd like a good explanation: Why do I need to print to generate/catch the error?
Is there another way to do this (without print to stdout) so that it doesn't generate all the output? Or should I just redirect stdout to /dev/null?

Thanks!

Comment on Fixing utf8 errors Download Code

Replies are listed 'Best First'.
Re: Fixing utf8 errors by graff (Chancellor) on Oct 31, 2015 at 04:07 UTC
Can you define or describe what "bad encoding (for my purposes)" refers to? Can you provide or describe a few samples of "bad encoding" and "ok encoding"? Is the string that is passed to "_checkit" supposed to be a (perl-internal) UTF8 string? Has STDOUT been set to use UTF8 encoding before "_checkit" is called? When the error message from a failed eval matches `/U\+(\S+)/`, what is the full text of the error message? (Do you get error messages that don't match that regex, and if so, what are they?) Please be explicit. The point is that I don't understand what sort of error you're "catching". There probably is a better way to catch it, but since I don't know what you're dealing with, I don't know what to suggest. A few lines of code that reads a few lines of relevant __DATA__ and calls "_checkit" would answer most or all of my questions.	[reply] [d/l]
Re^2: Fixing utf8 errors by Rodster001 (Pilgrim) on Oct 31, 2015 at 16:07 UTC
Here is an example: `print Dumper $text; $VAR1 = { 'string' => "Text\x{daed}", }; print $text->{'string'}; 'Unicode surrogate U+DAED is illegal in UTF-8 at line 5.` [download]	[reply] [d/l]
Re^3: Fixing utf8 errors by graff (Chancellor) on Nov 01, 2015 at 01:26 UTC
It's possible for UTF8-encoded text to contain code points in the range U+D800 - U+DFFF (Surrogate Pair Area), but this is always a sign that something bad was done to the data when it was converted into UTF8 encoding. I don't recall offhand what particular mistake(s) in transcoding would cause these code points to show up in UTF8 data, but there's a chance that your data originated in some 16-bit (fixed-width) encoding that was incorrectly converted to UTF8, and there may be better ways of "fixing" the problem than just deleting these odd characters. Still, if your experience with the data in question indicates that deletion is the "right thing" to do, it's easy to detect and "fix" the problem: `# Assuming $string is utf-8 endcoded: if ( my @badch = ( $string =~ /[\x{d800}-\x{dfff}]/g )) { warn sprintf( "%d surrogate code point(s) found\n", scalar @badch +); $string =~ tr/\x{d800}-\x{dfff}//d; }` [download] (Update - forgot to mention: you can read more about surrogate-pair characters in perlunicode. It may be that your UTF8 data started out as UTF16, but was incorrectly treated as if it were UCS2 when it was converted to UTF8. Converting to UTF8 in the correct way would yield different results.)	[reply] [d/l]
Re: Fixing utf8 errors by Anonymous Monk on Oct 30, 2015 at 18:59 UTC
It's because printing wide characters causes Perl to try to encode them into bytes (maybe through an :encoding(...) IOLayer). When you pass a character which cannot be represented in target encoding, an exception is raised. You can also try encode with different error handling modes (your current one is similar to FB_CROAK). Did I guess the error message ($@) you receive correctly? If not, please provide an example.	[reply]